WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=eXA2tjRZIbY

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/eXA2tjRZIbY.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:01.309 --> 00:00:03.309
[music]

s2
00:00:12.880 --> 00:00:14.799
Uh, can everyone hear me?

s3
00:00:14.799 --> 00:00:17.840
Okay, awesome.

s4
00:00:17.840 --> 00:00:20.640
Thank you so much uh for attending the session.

s5
00:00:20.640 --> 00:00:22.000
Um, my name is Vun.

s6
00:00:22.000 --> 00:00:24.320
I'm part of the team at Resolve AI.

s7
00:00:24.320 --> 00:00:26.480
We are building AI for prod.

s8
00:00:26.480 --> 00:00:30.480
uh basically we're building agents to run and fix your software.

s9
00:00:30.480 --> 00:00:40.960
So in this session we'll cover what are the six pillars of an agentic harness that you'll need in a system that is there to fix and run your software.

s10
00:00:40.960 --> 00:00:46.640
Uh before I get into the session just a quick show of hands just to understand for me to understand who's there in the audience.

s11
00:00:46.640 --> 00:00:48.800
I'm assuming most of you are engineers.

s12
00:00:48.800 --> 00:00:51.360
Uh how many of you have been on call before?

s13
00:00:51.360 --> 00:00:53.120
Uh quick show of hands.

s14
00:00:53.120 --> 00:00:54.079
Awesome.

s15
00:00:54.079 --> 00:00:58.320
Have you tried to use AI to help you with your on call?

s16
00:00:58.320 --> 00:01:00.719
Show of hands.

s17
00:01:00.719 --> 00:01:01.920
Awesome.

s18
00:01:01.920 --> 00:01:09.200
For the people who raise their hands, uh can you tell me that 90% of your on call work is being done by AI?

s19
00:01:09.200 --> 00:01:12.400
Uh can you comfortably say that?

s20
00:01:12.400 --> 00:01:12.880
Awesome.

s21
00:01:12.880 --> 00:01:15.680
Yes, that's what we realized when we started the company as well.

s22
00:01:15.680 --> 00:01:17.680
This is a really hard problem to solve.

s23
00:01:17.680 --> 00:01:21.040
So, and in this session, I want to share like our journey.

s24
00:01:21.040 --> 00:01:22.479
What is the evolution we went through?

s25
00:01:22.479 --> 00:01:27.520
what are some of the learnings we had specifically in building such an agentic harness.

s26
00:01:27.520 --> 00:01:33.280
Before I get in um I want to set context on uh what we are experiencing today.

s27
00:01:33.280 --> 00:01:44.720
So the first wave of AI really was focused on coding and uh truthfully I think most of the most for most of us the way we code has fundamentally changed

s28
00:01:44.720 --> 00:01:48.159
and I also wanted to share context on why that is the case.

s29
00:01:48.159 --> 00:01:52.000
Uh firstly code in itself is self-documenting.

s30
00:01:52.000 --> 00:01:58.079
It's easy for AI to actually parse and understand what's there to help you out with the next step.

s31
00:01:58.079 --> 00:02:01.040
Second, inherently code is very modular.

s32
00:02:01.040 --> 00:02:10.560
So it's easy for for AI to actually break it down as chunks and navigate to the next step and also help you extend and build and be build beyond it.

s33
00:02:10.560 --> 00:02:13.920
Third and the most important thing is code is single domain.

s34
00:02:13.920 --> 00:02:19.200
Um you do not need to teach AI multiple domains to reason across them.

s35
00:02:19.200 --> 00:02:25.599
it's easily able to navigate to the next step and that is one of the reasons why we have such defined benchmarks as well.

s36
00:02:25.599 --> 00:02:31.200
You can objectively assess how good is AI in solving this problem for you.

s37
00:02:31.200 --> 00:02:43.680
But in reality for engineering 70% of our time is spent on the other side not generating new software but rather in in production systems basically running and fixing software.

s38
00:02:43.680 --> 00:02:55.920
uh and this ranges through a broad range of issues as well right like operating and running software typically you can actually put it in three broad buckets like one is

s39
00:02:55.920 --> 00:03:10.319
u think of the healthcare analogy that's the easiest way I like to think about it as well one is the regular on call and maintenance think of this as regular small problems that are coming your way maybe you need a Tylenol for uh some

s40
00:03:10.319 --> 00:03:16.879
for some pain that you have maybe you need an Advil so These are the regular on call style alerts that you need to deal with.

s41
00:03:16.879 --> 00:03:21.040
Second category is the incidents where you need all hands on deck.

s42
00:03:21.040 --> 00:03:22.640
It's equivalent to a surgery.

s43
00:03:22.640 --> 00:03:23.840
Systems are shutting down.

s44
00:03:23.840 --> 00:03:29.680
You need everyone's attention who's uh who's involved or responsible to come in and fix it.

s45
00:03:29.680 --> 00:03:33.120
And the third category is more like your daily vitamins.

s46
00:03:33.120 --> 00:03:40.000
So you need to do a lot of things on a regular basis to keep to maintain the health of your system going on.

s47
00:03:40.000 --> 00:03:47.920
So be it things like taking a look at your infrastructure, taking a look at your costs or maybe some platform engineering work that you do to scale.

s48
00:03:47.920 --> 00:03:51.920
So these things are fundamentally hard because they're not exactly like code.

s49
00:03:51.920 --> 00:04:02.080
They cut across multiple different domains, multiple different tools and that is the repercussions that we are starting to see when the euphoria of initial initial euphoria of AI is starting to settle in.

s50
00:04:02.080 --> 00:04:04.480
And that's what you're seeing in the news right now.

s51
00:04:04.480 --> 00:04:15.599
So there's a lot more talk about the number of issues maybe AI generated code is starting to create in production and there is talk around what are the right frameworks to deal with these uh issues in code.

s52
00:04:15.599 --> 00:04:22.400
Second uh the era of bottomless AI is over like the unlimited AI is no longer there.

s53
00:04:22.400 --> 00:04:27.520
You can increasingly see conversations about tokens token optimization token efficiency.

s54
00:04:27.520 --> 00:04:35.440
How do you basically improve those architectures, those rigorous engineering architectures that you need to efficiently and precisely use AI?

s55
00:04:35.440 --> 00:04:47.919
And third, uh this is the conversation that is probably made popular by Satya Adela who's the CEO of Microsoft where he started talking about the architecture you need on top of your regular models

s56
00:04:47.919 --> 00:04:50.800
for any domain specific AI to work.

s57
00:04:50.800 --> 00:04:52.800
So there is a broad classification, right?

s58
00:04:52.800 --> 00:04:58.160
like whenever you look at a generation tasks, you don't have an output or an outcome in mind.

s59
00:04:58.160 --> 00:05:05.280
Uh you don't want you're not specifying to AI exactly how an image should look like, how a blog should look like or how a website should look like.

s60
00:05:05.280 --> 00:05:07.280
So that's where the euphoria sets in.

s61
00:05:07.280 --> 00:05:16.720
But when you think of a flip side of a task where you have a single correct answer that you want AI to get to that is when the disappointment starts to set in which

s62
00:05:16.720 --> 00:05:25.280
in most cases most people dub it as a last mile but that is the single longest mile that you have to run with AI that is why you you need these architectures

s63
00:05:25.280 --> 00:05:28.240
for you to focus on a specific problem.

s64
00:05:28.240 --> 00:05:34.800
Now beyond all this uh operating production systems is also a multiplayer and a multi system problem.

s65
00:05:34.800 --> 00:05:36.560
You're not just doing it on your own.

s66
00:05:36.560 --> 00:05:39.840
You need to pull inmemes for for different teams over time.

s67
00:05:39.840 --> 00:05:43.039
All of us have specialized in different categories of engineering.

s68
00:05:43.039 --> 00:05:44.720
We are put in different teams.

s69
00:05:44.720 --> 00:05:50.160
Uh we need to talk to to solve to resolve any issues across AI.

s70
00:05:50.160 --> 00:05:53.199
So sorry across your production systems.

s71
00:05:53.199 --> 00:05:57.199
All in all uh our company resolve AI started with a single thesis.

s72
00:05:57.199 --> 00:06:08.000
So we genuinely believe that all of this uh all of this work that you do on production systems like fixing issues or like uh running your software on a day-to-day basis should be done by agents in most part.

s73
00:06:08.000 --> 00:06:20.319
That's one of the reasons why I was asking can you confidently tell 90% of your work or more is being done by agents and engineers should just be running those agents where agents take care of like fixing and running your software.

s74
00:06:20.319 --> 00:06:25.360
So that's why we've built uh resolve AI as three categories of agents.

s75
00:06:25.360 --> 00:06:29.199
There are on call agents to fix your software issues that come in on a daily basis.

s76
00:06:29.199 --> 00:06:38.479
And there are incidents agents that will help you drive incident channels and get you get you from a complex incident to a root cause and fix.

s77
00:06:38.479 --> 00:06:42.560
And lastly, there are ambient background agents that will help you with your production tasks.

s78
00:06:42.560 --> 00:06:51.680
Be it things like deployment monitoring or any of the analysis that you have to run or any of the conditions that you trigger where you want a specific investigation to happen.

s79
00:06:51.680 --> 00:07:00.880
underneath all this is the same resolve agent architecture that is what we'll be talking about today uh and what it takes to get there to build it.

s80
00:07:00.880 --> 00:07:02.800
So let's start with a question.

s81
00:07:02.800 --> 00:07:11.280
So why is this even required right like why can't we just point the biggest model at production data models are of course getting better the reasoning capacity is getting better

s82
00:07:11.280 --> 00:07:25.120
uh to fix your software issues it it genuinely demos well whenever you focus on a particular use case and you build your harness on top of it shows up with a very good demo but slowly as you start to scale across use cases and teams

s83
00:07:25.120 --> 00:07:36.240
that is where uh some that is where very familiar cracks start to appear in the system so what are the kind of cracks that we have seen as you start to scale in the product along the way, right?

s84
00:07:36.240 --> 00:07:39.680
First thing you'll notice that of course models will have anchoring bias.

s85
00:07:39.680 --> 00:07:43.039
This is a very strong thing that you have to deal in your harnesses itself.

s86
00:07:43.039 --> 00:07:46.240
They they're designed to give you like a coherent answer.

s87
00:07:46.240 --> 00:07:54.560
Um and first of as and when new models start to come out with better reasoning capabilities, you also need to maintain the treadmill.

s88
00:07:54.560 --> 00:08:07.360
uh you also need to continuously change uh which evaluate firstly evaluate which model is suitable for the task and also update uh the right model for your use case and second and a little bit of an overlooked item.

s89
00:08:07.360 --> 00:08:16.080
So we always associate a model to a workflow or an outcome but every workflow has hundreds of different types of tasks.

s90
00:08:16.080 --> 00:08:25.680
So there are reasoning tasks, there are deterministic tasks that you go through, there are things like image gener image reasoning or you're reasoning across images, you're reasoning across SQL

s91
00:08:25.680 --> 00:08:27.520
or things like logs.

s92
00:08:27.520 --> 00:08:35.919
Now all the models frontier models that you see around each of them is good at a specific part or a specific uh kind of task.

s93
00:08:35.919 --> 00:08:45.519
So the orchestration engine that you need to imagine for uh moving across these models for a workflow first of all needs to keep pace with how models are evolving

s94
00:08:45.519 --> 00:08:53.920
and second it needs to be able to marry the best model for the best task and second failure mode uh is of course the context windows.

s95
00:08:53.920 --> 00:09:04.640
I know the context windows are expanding but context windows context windows are not just the problem uh for for solving for for solving a complex problem like this.

s96
00:09:04.640 --> 00:09:10.480
The second part is if you provide large enough context, models start to overexplore.

s97
00:09:10.480 --> 00:09:15.839
It'll start to come up with hypothesis that don't even exist or sometimes might not even get you to the right answer.

s98
00:09:15.839 --> 00:09:19.839
Now, if you supply very limited context, of course, they underexplore.

s99
00:09:19.839 --> 00:09:29.279
They won't be able to get to they won't actually have visibility into what other pathways exist or probably what other potential hypothesis can exist in solving a problem.

s100
00:09:29.279 --> 00:09:34.720
Especially for production incidents, this becomes a huge issue because telemetry is literally infinite.

s101
00:09:34.720 --> 00:09:38.959
You can start to create as many log lines as you want, as many metrics as you want.

s102
00:09:38.959 --> 00:09:47.440
And now when you start to marry that with your code and infrastructure, that is when real real cracks in system uh starts to appear.

s103
00:09:47.440 --> 00:09:52.959
Third and my and my most favorite one is uh defining causal reasoning.

s104
00:09:52.959 --> 00:09:55.040
So models are there to please you.

s105
00:09:55.040 --> 00:10:01.040
I mean, we've all come across these use cases where you push the model hard enough, uh, it'll start to agree with you in every different direction.

s106
00:10:01.040 --> 00:10:06.720
So, models are specifically designed to give you coherent answers, but not causality.

s107
00:10:06.720 --> 00:10:12.640
But production incidents, the only thing you're looking for is a causal chain of evidence like a detective.

s108
00:10:12.640 --> 00:10:20.959
What are the steps that took place that led to a particular incident so that you can actually fix uh the right issue instead of creating a patch.

s109
00:10:20.959 --> 00:10:31.839
So especially when you're solving a hard problem like production reasoning you need causal reasoning built into the mod built into the system not just coherent answers.

s110
00:10:31.839 --> 00:10:33.440
And the next one is of course guardrails.

s111
00:10:33.440 --> 00:10:42.560
I won't harp on this topic too much but all of us have read uh some news or the other where let's say AI is going and deleting a particular file system or like a database.

s112
00:10:42.560 --> 00:10:47.360
Uh so AI it's also this is a feature right not a bug.

s113
00:10:47.360 --> 00:10:53.760
uh AI might have determined that the cleanest possible fix is just deletion of that particular code snippet.

s114
00:10:53.760 --> 00:10:55.040
You can't fault it for it.

s115
00:10:55.040 --> 00:11:00.240
You just have to build guard rails uh around the system so that it works in the right fashion.

s116
00:11:00.240 --> 00:11:04.800
What is the least privilege access that AI can operate with in any given scenario?

s117
00:11:04.800 --> 00:11:07.040
And lastly, learning loops.

s118
00:11:07.040 --> 00:11:12.560
Any system that you build should be scalable across your team or your organization in the best case.

s119
00:11:12.560 --> 00:11:18.320
Every time you're using AI for production incidents, like I said at the starting, it is a multiplayer problem.

s120
00:11:18.320 --> 00:11:30.160
You have multiple different teams like SR, platform teams, uh backend engineers or service engineers, all of these people should be involved and most importantly all of these people should have the same context.

s121
00:11:30.160 --> 00:11:42.640
Now when you add temporal discussions on top of it like uh your AI system should have the same context of a previous investigation in the 100th investigation or else you're starting the wheel from the same time again.

s122
00:11:42.640 --> 00:11:46.000
So these are the failure modes where we've seen the cracks appear.

s123
00:11:46.000 --> 00:11:52.640
So that's why we've actually built an agent architecture that answers these same questions exactly.

s124
00:11:52.640 --> 00:12:01.440
So there are six major parts uh for any AI architecture that has to work in a specific domain.

s125
00:12:01.440 --> 00:12:06.800
The reason I'm mentioning it as a specific domain is models are there for general purpose reasoning.

s126
00:12:06.800 --> 00:12:12.959
Now when you want to harness these models into a specific answer, you need to take care of six different pillars.

s127
00:12:12.959 --> 00:12:15.760
The first one is model orchestration.

s128
00:12:15.760 --> 00:12:18.800
Like I mentioned, uh model orchestration is two layers.

s129
00:12:18.800 --> 00:12:22.320
one, how do you keep up with the model treadmill with the latest models?

s130
00:12:22.320 --> 00:12:25.839
And second, how do you match the best model for the task?

s131
00:12:25.839 --> 00:12:34.240
Is it Gemini for im uh image specific reasoning or is it OpenAI for deterministic steps or is it claude for open-ended investigations?

s132
00:12:34.240 --> 00:12:40.959
You need to constantly keep evaluating what is the best uh possible model for the task at that point in time.

s133
00:12:40.959 --> 00:12:44.959
So this is an engine that we've built uh up front very early.

s134
00:12:44.959 --> 00:12:46.880
Second is context engineering.

s135
00:12:46.880 --> 00:12:53.440
Now context engineering is often minimized to like uh the database or an execution solution.

s136
00:12:53.440 --> 00:12:54.959
Oh, are you using a graph rag?

s137
00:12:54.959 --> 00:12:56.160
Are you using a knowledge graph?

s138
00:12:56.160 --> 00:12:57.600
Are you using x y and z?

s139
00:12:57.600 --> 00:12:59.680
Those are all implementation details.

s140
00:12:59.680 --> 00:13:06.079
But what matters as a goal is what is the precise amount of context AI needs to solve that specific problem.

s141
00:13:06.079 --> 00:13:10.160
And more often than most often than not, it's a combination of techniques.

s142
00:13:10.160 --> 00:13:28.560
you might have to use a graph rag kind of a solution for AI to start investigations and when you actually further go into the next steps you need to determine you need to define very precise tool calls so that you don't burn through tokens very quickly when you're making multiple tool calls in in form of log queries

s143
00:13:28.560 --> 00:13:36.880
in form in form of metrics or dashboards or code queries etc. The third one is causal reasoning like I mentioned.

s144
00:13:36.880 --> 00:13:47.680
So this is one of the first uh principles that we've built into our system at resolve AI where uh the evident the root cause is always based on a causal chain of evidence.

s145
00:13:47.680 --> 00:13:51.920
What are the exact or precise steps that happened that led to this particular issue?

s146
00:13:51.920 --> 00:14:00.880
If you're not able to establish that chain of evidence, we have to give out that information with a low confidence level and point the user in a different direction.

s147
00:14:00.880 --> 00:14:12.320
This is true for any domain specific AI and this is true for resolve AI as well where we are pointing you in a different direction if we don't see the if we don't see the data beyond a particular point.

s148
00:14:12.320 --> 00:14:14.560
The next one like I said is governed actions.

s149
00:14:14.560 --> 00:14:16.240
This is quite straightforward.

s150
00:14:16.240 --> 00:14:26.880
Uh it depends upon every team, every different organization uh depending on uh depending on the access levels or probably the restrictions that you have or the industry you operate in.

s151
00:14:26.880 --> 00:14:30.800
What are the specific guardrails that you need to define for AI to operate under?

s152
00:14:30.800 --> 00:14:31.519
Is it read?

s153
00:14:31.519 --> 00:14:32.959
Is it right and under right?

s154
00:14:32.959 --> 00:14:34.639
What are the specific conditions for right?

s155
00:14:34.639 --> 00:14:38.800
Uh etc., etc. And lastly, a learning system.

s156
00:14:38.800 --> 00:14:40.160
So, think of it like this.

s157
00:14:40.160 --> 00:14:44.639
Every interaction that uh you're doing with an AI system is an opportunity for learning.

s158
00:14:44.639 --> 00:14:52.800
So, uh AI system has to learn not just from an investigation it is going through, but from a how from how a user is actually interacting with it.

s159
00:14:52.800 --> 00:14:54.880
Like am I giving you positive reinforcement?

s160
00:14:54.880 --> 00:14:57.519
So is this turning into a positive eval?

s161
00:14:57.519 --> 00:14:59.279
Am I giving you some negative reinforcement?

s162
00:14:59.279 --> 00:15:03.519
Am I am I guiding you or steering the investigation in a specific direction?

s163
00:15:03.519 --> 00:15:05.760
Those are different kinds of evals.

s164
00:15:05.760 --> 00:15:10.800
So that leads into the last piece which where eval is also a very critical step.

s165
00:15:10.800 --> 00:15:14.800
So this is the starting point and ending point for any agent architecture.

s166
00:15:14.800 --> 00:15:18.720
So the way we think about evals is in five different levels like I mentioned.

s167
00:15:18.720 --> 00:15:20.959
What is the positive reinforcement you can give?

s168
00:15:20.959 --> 00:15:23.120
What is the negative reinforcement you can give?

s169
00:15:23.120 --> 00:15:24.800
And can we trace your path?

s170
00:15:24.800 --> 00:15:26.800
How did you come up with this solution?

s171
00:15:26.800 --> 00:15:29.279
Can we score you like an engineer based on that?

s172
00:15:29.279 --> 00:15:32.480
How would the how would a best engineer perform this investigation?

s173
00:15:32.480 --> 00:15:34.399
And how do you compare against it?

s174
00:15:34.399 --> 00:15:38.480
And on top of it, how are you calibrating yourself as as AI?

s175
00:15:38.480 --> 00:15:40.639
How are you giving out that confidence level?

s176
00:15:40.639 --> 00:15:54.639
So all these are systematic eval infra evals that you build into one platform that you constantly have to push the architecture against whenever there's a new model, whenever there's a new use case or whenever there's an architecture change.

s177
00:15:55.600 --> 00:15:56.320
Cool.

s178
00:15:56.320 --> 00:15:57.279
There's enough talking.

s179
00:15:57.279 --> 00:16:01.199
So let me actually show you this in action on how it works in resolve.

s180
00:16:01.199 --> 00:16:04.959
Like I mentioned our platform uh is specifically there.

s181
00:16:04.959 --> 00:16:07.040
So there are two categories of agents that we work with.

s182
00:16:07.040 --> 00:16:11.040
One is an on call agent, one is an incident agent and of course the background agent.

s183
00:16:11.040 --> 00:16:16.639
So uh since like I've since most of you are engineers over here, you might be familiar with the system.

s184
00:16:16.639 --> 00:16:25.279
Of course most of your telemetry or observability uh alerts are firing into a collaboration tool like Slack or Microsoft Teams or whatever you might be using.

s185
00:16:25.279 --> 00:16:31.040
In this case, what we are seeing is Graphfana fire firing alerts into our common Slack channel.

s186
00:16:31.040 --> 00:16:36.320
Um so Resolve automatically picks up those alerts and kickstarts an investigation.

s187
00:16:36.320 --> 00:16:46.079
What you see here is uh it is throwing an error about a log skill and it starts to provide what is the what are the uh what is the root cause that it found for the

s188
00:16:46.079 --> 00:16:51.440
uh alert that came in and also the other contributing factors or other theories it rejected.

s189
00:16:51.440 --> 00:16:55.120
It gave a quick root cause and evidence and recommendations.

s190
00:16:55.120 --> 00:16:56.399
But let's actually deep dive.

s191
00:16:56.399 --> 00:17:00.399
So in most cases like I said this should be a multiplayer experience.

s192
00:17:00.399 --> 00:17:04.400
So let's actually go into the UI and see what's happening here behind the scenes.

s193
00:17:04.400 --> 00:17:12.400
So uh it the work starts off from resolve AI when it treats the alert that's coming in as a prompt in itself.

s194
00:17:12.400 --> 00:17:20.079
So it takes the prompt and it spins up a bunch of different agents actually two different categories of agents uh to get the work done.

s195
00:17:20.079 --> 00:17:22.880
So first category of agents are basically investigators.

s196
00:17:22.880 --> 00:17:35.520
So they are uh trained to actually uh go through the investigation just like how an engineer would firstly find out what is the issue gather more information from metrics and traces etc and figure out where the issue is happening from traces

s197
00:17:35.520 --> 00:17:45.280
logs and now start to correlate that with things like change events uh or any of the new deployments that would have gone in and your code and your infrastructure.

s198
00:17:45.280 --> 00:17:50.160
So primarily it's doing this translation or reasoning across four types of data sources.

s199
00:17:50.160 --> 00:17:55.919
your code, your infrastructure, your knowledge bases and your observability platforms.

s200
00:17:55.919 --> 00:18:06.400
Based on all that it starts to give out this uh root cause or this root cause what we've seen there are so one of the log skill is actually experiencing high high failure rate.

s201
00:18:06.400 --> 00:18:12.000
It trace the issue all the way down to some stale integrations that are still live in the system.

s202
00:18:12.000 --> 00:18:22.240
The interesting part is it's also showing out some other ruled out theories where generally we would start an investigation at the same time there was a GCP outage.

s203
00:18:22.240 --> 00:18:29.679
So you would actually uh this this would be a perfect correlation as well right going back to AI that is designed to give you coherent answers.

s204
00:18:29.679 --> 00:18:31.679
So a GCP outage happened at the same time.

s205
00:18:31.679 --> 00:18:40.400
So of course this might be the issue uh that we are seeing or there are other traffic spikes that it is experiencing uh without missing integration.

s206
00:18:40.400 --> 00:18:44.080
So uh that would be the other thread that you would start to pull on.

s207
00:18:44.080 --> 00:18:54.320
So let's say if I'm a new engineer, if I'm starting in okay resolve gave uh this as a root cause of course it's we have also designed to go operate with the least trust policy.

s208
00:18:54.320 --> 00:19:02.080
So let's say if I I can start to highlight and it'll start to it'll pull me in so that I can interrogate this further.

s209
00:19:02.080 --> 00:19:21.120
uh in this case let's say uh let's ask resolve uh is this related to uh GCP outage so it's giving out the answer here but there is other other precise questions that I ran I want to share with you so there was other deployment

s210
00:19:21.120 --> 00:19:32.160
errors that happened at the same time so that is what I was pushing resolve on going back to what I mentioned if you push uh an AI hard enough it'll start to agree with you we've designed the system to go against it.

s211
00:19:32.160 --> 00:19:36.559
It'll only base its answers based on the causal causal chain of evidence that it created.

s212
00:19:36.559 --> 00:19:41.360
Since it is able to establish that, it's able to give out a very coherent answer.

s213
00:19:41.360 --> 00:19:47.760
So, this uh also lets me actually pull in my teammates so that this truly becomes like a multiplayer experience.

s214
00:19:47.760 --> 00:19:55.679
Let's say if I give uh I can pull in, can you take a look at this?

s215
00:19:55.679 --> 00:20:01.919
This can pull in my teammate directly from Slack so that we can collaborate on this virtual war room experience.

s216
00:20:02.559 --> 00:20:16.320
This was a quick demo like I said uh the currently and what I talked about today is just one uh axis of the architecture right like now there is the other axis of the architecture where you just need to productize an AI

s217
00:20:16.320 --> 00:20:28.799
like if you want to put this in uh in front of any user or a customer you need to make sure it runs on a robust platform those are the other aspects that you have to think through but this is something that we are always familiar with

s218
00:20:28.799 --> 00:20:30.960
so that's been my time.

s219
00:20:30.960 --> 00:20:40.799
And uh if you're interested to learn more about uh about Resolve AI and which customers are currently using us and how are they using it, uh please check us out at the booth L28,

s220
00:20:40.799 --> 00:20:41.600
I believe.

s221
00:20:41.600 --> 00:20:46.880
Uh and by the way, before you leave, take out those cool bags uh that are right outside.

s222
00:20:46.880 --> 00:20:49.360
Thank you so much.

s223
00:21:02.199 --> 00:21:04.199
[music]