Automation and AI adoption for the transport network: 2025 Omdia survey results

1 00:00:00,860 --> 00:00:03,120 Hello everyone, and thank you for attending 2 00:00:03,120 --> 00:00:06,280 today's webinar sponsored by Ciena and Cisco. 3 00:00:06,720 --> 00:00:08,540 Before we begin, I want to 4 00:00:08,540 --> 00:00:10,410 cover a few housekeeping items. 5 00:00:10,560 --> 00:00:12,640 On the left-hand side of your screen is the 6 00:00:12,640 --> 00:00:14,980 Q&A box. If you have any questions during 7 00:00:14,980 --> 00:00:17,040 the webcast, you can type your question into 8 00:00:17,040 --> 00:00:19,400 the Q&A box and submit them to our speakers. 9 00:00:19,480 --> 00:00:21,740 All questions will be saved, so if we don't 10 00:00:21,740 --> 00:00:24,080 get to answer you, we may follow up via email. 11 00:00:24,280 --> 00:00:26,540 At the bottom of your audience, consider multiple 12 00:00:26,540 --> 00:00:29,180 application widgets you can use. If you have 13 00:00:29,180 --> 00:00:31,320 any technical difficulties, please click on the 14 00:00:31,320 --> 00:00:33,640 help widget. Here you can find answers to common 15 00:00:33,640 --> 00:00:36,380 questions. A copy of today's slide deck is 16 00:00:36,380 --> 00:00:39,120 available for download in the resource list widget. 17 00:00:39,420 --> 00:00:41,600 Towards the end of today's presentation, 18 00:00:41,600 --> 00:00:42,900 we'll ask for your feedback. 19 00:00:42,900 --> 00:00:45,140 A survey is already open on your screen 20 00:00:45,140 --> 00:00:47,260 and will only take one minute to complete. 21 00:00:47,360 --> 00:00:49,400 Your feedback is extremely helpful. 22 00:00:49,400 --> 00:00:51,880 An on-demand version of the webcast will 23 00:00:51,880 --> 00:00:53,900 be available about one day after the event 24 00:00:53,900 --> 00:00:56,160 and can be accessed using the same audience 25 00:00:56,160 --> 00:00:58,260 link that was sent to you earlier today. 26 00:00:58,410 --> 00:01:01,300 I would now like to turn the event over to our 27 00:01:01,300 --> 00:01:04,620 Omdia Senior Principal Analyst, Optical Networks 28 00:01:04,620 --> 00:01:07,320 and Transport, Sterling Perrin. Sterling? 29 00:01:08,590 --> 00:01:10,920 Thanks, Barbara. Hello. Welcome, everybody, 30 00:01:10,920 --> 00:01:15,440 to my first webinar of the year. Welcome to 2026. And 31 00:01:15,440 --> 00:01:19,280 our webinar on Automation and AI Adoption for the 32 00:01:19,280 --> 00:01:24,860 transport network a look at our late 2025 omdia survey 33 00:01:24,860 --> 00:01:29,000 results on the topic i will be the host and moderator 34 00:01:29,000 --> 00:01:32,440 for today and i'll introduce our speakers in a moment 35 00:01:32,440 --> 00:01:38,060 um uh here's a flow of of what we'll be talking about 36 00:01:38,060 --> 00:01:40,820 over the next hour uh next slide i'll talk about the 37 00:01:40,820 --> 00:01:43,620 project itself there's a survey and other components to 38 00:01:43,620 --> 00:01:47,160 it um and then we'll get into the meat of of what the 39 00:01:47,160 --> 00:01:50,060 survey was about, which is looking at use cases and 40 00:01:50,060 --> 00:01:52,880 benefits around automation and AI, looking at artificial 41 00:01:52,880 --> 00:01:56,260 intelligence, specifically some detail there, and 42 00:01:56,260 --> 00:01:59,060 then a first for us on the transport network looking 43 00:01:59,060 --> 00:02:02,820 at agentic AI, so some interesting data points there. 44 00:02:03,460 --> 00:02:06,360 At the end, we'll have questions and answers. Barbara's 45 00:02:06,360 --> 00:02:11,440 explained the process for Q&A. Just ask as we go, 46 00:02:11,440 --> 00:02:15,400 as they come up and we'll gather them up and get to 47 00:02:15,400 --> 00:02:17,440 as many as we can at the end we hope to have about 10 48 00:02:17,440 --> 00:02:20,700 minutes or so at the end always interesting to get 49 00:02:20,700 --> 00:02:23,820 your thoughts comments and questions in our webinars 50 00:02:24,120 --> 00:02:27,980 the project itself is survey based it's automation 51 00:02:27,980 --> 00:02:30,840 and ai for transport it's actually the fourth 52 00:02:30,840 --> 00:02:34,660 year so some of you may have joined our past 53 00:02:34,660 --> 00:02:38,160 webinars on results this is from the november 2025 54 00:02:38,160 --> 00:02:41,860 survey we conducted we had 80 respondents globally 55 00:02:41,860 --> 00:02:45,240 i'll talk through the demographics in a moment 56 00:02:45,440 --> 00:02:50,300 of of who responded but we look at transport you know 57 00:02:50,300 --> 00:02:53,900 broadly defined ip optical and converged ip and optical 58 00:02:54,170 --> 00:02:57,020 as the years have gone by we've more moved more and 59 00:02:57,020 --> 00:03:00,540 more into ai as a component of automation or complementary 60 00:03:00,540 --> 00:03:03,340 technology and as i mentioned this year really 61 00:03:03,340 --> 00:03:07,580 getting into agentic ai AI and how operators are starting 62 00:03:07,580 --> 00:03:13,000 to see agentic AI factoring into their strategies. 63 00:03:13,700 --> 00:03:15,840 The other components will be a 64 00:03:15,840 --> 00:03:17,520 white paper based on the results. 65 00:03:17,520 --> 00:03:19,700 Everybody on this webinar will be able to get 66 00:03:19,700 --> 00:03:22,220 a copy of that white paper. It's very detailed. 67 00:03:22,220 --> 00:03:24,980 Look at the results. I'm writing that now. That'll 68 00:03:24,980 --> 00:03:28,880 be available in early February. That'll go out. And 69 00:03:28,880 --> 00:03:31,200 then a couple of blogs that highlight some of the 70 00:03:31,200 --> 00:03:33,740 results. Those will go on light reading. Very happy to 71 00:03:33,740 --> 00:03:36,900 have both Ciena and Cisco as our sponsors this year 72 00:03:36,900 --> 00:03:40,400 and to have participants from each company helping 73 00:03:40,400 --> 00:03:44,060 me walk through and interpret the results that we 74 00:03:44,060 --> 00:03:48,300 have. It's very new data, really just before the holidays 75 00:03:48,300 --> 00:03:51,480 that this came in. The demographics, as you can 76 00:03:51,480 --> 00:03:55,100 see here, we really like to get a mix of what we define 77 00:03:55,100 --> 00:03:59,380 as Tier 1 and Tier 2 and 3 operators. So Tier 1, 78 00:03:59,380 --> 00:04:01,460 we define as a billion dollars or more in annual 79 00:04:01,460 --> 00:04:04,620 revenue, and a little over half the audience for this 80 00:04:04,620 --> 00:04:08,100 survey is classified as Tier 1. So we'll show some 81 00:04:08,100 --> 00:04:10,920 splits in the data there. And then another one, of 82 00:04:10,920 --> 00:04:14,180 course, is North American operators versus other regions, 83 00:04:14,180 --> 00:04:16,480 a little over half North America, which is typical 84 00:04:16,480 --> 00:04:18,620 in our surveys. And then we've got representation 85 00:04:18,620 --> 00:04:22,440 from EMEA, Asia-Pac, and Central and Latin America. 86 00:04:23,040 --> 00:04:26,320 let me there's more demographics detail that'll be 87 00:04:26,320 --> 00:04:28,380 in the white paper so you can see full but this 88 00:04:28,380 --> 00:04:31,140 kind of gives you a high level picture of who's 89 00:04:31,140 --> 00:04:34,360 responding let me introduce our speakers here happy to 90 00:04:34,360 --> 00:04:37,260 have marie fiala director of portfolio marketing 91 00:04:37,260 --> 00:04:40,560 with sienna representing sienna today hi marie welcome 92 00:04:40,610 --> 00:04:45,640 hi hello everyone good to have you on i look 93 00:04:45,640 --> 00:04:48,260 forward to your thoughts and then she is joined by omar 94 00:04:48,260 --> 00:04:51,020 sultan director of product management automation 95 00:04:51,020 --> 00:04:54,420 and AI with Cisco. Hi, Omar. Good morning to you. 96 00:04:54,980 --> 00:04:56,600 Good morning, and thanks to 97 00:04:56,600 --> 00:04:57,680 folks for joining us today. 98 00:04:58,660 --> 00:05:01,540 Yeah, looking forward to it. Let me, moving 99 00:05:01,540 --> 00:05:05,420 into the content of the survey, use cases and 100 00:05:05,420 --> 00:05:07,440 benefits, some of the high-level stuff before we 101 00:05:07,440 --> 00:05:12,300 get into more in the AI and agentic AI bits. 102 00:05:13,500 --> 00:05:15,180 I'll present a couple here, and then we'll 103 00:05:15,180 --> 00:05:17,860 get into the discussion. But this is a 104 00:05:17,860 --> 00:05:20,760 question we actually also asked last year, so 105 00:05:20,760 --> 00:05:23,280 it's good to kind of compare year to year. 106 00:05:24,360 --> 00:05:26,680 This is looking at the state of automation 107 00:05:26,680 --> 00:05:29,780 in telecom. Our audience is, we call 108 00:05:29,780 --> 00:05:31,500 them communication service providers. 109 00:05:31,620 --> 00:05:33,560 Telcos is kind of the older term 110 00:05:33,560 --> 00:05:35,820 used for that, telecom operators. 111 00:05:36,000 --> 00:05:40,400 And we base this question on five levels of autonomous 112 00:05:40,400 --> 00:05:43,380 networking, which is adapted from the tm forum which 113 00:05:43,380 --> 00:05:46,000 itself adapted from the automotive industry i think 114 00:05:46,000 --> 00:05:50,120 is most are aware it's not a perfect model but um 115 00:05:50,480 --> 00:05:53,660 people know it so if you ask a question around these 116 00:05:53,660 --> 00:05:57,420 five levels people have a general sense so uh we ask 117 00:05:57,420 --> 00:06:00,880 it uh and this is the results from from this year if 118 00:06:00,880 --> 00:06:06,280 you look at the today picture 70 of operators are in 119 00:06:06,280 --> 00:06:10,340 the partial automation phase uh or below on the global 120 00:06:10,340 --> 00:06:13,470 results so that's currently the the dark blue is 121 00:06:13,470 --> 00:06:16,680 is um where are you today in your phase of automation 122 00:06:16,680 --> 00:06:19,940 uh level two is uh partial automation it's really 123 00:06:19,940 --> 00:06:22,460 the earliest stage of what we would call true automation 124 00:06:22,460 --> 00:06:25,380 this is where an introduction of ai and machine 125 00:06:25,380 --> 00:06:30,200 learning start to come into the network um if you look 126 00:06:30,200 --> 00:06:35,860 out to 2028 the pinkish uh shade here in our color 127 00:06:35,860 --> 00:06:40,830 screen 50 color scheme 56 percent expect to be conditionally 128 00:06:40,840 --> 00:06:44,580 autonomous or higher by the end of 2028 over 129 00:06:44,580 --> 00:06:47,260 the next three years that would be either level three 130 00:06:47,260 --> 00:06:50,800 level four or level five in the the tm forum scheme 131 00:06:51,540 --> 00:06:55,580 level three is where the the largest group expects 132 00:06:55,580 --> 00:06:59,140 to be 26 percent highly autonomous i mean i'm 133 00:06:59,140 --> 00:07:01,340 sorry level four over the next three years is 134 00:07:01,340 --> 00:07:04,520 where the bulk of the audience expects to be at 135 00:07:04,520 --> 00:07:07,180 least the largest individual share. Highly 136 00:07:07,180 --> 00:07:11,420 autonomous is level four. So fairly aggressive plans, 137 00:07:11,420 --> 00:07:13,660 actually very consistent with last year's 138 00:07:13,660 --> 00:07:17,100 results in terms of that progression. So operators 139 00:07:17,560 --> 00:07:20,920 are surely consistent and have something in mind of 140 00:07:20,920 --> 00:07:24,280 where they want to go. Level four, not so much in level 141 00:07:24,280 --> 00:07:27,140 five, which we can certainly discuss as we go through. 142 00:07:27,640 --> 00:07:32,660 The other, as I mentioned up front, we talk about 143 00:07:32,660 --> 00:07:34,960 Tier 1 versus Tier 2 and Tier 3. I'll just 144 00:07:34,960 --> 00:07:37,880 highlight this here. It is fairly consistent. The 145 00:07:37,880 --> 00:07:42,300 Tier 1 operators are more aggressive in moving to 146 00:07:42,300 --> 00:07:45,800 AI versus their Tier 2 and Tier 3 counterparts. 147 00:07:45,800 --> 00:07:48,360 Again, we saw the same thing last year. So this 148 00:07:48,360 --> 00:07:51,070 is the same question. I just broke it out by the 149 00:07:51,160 --> 00:07:54,460 operator size. Tier 1, dark blue, teal is the 150 00:07:54,460 --> 00:07:56,700 Tier 2 and 3s. and then the left is the current 151 00:07:56,700 --> 00:08:01,880 state, and the right is the expected state three 152 00:08:01,880 --> 00:08:05,020 years from now. You can see the Tier 2s and 3s are 153 00:08:05,020 --> 00:08:07,680 significantly less advanced than Tier 1s today. 154 00:08:07,980 --> 00:08:10,340 Probably, you know, as you might expect, these 155 00:08:10,340 --> 00:08:12,640 are smaller operators, less sophisticated, 156 00:08:12,800 --> 00:08:15,260 but a pretty big gap of where they are today. 157 00:08:15,540 --> 00:08:16,680 On the right, 158 00:08:16,960 --> 00:08:19,220 interesting, you know, Tier 1s are still expected to 159 00:08:19,220 --> 00:08:21,600 be more advanced three years out from now, again as 160 00:08:21,600 --> 00:08:24,780 you would expect but um the the takeaway that that 161 00:08:24,780 --> 00:08:27,060 i had particularly on this one is that the tier two 162 00:08:27,060 --> 00:08:30,010 threes have pretty significant expectations for 163 00:08:30,010 --> 00:08:33,980 moving to automation uh over the next three years so as 164 00:08:33,980 --> 00:08:35,580 we go through the discussion we can kind of keep 165 00:08:35,580 --> 00:08:40,080 that in mind and you know tease out maybe how how how 166 00:08:40,080 --> 00:08:42,380 vendors like cisco and sienna and others can can help 167 00:08:42,380 --> 00:08:45,120 these smaller operators uh get to where they clearly 168 00:08:45,120 --> 00:08:47,280 want to go but but aren't uh certainly don't have 169 00:08:47,280 --> 00:08:50,820 the resources currently to do it uh with that let 170 00:08:50,820 --> 00:08:54,600 me uh let me jump into the discussion on this one um 171 00:08:54,740 --> 00:08:59,040 and omar i'll let you lead this one was on use cases 172 00:08:59,040 --> 00:09:02,120 and just for background for the audience so we we 173 00:09:02,120 --> 00:09:04,780 have a set of 11 use cases you can see them here and 174 00:09:05,110 --> 00:09:08,260 their standard set that we came up with that we used 175 00:09:08,260 --> 00:09:11,060 the same ones last year they're not perfect but it 176 00:09:11,060 --> 00:09:14,900 gives a a good you know standardized metric across 177 00:09:14,900 --> 00:09:17,720 multiple questions where we can ask about use cases 178 00:09:17,800 --> 00:09:20,760 uh the today picture network performance monitoring 179 00:09:20,760 --> 00:09:24,640 and inventory are topping the list of where automation 180 00:09:24,640 --> 00:09:28,480 is today looking out uh tomorrow which means over the 181 00:09:28,480 --> 00:09:30,680 next three years there are really strong expectations 182 00:09:30,680 --> 00:09:33,600 to adopt almost all of the use cases really all of 183 00:09:33,600 --> 00:09:38,500 them but 80 i highlight here um expect to adopt 10 of 184 00:09:38,500 --> 00:09:41,520 the 11. the one that was lagging a bit was network 185 00:09:41,520 --> 00:09:48,780 design and planning, but still quite impressive. Omar, 186 00:09:49,160 --> 00:09:51,280 thoughts and comments on this 187 00:09:51,280 --> 00:09:53,460 one? I'll let you start it off. 188 00:09:54,200 --> 00:09:56,880 Yeah, so I think one of the key things is there's 189 00:09:56,880 --> 00:09:58,740 nothing like really esoteric here. This is all 190 00:09:58,740 --> 00:10:02,710 bread and butter operations for a transport 191 00:10:02,860 --> 00:10:06,160 network, and you kind of see the investment across 192 00:10:06,860 --> 00:10:14,060 all these use cases in that pink-colored bar. And 193 00:10:14,060 --> 00:10:15,800 this is kind of what we see, too. The initial use 194 00:10:15,800 --> 00:10:19,340 cases were like more augmenting staff with AI 195 00:10:19,340 --> 00:10:22,440 to help them do kind of the humdrum work, do it 196 00:10:22,440 --> 00:10:26,340 faster or more efficiently or offload altogether. 197 00:10:26,440 --> 00:10:29,780 I think the other piece is things like planning 198 00:10:29,780 --> 00:10:32,780 and design and even testing management. 199 00:10:33,380 --> 00:10:33,980 you know, 200 00:10:34,300 --> 00:10:36,020 less traction there. I think those 201 00:10:36,020 --> 00:10:37,080 are the kinds of things where maybe 202 00:10:37,080 --> 00:10:38,640 people should be spending their time, 203 00:10:38,760 --> 00:10:40,180 kind of the higher order functions 204 00:10:40,180 --> 00:10:44,100 and, you know, kind of the automation 205 00:10:44,620 --> 00:10:46,660 and implementation actually frees them, 206 00:10:46,660 --> 00:10:48,220 allows them to do those kinds of things. 207 00:10:49,980 --> 00:10:51,880 Yeah, and then one point I'll make, so 208 00:10:51,880 --> 00:10:53,520 these are a set of use cases. And so 209 00:10:53,520 --> 00:10:55,250 when you see there's lack of adoption, 210 00:10:55,460 --> 00:10:57,140 it doesn't mean that operators 211 00:10:57,140 --> 00:10:59,280 aren't interested in that use case. 212 00:10:59,420 --> 00:11:01,620 And this is, you know, you understand it. I just 213 00:11:01,620 --> 00:11:03,040 want to make it clear for the audience. It doesn't 214 00:11:03,040 --> 00:11:05,220 mean that the use case isn't of interest. This is 215 00:11:05,220 --> 00:11:07,890 specifically trying to get where automation, where 216 00:11:07,890 --> 00:11:10,740 they see the applicability of automation. And 217 00:11:10,740 --> 00:11:13,500 then the question is, you know, are there things 218 00:11:13,500 --> 00:11:17,720 that are missing or not in applying AI to automation? 219 00:11:17,720 --> 00:11:21,000 But it's not any time there's a low finding on 220 00:11:21,000 --> 00:11:22,280 this or any of the questions. It doesn't mean the 221 00:11:22,280 --> 00:11:24,520 use case isn't important. It's, you know, we're 222 00:11:24,520 --> 00:11:27,580 really drilling into automation and AI a bit later. 223 00:11:27,680 --> 00:11:29,360 Marie, any comments or thoughts on this 224 00:11:29,360 --> 00:11:31,080 one before we go to the next data point? 225 00:11:31,890 --> 00:11:34,440 well the the one thing i wanted to point out 226 00:11:34,440 --> 00:11:36,960 here it was kind of surprising to me because 227 00:11:37,200 --> 00:11:40,220 network deployment and commissioning that use case 228 00:11:40,220 --> 00:11:45,020 was lower at 21 percent and when it comes to automation 229 00:11:45,020 --> 00:11:47,820 we you know we at santa we're actually seeing a 230 00:11:47,820 --> 00:11:50,500 real driver for automating those processes because 231 00:11:50,500 --> 00:11:54,160 of this scaling that's happening especially uh for 232 00:11:54,160 --> 00:11:57,960 data center build out and data center connectivity, 233 00:11:58,160 --> 00:12:00,280 right? The high capacity connectivity across 234 00:12:00,280 --> 00:12:04,100 data centers for expanding AI training 235 00:12:04,100 --> 00:12:07,220 geographically. So we're actually seeing, like, 236 00:12:07,220 --> 00:12:09,760 despite that being 21%, we're seeing a lot 237 00:12:09,760 --> 00:12:11,800 of interest in automating those processes. 238 00:12:13,180 --> 00:12:15,760 So this is another good point, Marie. 239 00:12:16,920 --> 00:12:21,780 The other thing about the audience for this survey 240 00:12:21,780 --> 00:12:26,660 is it's very, very telco heavy. um yeah and not uh 241 00:12:26,660 --> 00:12:30,020 hyperscaler or neocloud uh really much representation 242 00:12:30,020 --> 00:12:33,800 at all so it's interesting so you know ai uh 243 00:12:33,800 --> 00:12:38,120 connectivity dci certainly telcos are are in play in 244 00:12:38,120 --> 00:12:41,300 there to a degree but as you know the bulk of it 245 00:12:41,300 --> 00:12:44,540 right now is being dominated by the hyperscalers so 246 00:12:45,000 --> 00:12:48,200 interesting that maybe the hyperscalers 247 00:12:48,200 --> 00:12:50,140 certainly the hyperscalers and maybe telcos that 248 00:12:50,140 --> 00:12:52,020 are serving that particular application might 249 00:12:52,020 --> 00:12:55,500 be seeing something that for the broader set 250 00:12:55,500 --> 00:12:58,800 of, you know, telco, they're not there yet. 251 00:13:00,780 --> 00:13:03,920 So, you know, whether they're wrong 252 00:13:03,920 --> 00:13:05,320 to be missing it, I don't know. But 253 00:13:05,320 --> 00:13:06,920 interesting, interesting observation. 254 00:13:08,320 --> 00:13:11,140 Yeah, anyway, go ahead, Omar. 255 00:13:11,320 --> 00:13:13,080 I mean, if you look at the adoption data, 256 00:13:13,080 --> 00:13:15,060 right, there's still a good chunk of 257 00:13:15,060 --> 00:13:17,540 respondents that are still zero to one, 258 00:13:18,120 --> 00:13:20,480 zero to two in terms of where they are, where 259 00:13:20,480 --> 00:13:22,540 they want to be. And I think that's some 260 00:13:22,580 --> 00:13:24,800 of the stuff I'm referring to is like this 261 00:13:24,800 --> 00:13:27,180 and just bread and butter automation stuff 262 00:13:27,960 --> 00:13:30,840 that you can do today that folks are still 263 00:13:30,840 --> 00:13:33,020 trying to just get the basics in place. 264 00:13:34,400 --> 00:13:36,560 Very much bread and butter, I 265 00:13:36,560 --> 00:13:38,540 think is the way to look at it. 266 00:13:39,740 --> 00:13:42,780 And if you compare it to how these, I don't have a 267 00:13:42,780 --> 00:13:45,440 slide for it, but if you look a little bit of, you know, 268 00:13:45,440 --> 00:13:49,540 how they answered last year there is a bit of a 269 00:13:49,540 --> 00:13:52,940 movement there so i i think um you know some of them were 270 00:13:52,940 --> 00:13:56,280 certainly still in the mix but um i think they're 271 00:13:56,280 --> 00:13:58,760 still you know kind of finding their footing about 272 00:13:58,760 --> 00:14:03,400 where automation uh applies to a degree maybe some of 273 00:14:03,400 --> 00:14:06,310 that you know showing showing through in the results um 274 00:14:07,360 --> 00:14:10,760 let me uh move to the next one so this one was on uh 275 00:14:10,760 --> 00:14:14,200 digital twin marie i'll let you kick this one off um 276 00:14:14,410 --> 00:14:18,560 so the same thing we took the 11 use cases that we 277 00:14:18,560 --> 00:14:21,800 sort of standardized and asked about where digital twin 278 00:14:21,800 --> 00:14:24,220 is is most applicable and of course you're going to 279 00:14:24,220 --> 00:14:26,520 get different results from where you know applications 280 00:14:26,520 --> 00:14:29,340 are are today so it's it shouldn't be surprising 281 00:14:29,340 --> 00:14:31,080 that the list is going to look different from the last 282 00:14:31,080 --> 00:14:34,320 slide this is digital twin here network uh optimization 283 00:14:34,320 --> 00:14:36,320 traffic engineering and network performance 284 00:14:36,320 --> 00:14:39,340 monitoring were the top ones on the list and we had an 285 00:14:39,340 --> 00:14:42,920 additional six as you can see kind of going down the list 286 00:14:42,920 --> 00:14:45,420 where they're ordered from you know most applicability 287 00:14:45,420 --> 00:14:49,120 to least you can see where where it resonated um 288 00:14:49,940 --> 00:14:54,820 uh and they didn't uh just as the the quick um you 289 00:14:54,820 --> 00:14:58,440 know look it didn't correlate specifically with with 290 00:14:58,440 --> 00:15:00,600 the overall transport use case priorities on the 291 00:15:00,600 --> 00:15:03,180 last slide so digital twin is seems to be applying for 292 00:15:03,180 --> 00:15:05,680 specific things but the only other comment i'll 293 00:15:05,680 --> 00:15:08,560 make here before i let marie comment is um that 36 294 00:15:08,560 --> 00:15:12,660 percent not uh having plans for digital twin for 295 00:15:12,660 --> 00:15:17,640 transport i think that's fairly high um we we noted it 296 00:15:17,640 --> 00:15:20,700 last year also some some challenges to digital twins 297 00:15:20,700 --> 00:15:23,060 specific to transport so i think that's showing it 298 00:15:23,060 --> 00:15:27,280 up as well um but maria i'll let you open open it up 299 00:15:27,280 --> 00:15:30,880 first the use cases and and how they resonate with 300 00:15:30,880 --> 00:15:34,120 what you see and then also maybe addressing the um 301 00:15:37,860 --> 00:15:38,700 Yeah, 302 00:15:39,920 --> 00:15:42,120 the one thing I wanted to note, like, it is just 303 00:15:42,120 --> 00:15:44,680 a one-year horizon here. The way the question 304 00:15:44,680 --> 00:15:47,380 was phrased here is over the next year or so. 305 00:15:48,220 --> 00:15:51,600 So kind of in the very short term, there's 306 00:15:51,600 --> 00:15:53,560 like one-third of operators. You're right, 307 00:15:53,560 --> 00:15:56,400 that was, you know, it's quite high that 308 00:15:56,400 --> 00:15:59,320 a third have no plans for digital twin. 309 00:15:59,320 --> 00:16:02,320 But it really wasn't that surprising to me. just that 310 00:16:02,320 --> 00:16:05,400 there is some confusion around um what digital twin 311 00:16:05,400 --> 00:16:08,610 is or isn't um over and above you know simulation 312 00:16:08,840 --> 00:16:15,860 capabilities and um the uh the so i'll spend a bit of time 313 00:16:15,860 --> 00:16:18,760 on just on the next slide just uh going over how how 314 00:16:18,760 --> 00:16:22,640 it works in practice but um with respect to the results 315 00:16:22,640 --> 00:16:26,060 here i i did see that network design and planning 316 00:16:26,060 --> 00:16:31,180 it was um quite high on the list and in comparison 317 00:16:31,180 --> 00:16:35,590 though for automation, it was low on the list. So what 318 00:16:35,590 --> 00:16:39,560 that really brought home to me was that, you know, 319 00:16:39,560 --> 00:16:42,360 operators, they're looking to use network digital twin 320 00:16:42,360 --> 00:16:45,860 for those planning use cases to get better confidence 321 00:16:45,860 --> 00:16:48,740 in the results before they can even hope to automate. 322 00:16:48,740 --> 00:16:51,630 So that was one that really, that 323 00:16:51,920 --> 00:16:53,760 was a point that really struck me. 324 00:16:56,040 --> 00:17:00,780 Yeah, good point on the one-year horizon too. i i 325 00:17:00,780 --> 00:17:02,840 should have noted that a lot of times we ask about 326 00:17:02,840 --> 00:17:04,820 three years this year this time we really wanted to 327 00:17:04,820 --> 00:17:09,370 know these are very very near-term plans so um that 328 00:17:09,370 --> 00:17:12,730 that certainly would would impact the 36 percent um 329 00:17:12,730 --> 00:17:16,260 yeah any uh i know you want to kind of walk through on 330 00:17:16,260 --> 00:17:19,620 the next slide um before you do that i'll let omar 331 00:17:19,620 --> 00:17:21,840 comment um although marie did you have anything else 332 00:17:21,840 --> 00:17:23,740 to say on the results and then i'll kick it over 333 00:17:23,740 --> 00:17:29,820 to omar omar any thoughts on this one uh so i do think 334 00:17:29,820 --> 00:17:33,280 digital twin is is quite overloaded in terms of 335 00:17:33,690 --> 00:17:37,420 multiple definitions you know I think you can have 336 00:17:37,420 --> 00:17:40,960 digital twins with varying levels of fidelity and how 337 00:17:40,960 --> 00:17:44,980 rich in a complex environment they can model I think 338 00:17:44,980 --> 00:17:47,180 from our perspective we do see like a high fidelity 339 00:17:47,180 --> 00:17:51,460 digital twin is critical to an AI strategy I think 340 00:17:51,460 --> 00:17:54,170 for customers have trust in their systems they they're 341 00:17:54,170 --> 00:17:56,480 gonna want to be able to test a proposed change 342 00:17:56,480 --> 00:17:59,380 and see how it behaves before pushing the production 343 00:17:59,380 --> 00:18:01,960 if it can really automate that kind of tool chain. 344 00:18:01,960 --> 00:18:04,320 You know, they need that kind of ability 345 00:18:04,320 --> 00:18:07,300 to test and trust before they deploy, 346 00:18:07,520 --> 00:18:09,080 which is ultimately going to drive 347 00:18:09,080 --> 00:18:11,560 the need for a very high-fidelity 348 00:18:11,560 --> 00:18:13,780 digital twin to be able to pull it off. 349 00:18:14,920 --> 00:18:19,440 Yeah, and I do think some operators are still trying 350 00:18:19,440 --> 00:18:22,140 to understand, you know, even still what digital twin 351 00:18:22,140 --> 00:18:27,480 is as well. Fairly new concept, I think, for many. 352 00:18:28,180 --> 00:18:30,460 With that, why don't I let, Marie, I'll 353 00:18:30,460 --> 00:18:34,000 let you walk through a bit more detail here, 354 00:18:34,000 --> 00:18:35,280 and then we'll go to the next section. 355 00:18:36,360 --> 00:18:40,140 Sure. And this really picks up what Omar was talking 356 00:18:40,140 --> 00:18:43,440 about, about building trust. So just wanted to 357 00:18:43,440 --> 00:18:45,280 spend a bit of time, you know, so how does it really 358 00:18:45,280 --> 00:18:48,540 work? So a lot of people in the audience here, 359 00:18:48,540 --> 00:18:51,080 you're familiar that a digital twin is a virtual 360 00:18:51,080 --> 00:18:54,100 representation of the actual details of the physical 361 00:18:54,100 --> 00:18:56,080 network, the live network, everything from the 362 00:18:56,080 --> 00:18:58,780 network elements, configuration, behavior, et cetera. 363 00:18:58,780 --> 00:19:01,640 And these are complex, right? Like 364 00:19:01,640 --> 00:19:03,680 the physical transport network, 365 00:19:03,860 --> 00:19:07,320 it has multi-layer IP optical infrastructure from 366 00:19:07,320 --> 00:19:09,940 different vendors, and there's really a mix of legacy 367 00:19:09,940 --> 00:19:13,900 and modern devices. So some only have TL1 interfaces, 368 00:19:14,030 --> 00:19:17,400 some are GNI enabled for real-time streaming, et 369 00:19:17,400 --> 00:19:19,800 cetera. I have to take all of that into account. So 370 00:19:19,800 --> 00:19:23,540 it's quite difficult to construct and implement a 371 00:19:23,540 --> 00:19:27,100 network digital twin and have it accurate now what 372 00:19:27,100 --> 00:19:29,680 i'm showing here so it's all about the good data 373 00:19:29,680 --> 00:19:32,760 what the network can provide up order to the control 374 00:19:32,760 --> 00:19:36,360 layer and i'm showing here conceptually two 375 00:19:36,360 --> 00:19:38,660 streams but it could be one single stream it could 376 00:19:38,660 --> 00:19:41,880 actually be a common data source that both the network 377 00:19:41,880 --> 00:19:44,710 digital twin and the controller really consume 378 00:19:45,940 --> 00:19:50,640 consume the same stream so again just an abstract 379 00:19:51,690 --> 00:19:53,640 representation here. But when you have an 380 00:19:53,640 --> 00:19:56,080 accurate twin, you can execute all sorts of what 381 00:19:56,080 --> 00:19:59,960 -if scenarios to really risk-free tryout 382 00:19:59,960 --> 00:20:03,420 solutions before applying them to the network. 383 00:20:04,400 --> 00:20:06,980 And so what really makes it interesting 384 00:20:06,980 --> 00:20:09,220 is when you bring AI into the picture. 385 00:20:09,400 --> 00:20:12,080 And Omar stated that as well. Really, 386 00:20:12,080 --> 00:20:15,660 AI engines, AI, agentic AI frameworks, 387 00:20:16,100 --> 00:20:18,060 network digital twin and controllers that can 388 00:20:18,060 --> 00:20:20,240 all work together to bring about that optimal 389 00:20:20,640 --> 00:20:25,760 outcome they can you know using various various 390 00:20:25,760 --> 00:20:29,680 standards-based interfaces very accepted interfaces like 391 00:20:29,680 --> 00:20:33,100 when it comes to ai agent to agent interfaces and 392 00:20:33,100 --> 00:20:37,660 whatnot so ai generated recommendations when you put 393 00:20:37,660 --> 00:20:40,330 that in the mix they can be really safely validated 394 00:20:40,330 --> 00:20:42,940 in a digital twin environment before they're enacted 395 00:20:42,940 --> 00:20:46,920 on the network so if network digital twins aren't 396 00:20:46,920 --> 00:20:50,900 in that one-year time horizon of adoption. As operators 397 00:20:50,900 --> 00:20:53,880 do adopt AI strategies, they're going to become 398 00:20:53,880 --> 00:20:57,380 more and more important to really build that trust. 399 00:20:57,900 --> 00:21:01,360 Yeah, so that kind of summarizes my 400 00:21:01,970 --> 00:21:03,580 recovery. Yeah, great. Thank you. 401 00:21:03,900 --> 00:21:09,740 And trust throughout is quite a theme, I think, 402 00:21:09,740 --> 00:21:13,080 in the results and AI as well as you and Omar 403 00:21:13,080 --> 00:21:16,720 has just brought out. So I think quite rightly, 404 00:21:16,720 --> 00:21:18,860 we focused a lot of the survey this year on 405 00:21:18,860 --> 00:21:21,740 artificial intelligence. So let's kind of dig into 406 00:21:22,060 --> 00:21:24,240 some of the details here. This is not the first 407 00:21:24,240 --> 00:21:26,740 year. Last year, we focused a fair amount on AI 408 00:21:26,740 --> 00:21:29,400 as well, but we certainly built on it this year. 409 00:21:30,010 --> 00:21:31,780 And the first thing, of course, you 410 00:21:31,780 --> 00:21:35,180 want to look at for AI adoption for 411 00:21:35,180 --> 00:21:38,320 transport is what stage are operators at? 412 00:21:38,480 --> 00:21:41,220 So that was the question here for the global group. 413 00:21:41,280 --> 00:21:45,220 More than half of the service providers surveyed um 414 00:21:45,220 --> 00:21:49,240 habit uh that that middle bar uh you could see uh quite 415 00:21:49,240 --> 00:21:53,040 further along than than the other ones is is you know 416 00:21:53,040 --> 00:21:55,860 what i would call more the um the early phases of 417 00:21:55,860 --> 00:21:58,920 production networks so you're seeing ai into production 418 00:21:58,920 --> 00:22:02,720 networks for transport almost half of them um a very 419 00:22:02,720 --> 00:22:05,920 small share as you might expect in advanced phases um 420 00:22:05,920 --> 00:22:10,180 and still you know 20 not not yet using ai uh and i'll 421 00:22:10,180 --> 00:22:12,420 do a bit of a drill down on the next slide but marie 422 00:22:12,420 --> 00:22:16,400 um maybe let you comment on these results in terms of 423 00:22:16,400 --> 00:22:20,040 as expected not expected or any other thoughts on on 424 00:22:20,040 --> 00:22:24,360 how this came in yeah this this really does reflect 425 00:22:24,360 --> 00:22:26,860 what we're seeing in our customer engagements you know 426 00:22:26,860 --> 00:22:30,520 operators are trying out ai focusing on specific pilot 427 00:22:30,520 --> 00:22:34,040 use cases because it really takes a a fair amount of 428 00:22:34,040 --> 00:22:37,620 retooling of infrastructure and we'll get into that 429 00:22:37,620 --> 00:22:40,740 that there's you know privacy concerns regarding cloud 430 00:22:40,740 --> 00:22:44,480 access to large language models etc so it takes a 431 00:22:44,480 --> 00:22:47,340 fair amount of retailing and also retailing of processes 432 00:22:47,340 --> 00:22:52,480 to incorporate AI so and so it does take time so I 433 00:22:52,480 --> 00:22:55,680 guess I wasn't you know it's good that there is such a 434 00:22:55,680 --> 00:22:59,340 high number actually trying it out for specific uses 435 00:23:00,600 --> 00:23:04,260 yeah i mean you know certainly we we had a little 436 00:23:04,260 --> 00:23:06,700 bit of discussion on the hype certainly i don't think 437 00:23:06,700 --> 00:23:09,260 anybody's going to disagree the telco audience 438 00:23:09,260 --> 00:23:13,340 is going to be more conservative than the hyperscalers 439 00:23:13,340 --> 00:23:17,840 for for many many many reasons so um it always 440 00:23:17,840 --> 00:23:19,980 depends on kind of what perspective you're going in 441 00:23:19,980 --> 00:23:22,780 but i i would say i would agree a result like this 442 00:23:22,780 --> 00:23:27,320 is is is good news i think considering it's telco 443 00:23:27,320 --> 00:23:30,380 they're they're they're moving ahead at uh i think 444 00:23:30,380 --> 00:23:36,910 uh probably a reasonable um pace um let me um pull 445 00:23:36,910 --> 00:23:41,100 the next one here on and this was um this was 446 00:23:41,100 --> 00:23:43,620 another one of looking at the tier ones versus the tier 447 00:23:43,620 --> 00:23:46,280 two threes and and i wanted to highlight it here 448 00:23:46,540 --> 00:23:49,460 because just as an automation we're seeing different 449 00:23:49,460 --> 00:23:51,580 expectations when we look at the tier one operators 450 00:23:51,580 --> 00:23:55,360 versus the tier two threes in fact it's quite quite 451 00:23:55,360 --> 00:23:59,480 significant uh on the left i circled you can see 32 452 00:23:59,480 --> 00:24:03,220 percent of the um tier two threes not yet using uh ai 453 00:24:03,220 --> 00:24:06,780 for transport versus just 10 percent for uh for the 454 00:24:06,780 --> 00:24:10,740 tier one so the tier ones are clearly driving not just 455 00:24:10,740 --> 00:24:13,080 automation but particularly the the ai adoption 456 00:24:13,080 --> 00:24:17,480 versus uh the the smaller operators um marie i'll let 457 00:24:17,480 --> 00:24:20,180 you comment here and then um omar if if you wanted to 458 00:24:20,180 --> 00:24:22,420 to share any thoughts but what you know one one thing 459 00:24:22,420 --> 00:24:29,820 i would be interested in is uh how vendors can help 460 00:24:29,820 --> 00:24:33,280 if at all uh specifically the tier two tier threes 461 00:24:33,280 --> 00:24:36,320 move move a bit more aggressively and you know more 462 00:24:36,320 --> 00:24:41,080 in line with the larger operators yeah so i'll pick 463 00:24:41,080 --> 00:24:43,580 up on that in fact like the tier two tier threes 464 00:24:43,580 --> 00:24:46,320 sometimes they're they're more agile than the tier ones 465 00:24:46,320 --> 00:24:51,220 and so one way that we are seeing is they've actually 466 00:24:51,220 --> 00:24:54,220 the smaller operators have really adopted cloud 467 00:24:54,220 --> 00:24:57,860 infrastructure more quickly like they're even using you 468 00:24:57,860 --> 00:25:01,040 know for example um like network control solutions that 469 00:25:01,040 --> 00:25:03,260 are hosted in the in the cloud and that makes it 470 00:25:03,260 --> 00:25:07,050 easier for them to adopt ai technologies because ai 471 00:25:07,050 --> 00:25:11,220 applications are hosted in the cloud so in fact um it's 472 00:25:11,320 --> 00:25:14,680 somewhat easier like despite them despite these 473 00:25:14,680 --> 00:25:17,860 results showing that they're um they're they're not at 474 00:25:17,860 --> 00:25:21,380 the that's ai adoption stage these are the tier ones 475 00:25:21,380 --> 00:25:25,300 they will catch up quicker i think because they're 476 00:25:25,300 --> 00:25:28,080 all familiar with the cloud they don't they typically 477 00:25:28,370 --> 00:25:32,140 are again much more open to uh already leveraging 478 00:25:32,140 --> 00:25:34,180 the cloud for all sorts of applications 479 00:25:34,200 --> 00:25:36,120 Yeah, yeah, good points. 480 00:25:36,260 --> 00:25:37,800 Omar, any thoughts here? 481 00:25:38,470 --> 00:25:40,760 Yeah, I'll just kind of echo that, maybe say it more 482 00:25:40,760 --> 00:25:45,430 strongly. I think for Tier 2, Tier 2, Tier 3, they 483 00:25:45,430 --> 00:25:48,860 actually have the opportunity to leapfrog and 484 00:25:48,860 --> 00:25:51,700 maybe go from Tier 1 to, you know, Level 1, Level 2, 485 00:25:51,700 --> 00:25:54,600 to Level 3, Level 4, because they can, you know, 486 00:25:54,600 --> 00:25:56,600 they're simpler environments, simpler operations. They 487 00:25:56,600 --> 00:25:59,580 can consume off-the-shelf solutions better than 488 00:25:59,580 --> 00:26:02,280 Tier 1s or, you know, almost everything they do is 489 00:26:02,280 --> 00:26:04,960 bespoke. So there's a lot more heavy lifting for 490 00:26:05,040 --> 00:26:08,720 those operators to move through the autonomy levels. 491 00:26:09,800 --> 00:26:12,180 Yeah, interesting, interesting perspective. Let 492 00:26:12,180 --> 00:26:16,180 me, we have a couple of use cases we want to go 493 00:26:16,180 --> 00:26:19,380 through. I think both of you guys have a use case. 494 00:26:19,380 --> 00:26:22,220 Before we get there, let me, let's just kind of 495 00:26:22,220 --> 00:26:25,120 get into the use cases a bit. Omar, I'll let you 496 00:26:25,120 --> 00:26:28,280 start this one. So we asked which use cases are 497 00:26:28,280 --> 00:26:32,100 expected to benefit the most from AI. So, again, 498 00:26:32,100 --> 00:26:34,240 this is the standard set of use cases, and this 499 00:26:34,240 --> 00:26:37,920 is which ones where AI might be most applicable. 500 00:26:38,580 --> 00:26:41,660 Strong preference for the top three there, predictive 501 00:26:41,660 --> 00:26:44,180 analysis, network performance monitoring, and 502 00:26:44,180 --> 00:26:47,500 network troubleshooting, and root cause analysis. 503 00:26:47,500 --> 00:26:50,820 And if I look back, comparing last year to 504 00:26:50,820 --> 00:26:53,080 this year, because we asked the same question, 505 00:26:53,500 --> 00:26:58,900 there's a very consistent showing of the top five 506 00:26:58,900 --> 00:27:01,840 listed there. Certainly the top five is, or the 507 00:27:01,840 --> 00:27:03,580 top five each year, the order has moved a little 508 00:27:03,580 --> 00:27:07,100 bit, but, you know, statistically maybe not at all. 509 00:27:07,100 --> 00:27:10,340 So operators know, they seem to know where they 510 00:27:10,340 --> 00:27:12,840 want to apply AI, whether they've done it or not. 511 00:27:14,230 --> 00:27:16,080 So that's encouraging. It seems to be a 512 00:27:16,080 --> 00:27:18,380 strategy there. But Omar, thoughts, comments on 513 00:27:18,380 --> 00:27:20,600 this, and then we'll dig into these cases. 514 00:27:21,100 --> 00:27:23,180 Yeah, sure. I think there's an interesting 515 00:27:23,180 --> 00:27:25,720 difference between the first two and the third. 516 00:27:26,230 --> 00:27:28,680 So the first two is stuff you can do today, right? 517 00:27:28,680 --> 00:27:30,960 I think there's lots of tool sets out there for doing 518 00:27:30,960 --> 00:27:33,760 these kinds of things. I think there is a comfort 519 00:27:33,760 --> 00:27:36,940 level with operators, with the tools and technology, 520 00:27:37,680 --> 00:27:40,580 and AI is a very linear way to have 521 00:27:40,580 --> 00:27:43,060 maybe better analytics, better insights, 522 00:27:43,120 --> 00:27:44,820 being able to get more proactive. 523 00:27:45,880 --> 00:27:48,860 Root cause analysis is kind of the opposite, right? 524 00:27:48,860 --> 00:27:51,940 When I talk to our customers, it's really the 525 00:27:51,940 --> 00:27:57,940 holy grail of AI use cases, But I think it's also 526 00:27:57,940 --> 00:28:02,240 the most complex and maybe the least mature of the 527 00:28:02,240 --> 00:28:03,740 things out there because it requires a lot of 528 00:28:03,740 --> 00:28:06,620 knowledge that's specific to that environment and those 529 00:28:06,620 --> 00:28:09,060 operations, those kinds of things. But I think as 530 00:28:09,060 --> 00:28:11,900 we start to crack that, it has the potential to 531 00:28:11,900 --> 00:28:13,700 have the greatest upside in terms of productivity, 532 00:28:13,920 --> 00:28:15,580 customer experience, those kinds of things. 533 00:28:16,300 --> 00:28:20,120 All right. Great. Let me just move things along. 534 00:28:20,120 --> 00:28:23,420 I know each of you has a use case, so let me tee 535 00:28:23,420 --> 00:28:26,120 those up for you. Omar, I believe you're first. 536 00:28:26,400 --> 00:28:28,460 And just for the audio, what we wanted to do is, 537 00:28:28,460 --> 00:28:30,580 you know, we have a bunch of data, but where we can, 538 00:28:30,580 --> 00:28:33,580 we want to show things are real. So we tried to 539 00:28:33,580 --> 00:28:36,320 insert some specific use cases in there to, you know, 540 00:28:36,320 --> 00:28:39,980 kind of put reality to these pie charts and bar 541 00:28:39,980 --> 00:28:42,820 charts. So, Omar, I'll let you start with the one 542 00:28:42,820 --> 00:28:45,160 you picked, and then Marie, we'll let you go next. 543 00:28:45,460 --> 00:28:48,560 Sure. So we've been building an agentic platform 544 00:28:48,560 --> 00:28:51,680 for a couple of years now, and Config Drift is 545 00:28:51,680 --> 00:28:54,380 one of the first agents we are releasing shortly. 546 00:28:55,200 --> 00:28:58,540 And it's interesting in that it's designed to find 547 00:28:58,540 --> 00:29:01,060 compliance with golden configs for your devices. 548 00:29:01,260 --> 00:29:03,560 And typically the solutions that are 549 00:29:03,560 --> 00:29:05,560 available today, including ones that we sell, 550 00:29:05,700 --> 00:29:10,800 really require you having a list of your golden configs 551 00:29:10,800 --> 00:29:14,020 and being able to do comparisons with what's running 552 00:29:14,020 --> 00:29:16,560 on the device. This takes a very AI approach, and 553 00:29:16,560 --> 00:29:20,800 in terms of it uses more a human intuitive process to 554 00:29:21,090 --> 00:29:24,000 find variances. So if I show you a picture of, like, 555 00:29:24,000 --> 00:29:26,400 three dogs and a cat on a slide, your mind automatically 556 00:29:26,400 --> 00:29:29,480 is, oh, the cat's, you know, the thing that doesn't 557 00:29:29,480 --> 00:29:31,800 belong here without any other prompting from you. 558 00:29:31,820 --> 00:29:34,120 And essentially, this is what Config Drift does. 559 00:29:34,120 --> 00:29:36,820 It'll go through all your router configs or 560 00:29:36,820 --> 00:29:39,880 device configs. It'll train a small onboard model 561 00:29:39,880 --> 00:29:42,420 on those so it starts to understand, okay, 562 00:29:42,420 --> 00:29:47,040 what's normal, what's normal to vary, what's normal 563 00:29:47,040 --> 00:29:48,980 to be the same. And then it pulls out the 564 00:29:48,980 --> 00:29:51,400 things that are varying that probably shouldn't. 565 00:29:51,540 --> 00:29:53,740 and it generates a report for you. So the nice 566 00:29:53,740 --> 00:29:56,220 thing is there's no prep work. There's no training. 567 00:29:56,410 --> 00:29:58,560 You can just kind of run this out of the 568 00:29:58,560 --> 00:30:01,240 box to find issues in your configurations. 569 00:30:02,360 --> 00:30:03,360 All right. 570 00:30:03,720 --> 00:30:07,040 Excellent. And Marie, the one 571 00:30:07,040 --> 00:30:08,420 that you wanted to highlight. 572 00:30:08,700 --> 00:30:12,060 Yeah, well, I just wanted to set it up first and just 573 00:30:12,060 --> 00:30:16,220 talk about kind of the ways that users consume AI. 574 00:30:16,400 --> 00:30:19,820 You know, there's two kind of personalities, I call 575 00:30:19,820 --> 00:30:22,560 it here. One is that ad hoc query that we're all familiar 576 00:30:22,560 --> 00:30:24,660 with because we have it on our devices now, et 577 00:30:24,660 --> 00:30:28,020 cetera, where you ask a question, you get a result. And 578 00:30:28,020 --> 00:30:31,000 then there's continuously running AI-driven insights. 579 00:30:31,000 --> 00:30:34,240 And so these aren't mutually exclusive by any means, 580 00:30:34,240 --> 00:30:37,360 but I just wanted to bring it back to the results of the 581 00:30:37,360 --> 00:30:40,160 survey here, the top one, predictive analytics and 582 00:30:40,160 --> 00:30:44,970 network health analysis, that was the top use case selected 583 00:30:45,700 --> 00:30:48,680 by the survey respondents, right? like that that 584 00:30:48,680 --> 00:30:52,120 lends itself more towards those continuously um running 585 00:30:52,120 --> 00:30:55,590 ai driven insights because it's about prediction right 586 00:30:55,590 --> 00:31:00,320 really useful for uh for planning teams so uh it's 587 00:31:00,320 --> 00:31:03,040 just important to keep that in mind and make sure that 588 00:31:03,040 --> 00:31:06,280 you know at the the ai tools that you have support both 589 00:31:06,420 --> 00:31:09,560 these um these type of uh personalities i'll call 590 00:31:09,560 --> 00:31:12,760 it now though the the use case that we're seeing most 591 00:31:12,760 --> 00:31:15,560 prevalent um albeit that predictive analytics was 592 00:31:15,560 --> 00:31:19,600 the top we actually see customers looking for ai for 593 00:31:19,600 --> 00:31:22,720 that those troubleshooting uh scenarios right because 594 00:31:22,720 --> 00:31:26,060 that's the um that that's the highest priority 595 00:31:26,060 --> 00:31:28,320 you know if there's an outage if there's if there's 596 00:31:28,320 --> 00:31:30,900 issues on their network uh they don't want their uh 597 00:31:30,900 --> 00:31:35,480 their their name their name in the paper so so they're 598 00:31:35,480 --> 00:31:38,740 really looking for help there and what i've what 599 00:31:38,740 --> 00:31:42,440 i'm showing here is the reactive i would call it 600 00:31:42,440 --> 00:31:45,360 reactive assurance or network troubleshooting um 601 00:31:45,360 --> 00:31:48,620 scenario where if there's something in contact you know 602 00:31:48,620 --> 00:31:51,120 there's there's an alert there's some issue and you 603 00:31:51,120 --> 00:31:54,120 can use uh that ai assistant that's what we call 604 00:31:54,120 --> 00:31:57,020 it in terms of an ad hoc theory or it can be alerted 605 00:31:57,020 --> 00:31:59,860 it doesn't really have to be initiated by the user 606 00:32:00,520 --> 00:32:01,400 automated. 607 00:32:01,840 --> 00:32:04,480 And that AI system really helps 608 00:32:04,480 --> 00:32:06,700 you diagnose the issue first, 609 00:32:07,300 --> 00:32:08,740 troubleshoot it, and then 610 00:32:08,740 --> 00:32:11,060 provides recommended resolution. 611 00:32:11,460 --> 00:32:13,660 And this is all spelled out. And again, back to 612 00:32:13,660 --> 00:32:16,660 building trust, we kind of show a step-by-step sequence 613 00:32:16,660 --> 00:32:19,560 of what the issues are, what you can do to resolve 614 00:32:19,560 --> 00:32:21,660 the issue. But all this can be automated further 615 00:32:21,660 --> 00:32:25,800 down the road. again so it's just we're kind of um 616 00:32:25,800 --> 00:32:31,660 there's really a or a time frame where users need to 617 00:32:31,750 --> 00:32:34,560 get more and more familiar with the capabilities 618 00:32:34,560 --> 00:32:37,500 be able to trust the ai engines and um in 619 00:32:37,500 --> 00:32:40,240 order to then get to the next step of automation 620 00:32:40,240 --> 00:32:44,340 at autonomous level that we talked about 621 00:32:46,680 --> 00:32:51,700 yeah good good point on um you know we we think of 622 00:32:51,700 --> 00:32:54,320 things and i think we asked in this context but what 623 00:32:54,320 --> 00:32:57,940 what kind of bubbles up a lot is is automation for 624 00:32:57,940 --> 00:33:00,440 reliability and network uptime and kind of the flip 625 00:33:00,440 --> 00:33:03,820 side that is avoiding your name in the news i think 626 00:33:03,820 --> 00:33:06,960 uh to to your point uh is is the negative side of 627 00:33:06,960 --> 00:33:10,380 that and uh we're certainly seeing that uh quite 628 00:33:10,380 --> 00:33:14,400 quite a bit automation clearly is being eyed to um to 629 00:33:14,400 --> 00:33:17,220 both on the benefit side and to avoid the very very 630 00:33:17,500 --> 00:33:21,220 strong uh negative uh that could really do 631 00:33:21,220 --> 00:33:25,320 damage to a brand um yeah and this this one here 632 00:33:25,600 --> 00:33:30,380 hits on that point um maria i'll let you um uh 633 00:33:30,650 --> 00:33:33,540 make some observations here but you know the the 634 00:33:33,540 --> 00:33:35,560 most significant benefits of ai for transport 635 00:33:35,560 --> 00:33:40,540 reducing um reducing human error and and faster 636 00:33:40,540 --> 00:33:43,280 troubleshoot it's it's exactly it those two at the 637 00:33:43,280 --> 00:33:46,980 top um i don't know maybe we've already covered 638 00:33:46,980 --> 00:33:48,660 it but in any comments you want to make here 639 00:33:48,660 --> 00:33:50,920 certainly consistent with what you just described 640 00:33:51,200 --> 00:33:54,660 yeah yeah exactly so they want operators they want 641 00:33:54,660 --> 00:33:58,000 to use ai to avoid errors in the first place 642 00:33:58,000 --> 00:34:01,360 but if they do happen um you know to do that 643 00:34:01,360 --> 00:34:05,070 troubleshooting all the quicker uh smarter and And the 644 00:34:05,070 --> 00:34:08,380 only thing I wanted to point out here is that, 645 00:34:08,380 --> 00:34:11,080 like, the OPEX savings versus CapEx savings, like, 646 00:34:11,080 --> 00:34:13,780 we certainly do see an emphasis on OPEX savings 647 00:34:13,780 --> 00:34:17,380 that really want to use, you know, our customers 648 00:34:17,380 --> 00:34:21,400 are looking to use AI to save operational costs. 649 00:34:22,540 --> 00:34:25,900 And that domino effect on customer satisfaction, et 650 00:34:25,900 --> 00:34:28,080 cetera, it's not quite as quantifiable right now. 651 00:34:29,880 --> 00:34:32,000 It's not that much of a focus. 652 00:34:32,260 --> 00:34:36,080 So you are seeing today a strong focus 653 00:34:36,080 --> 00:34:38,520 on the OPEX savings from AI? Was that? 654 00:34:38,900 --> 00:34:40,220 Absolutely. Yeah. 655 00:34:40,840 --> 00:34:44,660 Okay. Yeah, that's interesting because I don't know 656 00:34:44,660 --> 00:34:46,600 if it was last year, but a different survey we 657 00:34:46,600 --> 00:34:50,220 did. We actually did poll operators on how much OPEX 658 00:34:50,220 --> 00:34:52,760 savings they're getting from AI and automation. 659 00:34:52,760 --> 00:34:56,740 And it was surprisingly low, at least currently. 660 00:34:57,240 --> 00:35:00,880 um so it's it's good to see that here it's it's 661 00:35:00,880 --> 00:35:03,020 certainly a goal and what you're seeing with 662 00:35:03,020 --> 00:35:06,620 your own um your own discussions and work with 663 00:35:06,620 --> 00:35:11,300 customers but um i i don't i don't know if you 664 00:35:12,220 --> 00:35:14,920 it's always a challenge to quantify the savings 665 00:35:14,920 --> 00:35:17,300 that's that's what that's what we do see in 666 00:35:17,300 --> 00:35:21,400 real life that yeah that's absolutely right 667 00:35:21,460 --> 00:35:24,100 uh let's hit the challenges and then move to 668 00:35:24,100 --> 00:35:27,120 the last section on Agentica. Actually, we're 669 00:35:27,120 --> 00:35:29,820 doing pretty good on time. So, good job, guys. 670 00:35:31,120 --> 00:35:33,300 Adoption challenges. Omar, I'll let you kick this 671 00:35:33,300 --> 00:35:34,880 off. I mean, there's a whole bunch of interesting 672 00:35:34,880 --> 00:35:37,700 stuff in here. The main challenge is integrating with 673 00:35:37,700 --> 00:35:40,960 existing OSS, BSS management is the biggest one. 674 00:35:41,480 --> 00:35:44,820 Data privacy and compliance and 675 00:35:44,820 --> 00:35:47,960 maturity are also quite high. 676 00:35:48,860 --> 00:35:51,220 And they do vary. I don't have a data point 677 00:35:51,220 --> 00:35:53,020 for it, but they did vary quite a bit by 678 00:35:53,020 --> 00:35:56,460 size of operator tier one. versus Tier 2, 3s. 679 00:35:56,460 --> 00:36:00,060 The Tier 2, 3s, lack of in-house resources is 680 00:36:00,060 --> 00:36:03,180 their main barrier, which could explain why 681 00:36:03,180 --> 00:36:06,260 they're not as far along as maybe they could be. 682 00:36:06,620 --> 00:36:07,580 Omar, 683 00:36:08,100 --> 00:36:12,920 thoughts on these results, and are there ones that 684 00:36:13,660 --> 00:36:16,960 aren't here but you're seeing in your own interactions? 685 00:36:17,580 --> 00:36:20,160 Yeah, certainly integration is the big one, 686 00:36:20,160 --> 00:36:21,720 and it's not just integration of the tool 687 00:36:21,720 --> 00:36:25,040 sets that customers already have, but also the 688 00:36:25,040 --> 00:36:27,200 MOPS, the processes, those kinds of things. 689 00:36:27,320 --> 00:36:29,820 I think the advantage of regular 690 00:36:29,820 --> 00:36:32,120 automation is just, you know, 691 00:36:32,740 --> 00:36:35,080 automating existing MOPS. And a lot of times what 692 00:36:35,080 --> 00:36:37,380 we're doing with AI, as an example, and we just used, 693 00:36:37,640 --> 00:36:40,140 you know, you're changing how you go about doing 694 00:36:40,140 --> 00:36:44,100 things and customers need to kind of figure out 695 00:36:44,100 --> 00:36:46,680 how they want to do that in their own environment. 696 00:36:46,920 --> 00:36:50,420 Certainly data privacy and where data goes is a huge 697 00:36:50,420 --> 00:36:54,320 concern. A lot of our CSPs are very much on-prem 698 00:36:54,320 --> 00:36:58,720 operations, so how they handle data and using things like 699 00:36:58,720 --> 00:37:01,790 public LLMs and stuff is a big topic of conversation. 700 00:37:03,520 --> 00:37:07,300 I think the other piece for them that maybe 701 00:37:07,300 --> 00:37:09,540 is not on here is things like understanding 702 00:37:09,540 --> 00:37:12,530 when AI is mature enough to put into production. 703 00:37:13,220 --> 00:37:15,720 i mean ai's probabilistic that's always 704 00:37:15,720 --> 00:37:17,480 going to be wrong some amount of the time 705 00:37:17,900 --> 00:37:21,600 and and customers need to develop criteria okay 706 00:37:21,600 --> 00:37:24,380 how good is good enough and and how do i handle it 707 00:37:24,380 --> 00:37:26,740 when it's wrong because those are going to happen 708 00:37:28,980 --> 00:37:32,900 yep good good points um yeah let's um 709 00:37:32,900 --> 00:37:36,440 marie any any quick comments here or um 710 00:37:38,470 --> 00:37:40,520 yeah i did want to talk just to the 711 00:37:40,520 --> 00:37:43,080 integration and just that case because 712 00:37:43,280 --> 00:37:45,080 It is really important, like 713 00:37:45,080 --> 00:37:47,320 these AI use cases, obviously, 714 00:37:47,640 --> 00:37:51,920 are complex. They require a real detailed level 715 00:37:51,920 --> 00:37:54,300 of network observability and specialized knowledge 716 00:37:54,300 --> 00:37:56,120 of the network. So you really need to have 717 00:37:56,120 --> 00:37:59,300 an AI solution, kind of call it full stack, like 718 00:37:59,300 --> 00:38:01,880 from the device to the control layer to OSS. 719 00:38:02,340 --> 00:38:07,500 And so important to have a solution that has those 720 00:38:07,500 --> 00:38:10,520 standards-based interfaces. I talked about in 721 00:38:10,520 --> 00:38:14,200 AA or MCP, also a model context protocol when you're 722 00:38:14,200 --> 00:38:16,660 leveraging your use tools when it comes to AI. 723 00:38:16,980 --> 00:38:21,020 So you really want to have that seamless 724 00:38:21,020 --> 00:38:23,940 experience to address use cases at the 725 00:38:23,940 --> 00:38:26,560 top level that our customer is facing. 726 00:38:28,620 --> 00:38:29,600 Okay. 727 00:38:30,740 --> 00:38:31,720 Great. 728 00:38:31,730 --> 00:38:34,700 We'll move into the last section and then we will 729 00:38:34,700 --> 00:38:37,740 have some time for questions. So we do have some in, 730 00:38:37,740 --> 00:38:41,080 but now is a good time to start asking, and we'll hit 731 00:38:41,080 --> 00:38:43,740 as many as we can on the time at the end. So agentic 732 00:38:43,740 --> 00:38:47,420 AI, as I mentioned, brand new for this year's 733 00:38:47,420 --> 00:38:50,380 transport survey. So Omar, I'll let you – I know you've 734 00:38:50,380 --> 00:38:53,180 got two slides here to just kind of quickly set the 735 00:38:53,180 --> 00:38:57,380 context and define agentic AI for us. So go ahead. 736 00:38:57,840 --> 00:39:02,000 Sure. So we look at agentic AI as pieces of software 737 00:39:02,000 --> 00:39:04,720 that can run out and do things for us. So, you 738 00:39:04,720 --> 00:39:07,460 know, today, or yesterday, if you wanted to find an 739 00:39:07,460 --> 00:39:09,300 answer to something, you could use your own tools, you 740 00:39:09,300 --> 00:39:11,640 could find an expert, and there's probably, you 741 00:39:11,640 --> 00:39:13,660 know, a big pile of data in your organization that 742 00:39:13,660 --> 00:39:15,940 probably has information, but you don't know what's 743 00:39:15,940 --> 00:39:18,860 there, it's hard to get it, you know, get to the TL1 744 00:39:18,940 --> 00:39:21,340 interfaces and stuff that Marie was talking about. 745 00:39:21,340 --> 00:39:25,340 But I think what agents are very good at today is 746 00:39:25,340 --> 00:39:28,020 giving you better information. You know, things like 747 00:39:28,020 --> 00:39:30,420 knowledge retrieval, pulling operational status, 748 00:39:31,140 --> 00:39:33,220 doing analytics, those kinds of things. Those are 749 00:39:33,220 --> 00:39:36,330 things that agents are very capable of doing today, and 750 00:39:36,330 --> 00:39:39,260 a lot of folks have those kinds of agents in use. 751 00:39:39,580 --> 00:39:43,060 The decision-making still falls to your staff and 752 00:39:43,060 --> 00:39:45,980 yourself, but agents let you make better, faster 753 00:39:45,980 --> 00:39:48,960 decisions by giving you better information. where we're 754 00:39:48,960 --> 00:39:52,280 really trying to get to as an industry is agents 755 00:39:52,280 --> 00:39:54,700 actually being able to give you better answers so you 756 00:39:54,700 --> 00:39:57,380 can ask very uh where today you still have to be 757 00:39:57,380 --> 00:40:00,840 very prescriptive saying hey i want you know my top 758 00:40:00,840 --> 00:40:04,600 five most busiest links and we can go you know we can 759 00:40:04,600 --> 00:40:07,240 very quickly pull that pull that stuff up you know 760 00:40:07,240 --> 00:40:09,910 where we want to go is much more broader more 761 00:40:09,910 --> 00:40:13,640 unbounded questions you can ask like hey why is my video 762 00:40:13,640 --> 00:40:16,840 traffic to seattle so choppy which is not necessarily 763 00:40:16,840 --> 00:40:18,900 clear what the cause is or what the resolution 764 00:40:18,900 --> 00:40:21,980 would be and have agents be able to, you know, figure 765 00:40:21,980 --> 00:40:24,140 out what the underlying components of the question 766 00:40:24,140 --> 00:40:28,020 are, go find answers, do reasoning analysis, those 767 00:40:28,020 --> 00:40:30,780 kinds of things, and not just give you information, 768 00:40:30,780 --> 00:40:33,200 but give you recommendations and give you answers. 769 00:40:33,240 --> 00:40:35,940 And some of the work like semantic encoding 770 00:40:35,940 --> 00:40:38,200 and knowledge graphs and just evolution, how 771 00:40:38,200 --> 00:40:40,720 we use LLMs, are really the drives behind 772 00:40:40,720 --> 00:40:43,720 getting us to those kind of operational models. 773 00:40:47,080 --> 00:40:50,800 all right uh you've got one did you want to hit this 774 00:40:50,800 --> 00:40:53,780 slide yeah i'll hit it real quick so one of one of 775 00:40:53,780 --> 00:40:57,600 the reasons you know why agents versus process automation 776 00:40:57,600 --> 00:41:00,300 goes back to the you know we talked about ai being 777 00:41:00,300 --> 00:41:03,340 probabilistic so it's kind of fuzzy and what it knows 778 00:41:03,340 --> 00:41:05,940 and what it doesn't know and one of the nice things 779 00:41:05,940 --> 00:41:09,240 is with regular process automation it's very static 780 00:41:09,240 --> 00:41:12,100 it's very rule-based you know if you say match this 781 00:41:12,100 --> 00:41:14,640 particular set of criteria then it'll knows what to do 782 00:41:14,640 --> 00:41:17,800 and if it doesn't like for an acl for example uh you 783 00:41:17,800 --> 00:41:19,780 know you'll get unpredictable results we'll just 784 00:41:19,780 --> 00:41:23,180 ignore that uh things with the nice thing with agents is 785 00:41:23,180 --> 00:41:27,200 they understand novel situations better so if it runs 786 00:41:27,200 --> 00:41:30,340 against you know comes again across a set of traffic 787 00:41:30,340 --> 00:41:33,720 it's not seen before but it's similar to other 788 00:41:33,720 --> 00:41:36,480 traffic that you've trained it on it knows what to do 789 00:41:36,480 --> 00:41:40,060 it'll make an attempt to you know to block the traffic 790 00:41:40,060 --> 00:41:43,820 or apply qos policy or those kind of things and learns 791 00:41:43,820 --> 00:41:46,120 and adapts. And this is kind of the key advancement 792 00:41:46,780 --> 00:41:50,340 with using agents over things folks are doing today. 793 00:41:51,020 --> 00:41:53,280 All right. So with that, yeah, thanks 794 00:41:53,280 --> 00:41:55,360 for that context, Omar. With that, 795 00:41:55,980 --> 00:42:00,090 the first question we'd asked in the survey was, 796 00:42:01,440 --> 00:42:05,180 how important will agentic AI be in a specific 797 00:42:05,180 --> 00:42:06,980 to your transport network over the next three 798 00:42:06,980 --> 00:42:10,140 years? So you see nearly half, 47%, expected to 799 00:42:10,140 --> 00:42:13,460 be either very important or absolutely critical. 800 00:42:13,670 --> 00:42:16,360 We ask these types of questions a lot. 801 00:42:16,360 --> 00:42:21,280 And when you get that level of percentage in those 802 00:42:21,280 --> 00:42:25,120 higher categories, they're very important or critical. 803 00:42:25,120 --> 00:42:29,680 You know, it's meaningful. It means this is something 804 00:42:29,680 --> 00:42:32,880 that's clearly resonating with the service providers. 805 00:42:32,880 --> 00:42:38,460 So encouraging there, but still certainly very new. 806 00:42:39,720 --> 00:42:43,380 I think maybe, for me, maybe a little 807 00:42:43,380 --> 00:42:45,140 bit more aggressive than I'd expected. 808 00:42:45,280 --> 00:42:48,340 But Omar, and then Marie, I'll let you 809 00:42:48,340 --> 00:42:51,140 comment as well. Omar, is this how you expected 810 00:42:51,140 --> 00:42:53,360 this result to come in, or did you think 811 00:42:53,360 --> 00:42:56,700 there would be more in the critical or? 812 00:42:57,710 --> 00:43:01,300 No, this is about what we predict, right? I mean, I 813 00:43:01,300 --> 00:43:04,380 think this becomes the mechanism for OPEX reduction is 814 00:43:04,380 --> 00:43:07,240 the ability to take tasks that your staff does and 815 00:43:07,240 --> 00:43:10,080 hand them off to agents and you know also the agents 816 00:43:10,080 --> 00:43:12,880 making their staff more effective and more productive 817 00:43:12,880 --> 00:43:15,200 with the things that they're going you know that they 818 00:43:15,200 --> 00:43:18,620 are still doing so this kind of lines up with uh with 819 00:43:18,620 --> 00:43:22,800 our world view okay sounds good marie any any last 820 00:43:22,800 --> 00:43:26,020 comments on this one yeah i i thought it was encouraging 821 00:43:26,020 --> 00:43:28,880 actually like in three years more than that yeah 822 00:43:28,880 --> 00:43:33,340 so um they're moving quick quickly relatively quickly 823 00:43:34,900 --> 00:43:37,340 that is an interesting thing with the three-year 824 00:43:37,340 --> 00:43:42,380 timeline we give um timelines for whatever reason are always 825 00:43:42,380 --> 00:43:46,240 more ambitious than reality uh it tells us a direction 826 00:43:46,240 --> 00:43:50,500 i find but um when i see three in a question i 827 00:43:50,500 --> 00:43:54,180 tend to in my mind instantly translate it to a five-year 828 00:43:54,180 --> 00:43:57,280 it seems to be more realistic so it tells us where 829 00:43:57,280 --> 00:44:01,160 they're going to get um i but you're right they they 830 00:44:01,160 --> 00:44:04,300 tend to not get there as is quickly um challenges always 831 00:44:04,300 --> 00:44:06,140 pop up and actually we're going to get to challenges 832 00:44:06,140 --> 00:44:09,120 in a moment they they tend to be a bit bigger than 833 00:44:09,120 --> 00:44:12,700 um especially with early technologies like this then 834 00:44:13,060 --> 00:44:16,630 then maybe uh what what operators are are anticipating 835 00:44:16,630 --> 00:44:21,620 um yeah so we asked and then there's one last slide 836 00:44:21,620 --> 00:44:25,520 after this and then we'll hit q a um you know which 837 00:44:25,520 --> 00:44:28,800 we asked which agentic ai capabilities are they planning 838 00:44:28,800 --> 00:44:30,640 for the transport network again within the next 839 00:44:30,640 --> 00:44:35,480 three years so kind of the midterm expectation uh and 840 00:44:35,480 --> 00:44:39,540 again so this is not just ai it's specific to agentic ai 841 00:44:39,540 --> 00:44:43,300 assisting with troubleshooting uh live optimization 842 00:44:43,300 --> 00:44:46,880 network performance and even the third one generating 843 00:44:46,880 --> 00:44:50,260 on-demand reports with natural language i would say 844 00:44:50,260 --> 00:44:54,260 are all fairly high for this type of question um and 845 00:44:54,260 --> 00:44:57,260 then you can see uh the the last three kind of fade 846 00:44:57,260 --> 00:44:59,920 away from from there and this was select all that apply 847 00:44:59,920 --> 00:45:04,900 so um you know certainly not all of them resonated uh 848 00:45:04,900 --> 00:45:09,540 but these top three um omar and then uh marie for this 849 00:45:09,540 --> 00:45:12,260 one will let you comment is this what you'd expect 850 00:45:12,840 --> 00:45:15,660 yeah i mean almost everyone i talk to once root cause 851 00:45:15,660 --> 00:45:17,960 analysis and troubleshooting you know it is probably 852 00:45:17,960 --> 00:45:20,540 has the biggest bang for the buck or easiest to 853 00:45:20,540 --> 00:45:23,880 justify roi for for for customers because usually 854 00:45:23,880 --> 00:45:26,880 there's hard tech you know hard financial costs associated 855 00:45:26,880 --> 00:45:29,780 with meantime to repair in time to identify those 856 00:45:29,780 --> 00:45:33,060 kinds of things um but we also see this is probably 857 00:45:33,060 --> 00:45:36,320 the hardest thing to pull off well um it's kind 858 00:45:36,320 --> 00:45:38,200 of interesting that finding documentation is the lowest 859 00:45:38,200 --> 00:45:40,480 on that since that's something that's uh that's 860 00:45:40,480 --> 00:45:43,780 very doable today with rag and those kinds of things 861 00:45:44,400 --> 00:45:50,700 right doable but maybe not as much benefit in in the 862 00:45:50,700 --> 00:45:53,900 end so maybe there was some prioritization uh ROI 863 00:45:53,900 --> 00:45:56,500 kind of in there. Yeah, that caught my eye as well. 864 00:45:56,860 --> 00:45:58,740 Marie, any thoughts on this one? 865 00:45:59,540 --> 00:46:03,920 Yeah, no, just kind of echoing what you guys 866 00:46:03,920 --> 00:46:07,920 said already, that it's really important about 867 00:46:08,200 --> 00:46:13,370 scalability as networks are scaling like ever so much 868 00:46:13,640 --> 00:46:18,480 more. You really need to get a handle on um troubleshooting 869 00:46:18,480 --> 00:46:22,220 and um and network assurance so that that really 870 00:46:22,220 --> 00:46:25,350 lends itself to agentic ai because we do that autonomous 871 00:46:25,350 --> 00:46:27,060 networking you've got these agents that are doing 872 00:46:27,060 --> 00:46:32,230 recently and acting on their own accord so um yeah yeah 873 00:46:36,000 --> 00:46:40,180 sorry as as i said there is a very consistent theme 874 00:46:40,180 --> 00:46:43,440 across all of the data here on on the troubleshooting 875 00:46:43,440 --> 00:46:46,480 and the performance and the reliability of the network 876 00:46:46,480 --> 00:46:50,140 you can really see where where transport is is 877 00:46:50,140 --> 00:46:54,700 focusing with all of these um different but but very 878 00:46:54,700 --> 00:46:59,100 tightly related technologies so i i found that consistency 879 00:46:59,140 --> 00:47:03,840 very um interesting and i think encouraging um Um, 880 00:47:06,180 --> 00:47:07,920 yeah, with that, let me go to the, 881 00:47:07,920 --> 00:47:10,990 this is the last one, uh, we had, um, 882 00:47:11,500 --> 00:47:15,980 and this is concerns around agentic AI and, uh, 883 00:47:15,980 --> 00:47:18,560 for, for transport networks, again, specifically, 884 00:47:18,560 --> 00:47:21,040 you know, this one was, uh, Tim, to me, 885 00:47:21,040 --> 00:47:22,200 interesting as an analyst, we always 886 00:47:22,200 --> 00:47:23,860 like to look at what are the challenges. 887 00:47:24,600 --> 00:47:28,080 Um, and, and I thought, you know, again, to, to my 888 00:47:28,080 --> 00:47:29,880 point I just made about improving accuracy and 889 00:47:29,880 --> 00:47:33,200 reliability driving a lot of what you know the the 890 00:47:33,200 --> 00:47:36,680 drivers for automation and ai and yet when you move 891 00:47:36,680 --> 00:47:40,060 to agentic ai the accuracy and reliability becoming 892 00:47:40,060 --> 00:47:45,020 a top barrier to doing that is an interesting i 893 00:47:45,020 --> 00:47:48,280 think dilemma for the operators and for the technology 894 00:47:48,280 --> 00:47:51,660 um the data privacy and security you know i 895 00:47:51,660 --> 00:47:54,080 wasn't surprised by that that's everywhere to do with 896 00:47:54,080 --> 00:47:57,490 automation and ai you know so transport to that. 897 00:47:58,860 --> 00:48:02,980 But yeah, this accuracy and reliability. 898 00:48:04,280 --> 00:48:07,660 Marie, I'll let you comment on this. 899 00:48:08,240 --> 00:48:12,400 And one thing I would like to delve into a little bit 900 00:48:12,400 --> 00:48:15,960 with you and with Omar is as vendors, how do you 901 00:48:15,960 --> 00:48:21,700 address that? Are these concerns coming up? And are you 902 00:48:21,700 --> 00:48:26,480 as vendors able to um to to address these to to a degree 903 00:48:27,810 --> 00:48:32,020 yeah well on that point accuracy and reliability 904 00:48:32,020 --> 00:48:35,640 like it really all comes down to starting from a um a 905 00:48:35,640 --> 00:48:38,520 solid foundation like clean network data is really 906 00:48:38,520 --> 00:48:43,250 important so that will then yield accurate reliable 907 00:48:43,250 --> 00:48:47,880 results so what we always you know um point our customers 908 00:48:47,880 --> 00:48:51,080 to is is clean up your your network you know we have 909 00:48:51,080 --> 00:48:53,780 network audits etc and you have to spend a fair amount 910 00:48:53,780 --> 00:48:56,880 of time in doing that to get that solid foundation 911 00:48:57,300 --> 00:49:00,600 and then also circling back to network digital twin 912 00:49:00,600 --> 00:49:04,180 right that's another way to get that reliability is use 913 00:49:04,180 --> 00:49:08,680 use that network digital twin capability to test that 914 00:49:08,680 --> 00:49:15,640 to validate those recommended ai solutions before before 915 00:49:15,640 --> 00:49:19,800 you apply them so uh those are two ways that we're 916 00:49:19,800 --> 00:49:23,000 addressing kind of that accuracy and reliability challenge 917 00:49:23,000 --> 00:49:25,610 the other thing in terms of data privacy concerns 918 00:49:26,140 --> 00:49:30,380 the the second highest challenge there is really 919 00:49:30,380 --> 00:49:33,000 enabling customers to use their own private cloud 920 00:49:33,260 --> 00:49:35,340 so when people talk about cloud they typically 921 00:49:35,340 --> 00:49:38,280 think of public cloud infrastructure for llms 922 00:49:38,280 --> 00:49:41,200 for ai but more and more especially the tier 923 00:49:41,200 --> 00:49:44,440 ones they're they're really want to you know 924 00:49:44,600 --> 00:49:47,420 put some guardrails around put some firewalls around 925 00:49:47,420 --> 00:49:50,300 it and move that technology to the private cloud maybe 926 00:49:50,300 --> 00:49:54,900 use slm small language models etc so that that's one 927 00:49:54,900 --> 00:49:58,200 way to address the data privacy and security concerns 928 00:49:59,000 --> 00:50:02,260 yeah we actually we don't have it in this presentation 929 00:50:02,260 --> 00:50:05,680 but um there was a separate question around uh types 930 00:50:05,680 --> 00:50:09,800 of uh approach to um to to uh you know where you 931 00:50:09,800 --> 00:50:12,750 want to house the data private versus public versus hybrid 932 00:50:12,750 --> 00:50:15,760 small language model do you want to make any comments 933 00:50:15,760 --> 00:50:23,460 on on on that particular result uh yeah i um i now 934 00:50:23,460 --> 00:50:25,980 forget what the answer was from that survey question 935 00:50:25,980 --> 00:50:29,640 i think that yeah well the private did did come in 936 00:50:29,640 --> 00:50:32,760 quite strongly um there was also a lot of interest in 937 00:50:32,760 --> 00:50:37,160 in uh in hybrid uh which you know maybe is a bit of a 938 00:50:37,160 --> 00:50:40,720 hedge public was right was minimal small average model 939 00:50:40,720 --> 00:50:43,880 was actually quite minimal too yeah go ahead yeah 940 00:50:43,880 --> 00:50:49,640 yeah so it's again this requires investment uh to uh to 941 00:50:49,640 --> 00:50:52,640 uh house things in in a private cloud infrastructure 942 00:50:52,690 --> 00:50:55,800 uh more investment up front anyways than the 943 00:50:55,800 --> 00:50:58,340 public cloud infrastructure but we yeah we are 944 00:50:58,630 --> 00:51:03,200 in our discussions we're seeing that um that uh move 945 00:51:03,200 --> 00:51:07,500 towards hybrid like you use the private cloud in some 946 00:51:07,500 --> 00:51:10,460 cases public cloud and others and in fact you can 947 00:51:10,460 --> 00:51:13,320 you can have uh one as a backup so to speak a backup 948 00:51:13,320 --> 00:51:17,300 of of the other in case there's some issues with 949 00:51:17,300 --> 00:51:22,000 connectivity or access so yeah it's uh it really nothing 950 00:51:22,000 --> 00:51:25,120 is coming out to the fore as a dominant right now 951 00:51:25,120 --> 00:51:29,120 again everyone's exploring and uh really at the early 952 00:51:29,120 --> 00:51:34,080 stages yeah that's it and you do kind of find that 953 00:51:34,080 --> 00:51:36,820 and when technologies are new it's it's kind of a 954 00:51:36,820 --> 00:51:39,940 buckshot approach so it it and sometimes that's the way 955 00:51:39,940 --> 00:51:46,240 it goes sometimes the industry you know finds a um 956 00:51:46,700 --> 00:51:50,080 finds it you know makes a decision and moves to that so 957 00:51:50,080 --> 00:51:53,980 i would probably say it's too early to tell we're 958 00:51:53,980 --> 00:51:56,320 used to so the small language mod i'm sorry i put you 959 00:51:56,320 --> 00:51:59,700 on that i realize we don't have the data in front of 960 00:51:59,700 --> 00:52:03,700 us but i know the small language model is low um do 961 00:52:03,700 --> 00:52:06,640 you feel operators are kind of missing an opportunity 962 00:52:06,640 --> 00:52:09,470 for small language model or would you expect that 963 00:52:09,470 --> 00:52:14,420 that would be you know more minority niche i think it's 964 00:52:14,420 --> 00:52:17,320 still evolving uh what we're seeing currently like 965 00:52:17,320 --> 00:52:20,840 right now is just really leveraging that um the reg 966 00:52:21,000 --> 00:52:23,440 technology like retrieval augmented generation 967 00:52:23,440 --> 00:52:26,960 right where you you actually tap into this specific 968 00:52:26,960 --> 00:52:30,620 knowledge base applicable to your network before 969 00:52:30,620 --> 00:52:36,360 then using a general llm um so that is working 970 00:52:36,600 --> 00:52:41,040 working well right now um i guess i'd say uh in in 971 00:52:41,040 --> 00:52:44,960 the interim so uh maybe that'll then move to the 972 00:52:44,960 --> 00:52:49,860 small language model later on okay um omar we'll 973 00:52:49,860 --> 00:52:51,740 let you have the final words we do have some good 974 00:52:51,740 --> 00:52:55,220 questions in so i want to get to those um on the 975 00:52:55,220 --> 00:52:57,980 challenges and maybe really kind of homing in 976 00:52:57,980 --> 00:53:03,120 specifically on how cisco as a major supplier is helping 977 00:53:03,120 --> 00:53:07,260 to address some of these around agentic ai sure so 978 00:53:07,260 --> 00:53:09,760 uh you know something i touched on earlier i think 979 00:53:09,760 --> 00:53:12,520 accuracy and reliability a big piece of that sits 980 00:53:12,520 --> 00:53:15,040 with customers right they need to understand what 981 00:53:15,380 --> 00:53:18,460 level model performance is acceptable for them. 982 00:53:18,460 --> 00:53:22,480 So if a model has 90% performance, 90% accuracy, 983 00:53:22,500 --> 00:53:25,050 is that good enough to put in production? But also, 984 00:53:25,620 --> 00:53:27,380 what processes do they put in 985 00:53:27,380 --> 00:53:29,000 place to handle the other 10%? 986 00:53:29,140 --> 00:53:32,500 If you have an employee that's wrong 10% of 987 00:53:32,500 --> 00:53:36,500 the time, you have mechanisms to deal with 988 00:53:36,500 --> 00:53:38,620 that. And you need similar things for the 989 00:53:38,620 --> 00:53:41,190 agents, the technologies that you deploy. 990 00:53:41,500 --> 00:53:46,000 From our perspective, from a feature perspective, 991 00:53:46,000 --> 00:53:49,780 We try and put confidence intervals on any output 992 00:53:49,780 --> 00:53:52,900 that we provide back to a system or back to a user 993 00:53:52,900 --> 00:53:54,760 in terms of, okay, this is the answer and we have 994 00:53:54,760 --> 00:53:57,900 this percent confidence that this is a good answer 995 00:53:57,900 --> 00:54:00,220 and let the human decide whether they're good 996 00:54:00,220 --> 00:54:02,010 with that or they want to do some further research. 997 00:54:02,550 --> 00:54:04,300 Other than that, the kind of usual things like 998 00:54:04,300 --> 00:54:07,420 transparency, things like dry run capabilities, we 999 00:54:07,420 --> 00:54:09,920 talked about digital twin down the road so it's less 1000 00:54:09,920 --> 00:54:11,830 of a roll of the dice from a customer perspective 1001 00:54:12,280 --> 00:54:14,940 if they start deploying some of these technologies. 1002 00:54:15,820 --> 00:54:20,060 All right. So safe to say from your perspective, 1003 00:54:20,060 --> 00:54:22,500 yes, the challenge is there, but they're all 1004 00:54:23,360 --> 00:54:25,860 addressable with the vendors in combination. It 1005 00:54:25,860 --> 00:54:27,300 sounds like things that the service providers 1006 00:54:27,300 --> 00:54:32,520 themselves also need to do to address trust. Yeah. 1007 00:54:32,880 --> 00:54:37,720 Good stuff. Let's jump into the Q&A. We've 1008 00:54:37,720 --> 00:54:40,000 got a few minutes left. We won't get to them 1009 00:54:40,000 --> 00:54:44,360 all, but I know our sponsors will be happy 1010 00:54:44,360 --> 00:54:48,580 to follow up one-on -one on ones we miss. 1011 00:54:49,160 --> 00:54:52,100 You know, this one, Omar, I'll start with you. 1012 00:54:55,660 --> 00:54:58,100 It relates to the agentic AI, and it's 1013 00:54:58,100 --> 00:54:59,600 a question of interest to me as well. 1014 00:54:59,600 --> 00:55:01,440 Question is, are agents always 1015 00:55:01,440 --> 00:55:04,760 based on large language models? 1016 00:55:04,960 --> 00:55:08,520 And really, for me, part of it is teasing out, you 1017 00:55:08,520 --> 00:55:10,560 know, you have AI, you have agentic AI, and you have 1018 00:55:10,560 --> 00:55:12,820 large language models, and I think large language 1019 00:55:12,820 --> 00:55:17,060 models tends to get the bulk of attention. I think 1020 00:55:17,060 --> 00:55:20,100 there's some confusion there. So from your perspective, 1021 00:55:21,210 --> 00:55:23,680 the role of large language models, 1022 00:55:23,680 --> 00:55:25,980 specifically with this agentic AI? 1023 00:55:26,120 --> 00:55:27,720 Yeah, I'll give you an emphatic 1024 00:55:27,720 --> 00:55:30,220 no for a couple of reasons. 1025 00:55:30,220 --> 00:55:34,240 I think agents, you know, AI is a bunch of different 1026 00:55:34,240 --> 00:55:37,380 technologies, just not LLMs. And sometimes other 1027 00:55:37,380 --> 00:55:40,860 technologies are better solutions than LLMs are. a lot 1028 00:55:40,860 --> 00:55:44,080 of traffic network traffic analysis is time series 1029 00:55:44,080 --> 00:55:47,240 analysis um which is kind of machine learning and we 1030 00:55:47,240 --> 00:55:50,720 have lots of proven high accuracy solutions for handling 1031 00:55:50,720 --> 00:55:54,500 that so i think at some point you take a uh you 1032 00:55:54,500 --> 00:55:57,180 know you take a traditional technical analysis and 1033 00:55:57,180 --> 00:56:00,400 not just chase the bright shiny and say okay lms are 1034 00:56:00,400 --> 00:56:04,460 cool but um you know we have better ways maybe not as 1035 00:56:04,460 --> 00:56:07,340 **** but you know more accurate more functional ways 1036 00:56:07,340 --> 00:56:11,240 of doing things i think also based on you know tying 1037 00:56:11,240 --> 00:56:14,120 back to the conversation you were just having you 1038 00:56:14,120 --> 00:56:16,020 know lm might be the better solution but it's not 1039 00:56:16,020 --> 00:56:18,160 going to be an acceptable solution from the security or 1040 00:56:18,160 --> 00:56:21,160 data privacy perspective so then yeah maybe you do 1041 00:56:21,280 --> 00:56:24,100 use more on-prem technologies like you know small 1042 00:56:24,100 --> 00:56:28,520 model run on prime or or stick to uh heuristics or 1043 00:56:28,520 --> 00:56:32,260 algos or machine learning instead i think over time a 1044 00:56:32,260 --> 00:56:35,580 lot of them will move to lms for for other reasons 1045 00:56:35,580 --> 00:56:38,080 and their time series foundation models based on lms 1046 00:56:38,080 --> 00:56:40,240 that are starting to become available but certainly 1047 00:56:40,240 --> 00:56:44,820 today uh often that lms are not the best uh you know 1048 00:56:44,820 --> 00:56:50,120 thing to start off with okay excellent um let's uh 1049 00:56:50,120 --> 00:56:55,120 marie a couple couple came in around digital twin uh 1050 00:56:55,120 --> 00:56:57,700 i'll let you start those and i guess omar if you 1051 00:56:57,700 --> 00:57:02,300 wanted to chime in as well but um the first one that 1052 00:57:02,300 --> 00:57:06,320 that came in was is the network digital twin uh fully 1053 00:57:06,320 --> 00:57:09,040 virtualized and commodity hardware does it require 1054 00:57:09,340 --> 00:57:12,520 gpus and tpus so kind of a hardware question 1055 00:57:12,520 --> 00:57:15,400 around uh digital twin does sienna have any 1056 00:57:15,400 --> 00:57:21,660 perspective on that one yeah it uh it can be just 1057 00:57:21,660 --> 00:57:23,840 virtualized on commodity hardware it doesn't 1058 00:57:23,840 --> 00:57:27,120 require like the the specialized uh processing 1059 00:57:27,120 --> 00:57:30,980 that that we uh now know and expect from uh 1060 00:57:31,100 --> 00:57:34,100 from gen ai tools from large language models like 1061 00:57:34,100 --> 00:57:39,160 gb so so certainly um as as it i guess i'd say 1062 00:57:39,160 --> 00:57:42,060 it's all in the implementation whether or not it's 1063 00:57:42,060 --> 00:57:46,700 actually fully integrated within uh within uh um 1064 00:57:47,380 --> 00:57:50,040 agentic AI framework, I'd say, within the 1065 00:57:50,040 --> 00:57:54,180 model itself. But for now, yeah, certainly it 1066 00:57:54,180 --> 00:57:57,440 doesn't require specialized GPUs or QPUs. 1067 00:57:58,180 --> 00:58:01,040 Okay. And then the other one, 1068 00:58:01,140 --> 00:58:03,560 which is part of a larger 1069 00:58:04,040 --> 00:58:08,780 trend, and it came out in the survey as well, just 1070 00:58:08,780 --> 00:58:12,260 generally about the legacy network and integration with 1071 00:58:12,260 --> 00:58:15,560 existing. This question here was specific to Digital 1072 00:58:15,560 --> 00:58:19,700 Twin, and it was actually asked as well more generally, 1073 00:58:19,700 --> 00:58:23,240 but maybe on the Digital Twin part. For Digital Twin, 1074 00:58:23,500 --> 00:58:26,160 it looks like all network elements should support SDN. 1075 00:58:26,180 --> 00:58:30,260 How do you deal with the legacy network elements in 1076 00:58:30,260 --> 00:58:33,920 trying to build Digital Twin? And maybe that's one 1077 00:58:33,920 --> 00:58:38,480 of the biggest challenges to adoption, why it isn't 1078 00:58:38,480 --> 00:58:42,960 happening super quick but yeah any there yeah it certainly 1079 00:58:42,960 --> 00:58:45,420 is a challenge because those legacy network elements 1080 00:58:45,420 --> 00:58:48,580 you know they did just not as programmable like 1081 00:58:48,580 --> 00:58:52,680 not only do they not share a telemetry northbound in as 1082 00:58:52,680 --> 00:58:56,560 uh i'd say smooth and in real time a manner as as 1083 00:58:56,560 --> 00:59:00,080 modern ones but um they they also aren't quite as 1084 00:59:00,080 --> 00:59:03,660 programmable so certainly there there's um you know specialized 1085 00:59:03,660 --> 00:59:06,500 techniques that you can use to circumvent that 1086 00:59:06,500 --> 00:59:10,380 and to get a to get an accurate view of the network 1087 00:59:10,380 --> 00:59:12,760 again it's not all about real-time streaming you can 1088 00:59:12,760 --> 00:59:16,260 have a network digital twin that has very good capabilities 1089 00:59:16,260 --> 00:59:20,220 that operates on snapshots of data that uh 1090 00:59:20,220 --> 00:59:24,840 snapshots on network data deltas of what's changed etc that 1091 00:59:24,840 --> 00:59:28,200 will then be able to address the use cases and you 1092 00:59:28,200 --> 00:59:31,080 can still run those very valuable what-if scenarios. 1093 00:59:31,900 --> 00:59:34,520 So, yeah, certainly something that has to be taken 1094 00:59:34,520 --> 00:59:37,160 into mind. It is challenging, but it's doable. 1095 00:59:38,300 --> 00:59:39,920 All right. Excellent. We'll 1096 00:59:39,920 --> 00:59:42,140 close out with one last question. 1097 00:59:43,240 --> 00:59:46,820 Omar, I'll let you take it and 1098 00:59:46,820 --> 00:59:49,220 kind of in context, actually. 1099 00:59:50,460 --> 00:59:53,680 We had a question about workforce impacts. And there 1100 00:59:53,680 --> 00:59:59,700 is some anxiety in the market about impacts of AI 1101 00:59:59,700 --> 01:00:01,960 on workforce. It showed up in the survey. There's 1102 01:00:01,960 --> 01:00:06,560 some, not dire concern, but it certainly was a finding. 1103 01:00:06,560 --> 01:00:08,840 The question here is somebody's wondering, how do 1104 01:00:08,840 --> 01:00:12,480 I AI-proof my career? Is there things moving to AI? 1105 01:00:12,840 --> 01:00:17,000 The people on this call, what do they do? 1106 01:00:17,000 --> 01:00:19,880 Omar, I'll let you close out with that one. 1107 01:00:20,320 --> 01:00:22,080 I think the thing to do is 1108 01:00:22,080 --> 01:00:23,700 stay on top of the technology, 1109 01:00:23,780 --> 01:00:26,620 you know, find a couple of blogs, 1110 01:00:26,620 --> 01:00:27,850 find a couple of YouTube channels, 1111 01:00:28,480 --> 01:00:31,080 those kinds of things, and just understand 1112 01:00:31,080 --> 01:00:33,320 and stay on top of what's going on, 1113 01:00:33,520 --> 01:00:34,720 and then try and figure out 1114 01:00:34,720 --> 01:00:36,100 how it applies to your day job. 1115 01:00:36,640 --> 01:00:38,780 You know, I think the trick for the folks who will 1116 01:00:38,780 --> 01:00:41,720 be successful is they can look at what's, you know, 1117 01:00:41,820 --> 01:00:44,020 what the newest technologies are and figure out how to 1118 01:00:44,020 --> 01:00:45,520 apply them or turn around and say, you know what, 1119 01:00:45,520 --> 01:00:48,900 this is not great. this is this will not work in our 1120 01:00:48,900 --> 01:00:51,540 environment for these reasons and that's the kind of 1121 01:00:51,660 --> 01:00:53,960 insights that their leadership will appreciate 1122 01:00:54,730 --> 01:00:58,420 okay uh yeah and certainly one of the um 1123 01:00:58,640 --> 01:01:01,120 the findings throughout was there there is certainly 1124 01:01:01,120 --> 01:01:04,600 um an interest in humans still in in the loop you know 1125 01:01:04,600 --> 01:01:08,000 to that trust and reliability point um with that 1126 01:01:08,000 --> 01:01:11,140 we're a little over so we'll close out here um really 1127 01:01:11,140 --> 01:01:13,880 enjoyed working on this project uh with you marie and 1128 01:01:13,880 --> 01:01:16,720 omar and cisco and and the Ciena teams. Thanks a lot 1129 01:01:16,720 --> 01:01:19,660 for your insights on this call and all the work in 1130 01:01:19,660 --> 01:01:22,660 putting this together. Thanks to the audience for tuning 1131 01:01:22,660 --> 01:01:26,640 in and for your great questions and look for the 1132 01:01:26,640 --> 01:01:28,820 white paper to come out over the next couple of weeks. 1133 01:01:28,820 --> 01:01:30,200 Thanks a lot, everyone. 1134 01:01:30,740 --> 01:01:31,900 Thank you. 1135 01:01:33,860 --> 01:01:34,780 Thanks.