Building the Context Flywheel for AI Data Agents

WEBVTT
NOTE
Transcription provided by Podhome.fm
Created: 07/06/2026 01:22:36
Duration: 3633.772
Channels: 1
1
00:00:11.360 --> 00:00:15.440
Hello, and welcome to the Data Engineering Podcast, the show about modern data management.
2
00:00:16.155 --> 00:00:28.474
Your host is Tobias Maci, and today I'm interviewing Prakulpa Sankar about strategies for building a context flywheel for your data agents. So, Prakulpa, welcome back. And for anybody who hasn't heard your previous appearances, if you can just give a quick introduction.
3
00:00:29.279 --> 00:00:33.199
Yeah. My name is Prapepa. I'm one of the founders of ATLAN.
4
00:00:33.440 --> 00:00:37.360
At ATLAN, we are building a context layer for AI.
5
00:00:37.360 --> 00:00:47.315
I have been fascinated by the topic of how to build shared context for what I used to call the humans of data for now a better part of a decade.
6
00:00:47.875 --> 00:01:01.530
I started as a data practitioner myself, ran these large scale data projects, realized that data engineers, analysts, scientists, machine learning folks, business, what does business even mean? These people all have their own version of context,
7
00:01:01.930 --> 00:01:02.650
which made
8
00:01:02.890 --> 00:01:08.890
answering a simple question like number on a dashboard is broken, we don't know why, really complicated,
9
00:01:08.890 --> 00:01:13.105
like seven systems and five people were involved in answering that question.
10
00:01:14.145 --> 00:01:19.585
ATLAN was founded on the mission of building a shared context and collaboration layer for the humans of data,
11
00:01:20.065 --> 00:01:22.225
which we have now realized as
12
00:01:22.480 --> 00:01:33.360
far more sort of valuable in a world where AI agents are starting to take over some of those tasks, which largely, you know, humans used to be the glue for this context inside organizations.
13
00:01:33.360 --> 00:01:36.994
And so I've been on that journey for now, better part of,
14
00:01:37.395 --> 00:01:40.195
well, this decade with our customers.
15
00:01:40.674 --> 00:01:43.715
And do you remember how you first got started working in the data space?
16
00:01:44.194 --> 00:01:50.060
Yeah. Well, I have been in the data space since I was 21 years old. I,
17
00:01:50.700 --> 00:01:53.100
at the time was doing work in
18
00:01:53.340 --> 00:01:54.939
finance and data.
19
00:01:55.259 --> 00:01:56.380
And I
20
00:01:56.619 --> 00:01:58.539
had done some work with nonprofits.
21
00:01:58.860 --> 00:02:10.665
I said these big problems like healthcare and education and infrastructure, they don't use data and it feels like they should. And let's go do something about that. So my co founder, Byron and I, we founded a company called SocialCops that
22
00:02:10.905 --> 00:02:16.265
was founded on the mission of bringing data science to real world critical problems.
23
00:02:16.505 --> 00:02:17.465
And that meant
24
00:02:17.960 --> 00:02:24.120
at the time going and partner with agencies that already had reach and scale. So it was large governments.
25
00:02:24.120 --> 00:02:27.320
It was the UN, the World Bank, the Gates Foundation.
26
00:02:27.880 --> 00:02:43.725
And we used to be their data team. We used to build data platforms for these problems at scale. And that is really where I learned everything that I learned about building and running data teams and how complex and chaotic it can get to actually make a data project come alive.
27
00:02:44.125 --> 00:02:46.285
And as you mentioned, you started
28
00:02:46.990 --> 00:03:02.110
ATLAN with this mission of collecting all of these disparate data systems and the metadata and semantics around that so that humans could make sense of all of the information that was available and how it was being used and how it can be used.
29
00:03:02.695 --> 00:03:11.255
Obviously, that has had a dramatic shift in terms of the scale and scope and workflow around it with the introduction of all of these agentic capabilities,
30
00:03:11.655 --> 00:03:13.655
especially since the beginning of this year.
31
00:03:14.215 --> 00:03:21.260
And I'm wondering if you can talk through some of the major notable changes that you've seen in terms of the
32
00:03:21.340 --> 00:03:23.100
roles and responsibilities
33
00:03:23.100 --> 00:03:29.100
for these data catalogs and metadata systems as agents become more of the actual
34
00:03:29.100 --> 00:03:29.980
operational
35
00:03:29.980 --> 00:03:30.540
substrate.
36
00:03:31.614 --> 00:03:33.375
What's interesting about
37
00:03:34.015 --> 00:03:36.655
the moment we are in AI, on one hand,
38
00:03:37.535 --> 00:03:39.375
I live in San Francisco
39
00:03:39.375 --> 00:03:43.375
and I walk around the streets and everyone's like, AGI is here.
40
00:03:44.130 --> 00:03:44.770
And
41
00:03:45.730 --> 00:03:54.530
every day the models keep getting better. The benchmarks keep getting better. Two years ago, it couldn't pass the bar. Today, it's the top 1% of test scorers.
42
00:03:54.770 --> 00:03:55.730
In the last decade,
43
00:03:56.305 --> 00:03:58.945
intelligence has compounded 1000x.
44
00:03:58.945 --> 00:04:01.585
In the last six months, intelligence has compounded
45
00:04:02.385 --> 00:04:03.265
2x.
46
00:04:03.905 --> 00:04:05.825
So by any parameter,
47
00:04:06.225 --> 00:04:08.705
exponential growth in every intelligence
48
00:04:08.945 --> 00:04:10.305
benchmark that you can have.
49
00:04:11.180 --> 00:04:16.140
On the other hand, I work in the real world. I work with real companies and real enterprises.
50
00:04:16.460 --> 00:04:20.140
And when I ask them, how are things going? Is AI
51
00:04:20.220 --> 00:04:21.020
useful?
52
00:04:21.500 --> 00:04:23.020
Almost always
53
00:04:23.435 --> 00:04:24.555
eyes drop.
54
00:04:24.715 --> 00:04:29.514
You know, somebody talks about this one pilot that they're making its way to production.
55
00:04:30.074 --> 00:04:30.635
The
56
00:04:30.875 --> 00:04:33.435
data on this could not be more stark.
57
00:04:33.675 --> 00:04:34.634
56%
58
00:04:34.634 --> 00:04:38.235
of CEOs report zero financial benefit with AI.
59
00:04:38.750 --> 00:04:42.190
You know, one of five projects make it to production.
60
00:04:42.590 --> 00:04:45.470
And so there's a gap, whichever way you argue this,
61
00:04:45.790 --> 00:04:47.470
AI being valuable,
62
00:04:47.470 --> 00:04:51.150
we've probably proven that. AI being useful, we haven't really proven that.
63
00:04:51.630 --> 00:04:53.310
And so what's the gap?
64
00:04:54.045 --> 00:04:57.885
And this is where I like to always go back to understanding humans
65
00:04:57.885 --> 00:05:03.085
and what it takes to make humans work before you figure out what it takes to make AI systems work.
66
00:05:03.405 --> 00:05:05.965
And hidden there is in plain sight is the answer.
67
00:05:07.320 --> 00:05:09.880
Companies in the real world care about performance.
68
00:05:10.120 --> 00:05:13.240
Performance is about driving outcomes in the real world.
69
00:05:13.960 --> 00:05:16.600
What is performance a function of? Performance
70
00:05:17.560 --> 00:05:20.360
is a function of, I like to think, intelligence,
71
00:05:21.355 --> 00:05:22.715
but also context.
72
00:05:23.595 --> 00:05:31.515
If you think about human job performance, less than 10% of human job performance is explained by cognitive intelligence.
73
00:05:32.475 --> 00:05:38.130
Which makes sense, right? Like would you say your best employee is also the person that scored the highest on the SATs?
74
00:05:38.370 --> 00:05:40.530
Or would you say they're the person that,
75
00:05:40.770 --> 00:05:48.050
you know, takes feedback the fastest and learns and grows and, you know, just, you know, absorbs institutional
76
00:05:48.050 --> 00:05:53.855
knowledge and, you know, works really hard? Like, is that what the best person in your team looks like?
77
00:05:54.175 --> 00:05:59.294
And almost every time I say that people will say, yeah, of course, that's what the best person on my team looks like.
78
00:05:59.854 --> 00:06:03.055
For the last hundred years, we have built infrastructure
79
00:06:03.055 --> 00:06:08.730
that helps humans become great at their jobs. When they join a company, they are onboarded.
80
00:06:09.370 --> 00:06:10.010
Given
81
00:06:10.410 --> 00:06:11.450
a mentor.
82
00:06:11.770 --> 00:06:18.890
They go and attend and shadow other people in their team. They learn in these conversations.
83
00:06:19.130 --> 00:06:34.870
Their manager gives them feedback. They learn that way. They make mistakes on the job, and then they learn through that. These are all the things that it takes for humans to get really knowledge and skills and expertise that it takes to get really good at their job. And that's what's missing. Today,
84
00:06:35.110 --> 00:06:43.590
have AI agents that are basically the smartest intern we can ever hire, but there is no way to really onboard them onto our context.
85
00:06:43.669 --> 00:06:44.230
And
86
00:06:44.550 --> 00:06:47.110
this is what I think of as the next frontier.
87
00:06:47.535 --> 00:06:51.055
I think the next frontier is contextual intelligence.
88
00:06:51.455 --> 00:06:58.655
It's like, and I truly mean this. I don't think there is a concept of intelligence without it being contextual. It's like an academic construct.
89
00:06:59.535 --> 00:07:08.570
Just measuring cognitive intelligence, it's not really useful in a real world setting doesn't really mean anything. And the next frontier is going to be about how do we actually
90
00:07:08.730 --> 00:07:13.530
take and harness this intelligence by giving it the context that our companies have to make it useful.
91
00:07:14.455 --> 00:07:21.335
And to that point of onboarding and the idea that we've been doing this for a long time, I think that
92
00:07:22.134 --> 00:07:28.615
the fact that we have to do it at such an accelerated pace and over and over again, really stress tests the
93
00:07:29.040 --> 00:07:32.960
systems and protocols that we've had in place to do that onboarding where
94
00:07:33.120 --> 00:07:44.935
because we're working with humans who are able to do a lot of their own discovery and self learning and self direction or find the right person to go out and ask in physical space, we have been fairly
95
00:07:45.014 --> 00:07:45.975
lackadaisical
96
00:07:45.975 --> 00:07:58.455
in terms of the level of rigor and detail that we put into that onboarding setup. And so there has been a lot of delay and pain for newcomers to a team to be able to actually get up to speed and be effective where
97
00:07:58.615 --> 00:08:20.985
it's not unheard of for somebody to say that I don't actually expect anybody to really be fully up to speed for six months after they start, which when you really think about it is fairly insane. And so the fact that we have to do this over and over again and we're bootstrapping these agents from scratch every time forces us to be more deliberate in terms of ensuring that we actually have all of these details and
98
00:08:21.145 --> 00:08:28.745
protocols and information more explicitly codified for the agents to be able to bootstrap, which also helps the humans who are going through that same process.
99
00:08:29.880 --> 00:08:34.120
Yeah, absolutely. In some ways, I think AI has shined the
100
00:08:35.880 --> 00:08:36.840
inefficiencies
101
00:08:36.840 --> 00:08:39.560
that exist in our human organizations
102
00:08:39.720 --> 00:08:41.320
that humans glue together.
103
00:08:42.035 --> 00:08:43.875
And now we realize
104
00:08:43.955 --> 00:08:50.755
that they're missing in a very explicit way. And that really, I do believe there will be a
105
00:08:51.795 --> 00:08:56.280
renaissance of organ like a new kind of learning organization,
106
00:08:56.280 --> 00:08:59.560
you know, like human like, what does it take to build a really successful
107
00:08:59.800 --> 00:09:10.680
company in the AI world will be dependent on how companies find a way to actually build new ways of structuring and building these and encoding these in the way the organizations itself run.
108
00:09:11.455 --> 00:09:12.335
And
109
00:09:12.335 --> 00:09:14.975
for the context of these
110
00:09:15.135 --> 00:09:22.255
metadata layers and the semantic details that we bake into them, I think another piece is that
111
00:09:22.495 --> 00:09:32.790
particularly for human analysts, they're able to do a lot of that information gathering, and there is that implicit knowledge, that institutional knowledge that gets built up over time where
112
00:09:33.029 --> 00:09:47.375
because they've had to run the same query 15 times or because they've had to ask the question of the sales team or the marketing team 15 different times to make sure that everybody's in agreement, they have the necessary context to make sure that the queries that they are producing
113
00:09:47.375 --> 00:10:01.100
are accurate for the organizational context where an LLM might be able to fetch some of that detail and maybe go through query logs of queries that were generated by humans to be able to glean that information.
114
00:10:01.819 --> 00:10:15.365
If you then persist that going forward, it makes everybody's job easier. And so I'm curious how you're seeing the nature and shape of these context layers and the metadata systems evolve to
115
00:10:15.525 --> 00:10:17.220
more explicitly incorporate
116
00:10:17.220 --> 00:10:23.300
and surface those types of details that maybe we wouldn't have bothered investing the resources into
117
00:10:23.700 --> 00:10:26.500
engineering into those catalog systems.
118
00:10:26.900 --> 00:10:29.300
Yeah, absolutely. So let's
119
00:10:29.940 --> 00:10:30.740
unpack
120
00:10:31.305 --> 00:10:35.945
what a human analyst does today and what it takes for them to get really good at their jobs.
121
00:10:36.905 --> 00:10:40.265
Let's say you get a question and the question is something like,
122
00:10:40.825 --> 00:10:43.465
tell me what my top 10 new customers are.
123
00:10:44.380 --> 00:10:46.140
Sounds like a really simple question.
124
00:10:46.699 --> 00:10:53.820
Actually, a really hard question to answer because the first question is like, well, who's asking? What decision are they looking to make?
125
00:10:54.540 --> 00:11:03.825
If sales is asking, then it might be the top 10 customers by revenue. If it is marketing, it's probably the top 10 customers by
126
00:11:03.905 --> 00:11:05.985
brand and logo referenceability.
127
00:11:05.985 --> 00:11:09.825
If it's product, it might be by adoption of the product.
128
00:11:10.305 --> 00:11:26.400
And so even before we get into in the data world, talk a lot about semantics and how do we define a customer and how do we there's actually institutional knowledge that if you're an analyst and someone's asking you a question from the marketing team, you already have some sense of what this means.
129
00:11:27.615 --> 00:11:29.455
Then I use this word new
130
00:11:29.695 --> 00:11:41.930
in my conversation. What are the top 10 new customers? At ATLAN, we run our business on a quarterly basis. So if someone says new customer, it typically means a customer that signed in the last quarter. This is not documented anywhere.
131
00:11:42.089 --> 00:11:45.130
Nobody ever tells anybody this explicitly.
132
00:11:45.529 --> 00:11:51.370
You just learn it because you joined the town hall and you joined the all hands. Like, this just becomes how you think about the business.
133
00:11:52.185 --> 00:11:56.745
And then there is the, what does a customer mean? And how do we define
134
00:11:56.905 --> 00:11:57.785
customer?
135
00:11:57.785 --> 00:12:06.905
And how do we make that explicit? And it's going to be different in finance and it's going to be different in sales, depending on again, are you reporting to the board and or are you running like
136
00:12:07.220 --> 00:12:17.220
an acceleration program or a freemium program or so on? And then eventually we get down to the data itself. And how do I actually measure, like what tables do I use? What descriptions
137
00:12:17.220 --> 00:12:18.420
do I use? And so on.
138
00:12:20.115 --> 00:12:25.475
These are the four layers of first knowledge. This is the first piece that I think of. This is all knowledge.
139
00:12:26.115 --> 00:12:31.315
And there are different types of knowledge in here. There's institutional knowledge. There is semantic knowledge.
140
00:12:31.810 --> 00:12:45.410
There is process knowledge. There is user and org knowledge. This is the first step of knowledge that the human analyst brings together. Then let's make this a little bit more interesting. Typically, the analyst is not asked a question saying,
141
00:12:45.905 --> 00:13:01.170
what are my top 10 new customers? Typically, the analyst is asked a question that says, why did my revenue drop? Or why did my revenue go up? And this is where, for example, there is expertise that people learn. So for example, you learn that you know, July is a seasonal quarter,
142
00:13:01.570 --> 00:13:03.810
especially if you do business in Europe,
143
00:13:04.370 --> 00:13:13.649
because Europe tends to be, you know, on vacation in the summer. So the first thing that an analyst does when a revenue drops is that they first check Europe
144
00:13:14.345 --> 00:13:32.090
and then they say, okay, is this coming because of Europe? And is there a drop just in Europe? Is it seasonal? Is it if it's not if it's seasonal, then this is nothing to worry about. So then they check, okay, well, how are my other regions doing? Have I had a drop in pipeline in other regions? And then they come back with And this is expertise
145
00:13:32.250 --> 00:13:36.570
that humans learn over time in terms of the business and how the business operates.
146
00:13:37.050 --> 00:13:44.715
And that's the second layer of what people need, which is how do I do a what if analysis in my business? How do I think about understanding
147
00:13:44.715 --> 00:14:05.470
the root cause analysis of certain things that happen in my business? How do I think about seasonality trends and so on? And then eventually, then we have the tools. This is really where like, how do I run? How do I go run a query against my CRM? How do I bring a text to SQL together in the right way? This is what I think of as context. Context is a function of knowledge,
148
00:14:05.950 --> 00:14:06.750
skills,
149
00:14:07.070 --> 00:14:09.150
and tools that all need to come together.
150
00:14:09.805 --> 00:14:15.405
The interesting thing about this problem is that we are now at a point where
151
00:14:16.285 --> 00:14:17.485
AI itself
152
00:14:17.485 --> 00:14:18.765
can bootstrap
153
00:14:18.765 --> 00:14:22.205
a large part of the knowledge inside organizations.
154
00:14:22.690 --> 00:14:29.330
So one of the things we have seen a dramatic success in the last roughly, I'd say three and a half months
155
00:14:29.570 --> 00:14:30.290
is
156
00:14:30.450 --> 00:14:32.690
we have unleashed
157
00:14:32.850 --> 00:14:35.570
essentially what we call our context harness
158
00:14:35.895 --> 00:14:37.895
on the foundational
159
00:14:38.135 --> 00:14:44.455
data systems inside Novo. There's already a lot of information about the business that's hidden in the business systems.
160
00:14:45.415 --> 00:14:56.200
If you connect the Salesforce to the warehouse, to the BI and how people are running queries on your BI, You're able to bring that back. You have end to end lineage and then you can start reverse constructing.
161
00:14:56.279 --> 00:15:15.895
How does this business run? How does this business work? How do you think about all of these questions that I already talked about? How do you reverse construct a metrics tree? How do you reverse construct an ontology? Can you start thinking about seasonality trends in the right way? And so the first thing we've seen as an evolution, as we think about the context here is how do you bootstrap?
162
00:15:16.535 --> 00:15:18.375
And how do you bootstrap
163
00:15:18.455 --> 00:15:22.375
context from all the existing business systems that exist inside the organization?
164
00:15:23.470 --> 00:15:25.710
And the second thing that we see is
165
00:15:26.510 --> 00:15:30.510
how do we bring this into the development workflow
166
00:15:30.590 --> 00:15:31.630
of agents.
167
00:15:31.950 --> 00:15:37.070
And so we have now a construct we call a context repo. Think of it as a GitHub repo equivalent,
168
00:15:37.464 --> 00:15:39.305
but for enterprise context.
169
00:15:39.705 --> 00:15:48.345
It becomes the unit of portable context for a use case. And this is where essentially iteration loops, simulation loops,
170
00:15:48.825 --> 00:15:52.425
one of our core philosophies is AI itself should go
171
00:15:53.220 --> 00:15:55.860
do most of the work and ask humans for help.
172
00:15:56.580 --> 00:16:01.140
So can it read through for all of these use cases and say, hey,
173
00:16:02.580 --> 00:16:03.700
you define
174
00:16:04.020 --> 00:16:14.155
your revenue differently in sales and finance. For this use case, which of the two should I pick? Human, tell me. And the human approves or rejects. So how do you build those kinds of loops?
175
00:16:14.395 --> 00:16:16.795
And then turn this into a portable unit
176
00:16:16.955 --> 00:16:18.635
that then builds
177
00:16:19.115 --> 00:16:20.555
on the context over time.
178
00:16:21.540 --> 00:16:33.460
One of the major barriers to entry for adopting a metadata catalog or doing a master data management project or building a business glossary has been the
179
00:16:34.565 --> 00:16:44.885
activation energy required to get to the point where it's actually useful and worthwhile where for the first months or years, depending on how the project is managed,
180
00:16:45.845 --> 00:16:59.430
you're spending a lot of time and energy and potentially money in the process of populating all of this information. But until you hit a particular critical mass where the majority of the information is present and correct and trustworthy,
181
00:16:59.590 --> 00:17:12.315
nobody's going to bother looking at it. And a lot of times, you never get past that point where you've actually reached that threshold. And so it eventually will just quietly die off until the next time around that somebody says, oh, we really need this system,
182
00:17:12.715 --> 00:17:16.154
and then you rinse and repeat. And so I'm curious how you're seeing
183
00:17:16.395 --> 00:17:17.995
these agentic capabilities
184
00:17:18.870 --> 00:17:35.584
reduce the amount of investment and time and activation energy required to get past that threshold where the metadata system or the business glossary or the master data management golden records are to a level of completeness and quality that the business will actually rely on them.
185
00:17:36.304 --> 00:17:40.304
Yeah, absolutely. This has been, I feel like a labor of love
186
00:17:40.544 --> 00:17:44.304
over the last many years for us at ATLAN. So 2022,
187
00:17:44.705 --> 00:17:48.945
in referring to early twenty twenty three, we were the first company to launch ATLAN AI in the category.
188
00:17:49.549 --> 00:17:50.990
And we
189
00:17:50.990 --> 00:17:55.950
kept iterating on it and it was a human assisted workflow to go generate
190
00:17:56.270 --> 00:17:57.710
some of this context.
191
00:17:58.030 --> 00:18:09.605
And we pushed the boundary as much as we could. Like, you know, honestly, were able to get about 75, 80% accuracy on this thing. It was the highest it could go. And that's when we realized that
192
00:18:10.325 --> 00:18:11.605
infrastructure
193
00:18:11.684 --> 00:18:18.725
had never been built for a world where AI was the primary producer of context. Metadata infrastructure itself was not ready for that world.
194
00:18:19.750 --> 00:18:21.990
And so what we did was we actually
195
00:18:22.150 --> 00:18:24.389
rebuilt the foundation of ATLAN completely.
196
00:18:25.030 --> 00:18:30.150
So now ATLAN is built on what we call a context lake house. It's an iceberg native file format.
197
00:18:31.135 --> 00:18:34.254
It allows us to leverage the same compute
198
00:18:34.414 --> 00:18:36.894
that we would leverage on our data systems.
199
00:18:37.535 --> 00:18:45.615
And so all metadata and traversal that an agent on Athlean can do is able to leverage this foundational Lakehouse infrastructure, which means that
200
00:18:46.240 --> 00:18:51.519
you can do compute level operations that you do on Snowflake and Databricks on your data, on your metadata.
201
00:18:51.760 --> 00:19:04.935
That was the first big architectural unlock that we needed to do to be able to get to them. And, of course, as you do that, like how do you think about graphs and how do you think about relationship traversal and how do you think about vectorization
202
00:19:04.935 --> 00:19:11.895
and how do you think about all of these things that it takes for AI to be able to truly work on your context at scale.
203
00:19:13.015 --> 00:19:14.215
That unlock
204
00:19:14.309 --> 00:19:26.309
was massive. Sometime middle of last year was when we announced the context lakehouse, and that was a massive unlock. The second thing that we did was we have done a very, very big investment into our own harness
205
00:19:26.549 --> 00:19:27.590
for context.
206
00:19:28.015 --> 00:19:29.054
And
207
00:19:29.054 --> 00:19:33.134
this is where we have realized context quality really compounds.
208
00:19:34.335 --> 00:19:40.255
So I'll give you an example. Honestly, you and I can today go take a table. We can go give it to ClotCode,
209
00:19:40.255 --> 00:19:50.409
and we can say, describe this table, give me descriptions. And Cloud Code does a kinda like okay job, like decent job, but not really accurate enough for me to push it into production.
210
00:19:51.370 --> 00:19:57.715
But imagine you're able to take the best AI that you have and intelligence you have, and you get it to go read
211
00:19:58.355 --> 00:20:03.075
your system of record and where your data was actually created in your database.
212
00:20:03.315 --> 00:20:11.075
Then you get it to read downstream how everybody is using it and how they're querying it and what is the usage business patterns against it.
213
00:20:11.900 --> 00:20:13.899
You connect all of this together,
214
00:20:14.299 --> 00:20:17.259
and then you have AI write a column description.
215
00:20:17.580 --> 00:20:19.980
A column description is extremely accurate.
216
00:20:20.460 --> 00:20:23.995
Today, are at a point where our customers tell us 89%
217
00:20:23.995 --> 00:20:27.434
of our customers say that it is as good or better than humans
218
00:20:27.915 --> 00:20:32.475
from an accuracy perspective. This was a tipping point. We got this tipping point this quarter
219
00:20:32.715 --> 00:20:34.154
with our context agents.
220
00:20:34.395 --> 00:20:44.889
Now, if I have high quality column descriptions, I can do a really good job with domain tagging. And I can start seeing, okay, these columns and tables belong to sales. These columns and tables belong to Crunch.
221
00:20:45.370 --> 00:20:49.609
Then I can read through the SQL, and then I can start reverse constructing metrics.
222
00:20:50.970 --> 00:20:52.650
And then the accuracy of that's pretty high.
223
00:20:53.365 --> 00:20:58.804
Once I have that and then I have semantics from my BI, I can start reverse constructing relationships.
224
00:20:58.804 --> 00:21:00.565
The accuracy of that's pretty high.
225
00:21:00.885 --> 00:21:04.164
And so what we have done over time is
226
00:21:04.804 --> 00:21:10.019
compound on the quality of this harness. We're now at a point where my customers are telling me to
227
00:21:10.340 --> 00:21:11.220
overwrite
228
00:21:11.220 --> 00:21:12.659
the human written
229
00:21:12.820 --> 00:21:13.620
context
230
00:21:13.620 --> 00:21:26.115
with AI generated context. I think we've really hit this tipping point where the quality of of what AI can do is at a point where it is it's truly production ready. And what I'm most excited by this is exactly what you said, Tobias. The
231
00:21:26.755 --> 00:21:29.475
biggest challenge with these programs
232
00:21:29.715 --> 00:21:47.480
has always been that it's really hard to get the activation energy for humans to go do this work. Everybody believes it's important work. Nobody has the time and the energy to go invest in this work. I should know, I wrote a book about it. Like last year, I wrote this whole book about how
233
00:21:47.480 --> 00:21:55.784
it takes to drive the people and the process that it takes to make this program successful. We studied all these implementations and we were like, this is what it takes to make this successful.
234
00:21:56.585 --> 00:22:02.664
And this year, I believe we're finally at this point where you don't need to do that anymore. You can actually get started,
235
00:22:02.905 --> 00:22:04.264
connect your systems,
236
00:22:04.345 --> 00:22:04.985
run AI,
237
00:22:05.510 --> 00:22:09.990
and within two weeks you can go live. And we're seeing this constantly across our ecosystem
238
00:22:10.390 --> 00:22:15.990
and customers today. And that's what I'm most excited by because context is a living compounding system.
239
00:22:16.870 --> 00:22:22.275
And if AI can do most of the work and only ask humans for help, it changes the dynamics
240
00:22:22.275 --> 00:22:24.274
of what's possible with it.
241
00:22:25.955 --> 00:22:30.674
Would that aspect of being production ready, the trustworthiness,
242
00:22:30.914 --> 00:22:31.715
obviously,
243
00:22:31.875 --> 00:22:34.130
LMs are notorious
244
00:22:34.130 --> 00:22:38.850
for making things up when they are left to their own devices.
245
00:22:39.250 --> 00:22:45.090
Having that context to give it grounding is necessary to avoid those situations.
246
00:22:46.205 --> 00:22:54.764
But then there's still that problem of bootstrapping where maybe you don't have enough context to give it the confidence. And I'm wondering how you are managing
247
00:22:55.405 --> 00:22:56.205
the
248
00:22:56.285 --> 00:23:11.450
bootstrapping of that context and managing the initial trust signals for the LLM to be able to build an appropriate flywheel so that it is building on trusted information rather than just going nuts, collecting everything it can find,
249
00:23:11.995 --> 00:23:17.994
making its own assumptions about details, and then putting that into the catalog that then feeds further
250
00:23:18.154 --> 00:23:19.034
hallucinations
251
00:23:19.034 --> 00:23:20.474
or disinformation
252
00:23:20.635 --> 00:23:36.830
as far as the ways that that data is supposed to be represented or just some of the ways that you maybe need to start with a narrower scope to build that flywheel and just the overall approach for organizations who don't already have an established set of data catalogs and semantics?
253
00:23:37.310 --> 00:23:38.350
Yeah. Absolutely.
254
00:23:38.670 --> 00:23:42.065
So the way I like to define it to customers is
255
00:23:42.225 --> 00:23:43.024
there's
256
00:23:43.585 --> 00:23:47.345
two practices that you have to start in almost in parallel.
257
00:23:47.585 --> 00:23:53.505
The first is what I like to think of as left to right. And that's what I just described to you, right? Connect your data systems
258
00:23:53.820 --> 00:23:56.940
that AI do most of the work to reverse construct your business.
259
00:23:57.259 --> 00:23:58.059
Then
260
00:23:58.620 --> 00:24:02.219
there's a right to left. And I like to align this to
261
00:24:02.539 --> 00:24:03.259
your
262
00:24:03.340 --> 00:24:09.019
use cases that you want to go live with. So let's say I want to pick a
263
00:24:09.985 --> 00:24:14.304
seasonal revenue forecasting agent for my finance team.
264
00:24:15.745 --> 00:24:24.380
Now, what happens here? Why does it take really long to make an agent go live? The first, you alluded to this, it's really hard to bootstrap the context.
265
00:24:24.540 --> 00:24:35.100
Today, it's super easy to actually create an agent. It's really hard to give it the context that it takes to make it go live in production. I've had customers stuck in testing hell for
266
00:24:35.495 --> 00:24:45.255
five months where they are constantly iterating and it's a nondeterministic system. So how do you even test? They know that if they don't get to 80%
267
00:24:45.255 --> 00:24:47.415
accuracy, their business will typically
268
00:24:48.420 --> 00:24:55.940
abandon it. I had a customer the other day who got really excited. They launched about thousand Genie rooms. This is Databricks' AI agent.
269
00:24:56.500 --> 00:25:02.420
And within a month, 90% were abandoned because it didn't have the trust loop, the hallucination,
270
00:25:02.420 --> 00:25:04.784
all of these challenges that we've all seen.
271
00:25:05.905 --> 00:25:15.424
So the first problem is how do you bootstrap and then how do you get it to that accuracy threshold so that you can actually launch it to the business?
272
00:25:15.745 --> 00:25:17.825
Here the approach we've taken is
273
00:25:19.399 --> 00:25:21.159
take the AI first approach
274
00:25:21.399 --> 00:25:29.239
and use all your existing context to reverse construct and bootstrap. So we have a product that we call context engineering workspace.
275
00:25:29.799 --> 00:25:40.414
What it does is think of it as you just come in and define your spec. And you're typically doing this in a cursor or a cloud code or wherever you're building. And you say, this is the agent I'm looking to build.
276
00:25:40.655 --> 00:25:42.255
What it does is it traverses
277
00:25:42.255 --> 00:25:46.815
all the context that you already have and it bootstraps and says, okay, what is
278
00:25:46.830 --> 00:25:51.869
the top context that I need for this use case? So it already does that initial bootstrapping.
279
00:25:52.350 --> 00:25:53.869
The second thing that it does
280
00:25:54.110 --> 00:26:01.150
is it then goes and reads all the context about usage and use cases, and it generates simulation environments.
281
00:26:01.625 --> 00:26:05.145
So the AI itself is going and running simulations on,
282
00:26:05.305 --> 00:26:25.030
okay, well revenue forecasting, let me look at all the business dashboards that are already associated with this. Let me look at how these people are using it. What are the queries they're running? Based on this, let me simulate who are the personas that might be using this. Based on this, let me simulate all the questions they might possibly ask from this agent. Right? So again, this goes back to the context flywheel.
283
00:26:25.110 --> 00:26:29.110
Because you have that base context, you're able to do a much better job of generating these simulations.
284
00:26:29.705 --> 00:26:41.305
Then AI does the first job of basically saying, what do we already have context for? And what's the accuracy? It runs the Text to Sequel. What do we not have context for? So then what it does is it brings it to the humans
285
00:26:41.880 --> 00:26:45.480
and then says, hey, here's the ten, twelve
286
00:26:45.880 --> 00:26:56.680
context pieces we need to add, that we need to improve context for. And this is how essentially our teams improve and run the simulation context engineering workflow to get it to production.
287
00:26:57.875 --> 00:27:02.594
Once you launch into production, one of our core theses is is that
288
00:27:02.835 --> 00:27:05.714
this is a living, breathing thing. This is a discipline.
289
00:27:05.875 --> 00:27:09.554
That with AI, there is no such thing as GA.
290
00:27:10.034 --> 00:27:12.434
You are constantly improving your AI system.
291
00:27:13.020 --> 00:27:16.940
And so the last part of this, and we bring this into the context repo flow,
292
00:27:17.180 --> 00:27:19.260
is we pull traces
293
00:27:19.660 --> 00:27:21.020
and we pull memory.
294
00:27:21.500 --> 00:27:27.660
And then we have AI again, we have a set of curation agents that are sitting on top of traces and evals,
295
00:27:28.144 --> 00:27:32.864
and basically bringing it back and saying, okay, here's how we should improve context over time
296
00:27:33.184 --> 00:27:39.984
for these use cases. So this is really where we have customers who would say things like when we launched this agentic use case,
297
00:27:40.304 --> 00:27:41.504
we started with
298
00:27:41.904 --> 00:27:43.184
50% autonomy.
299
00:27:44.000 --> 00:27:45.839
And now we're going to,
300
00:27:46.240 --> 00:27:52.239
you know, 90% autonomy. And the way we do that is by improving context over time for the system.
301
00:27:54.240 --> 00:27:58.195
With the fact that you have humans who are collaborating
302
00:27:58.195 --> 00:28:00.355
and agents who are collaborating,
303
00:28:00.595 --> 00:28:04.274
what do you see as the value in identifying
304
00:28:04.275 --> 00:28:07.475
the originator of a piece of information? Is
305
00:28:08.130 --> 00:28:27.365
implicitly more trustworthy if it's a human operator who is making a change, or do people typically see that because the agent is going to be more tireless and maybe have more attention to detail that I should trust what the agent puts in and not the humans? It's just some of the ways that you surface that level of detail as far as who provided this information,
306
00:28:27.445 --> 00:28:38.080
at what point, and what is the grounding for that level of detail, and then how you manage the the change process for adding new details or adding new fields or new metadata?
307
00:28:38.880 --> 00:28:39.520
Yeah.
308
00:28:41.280 --> 00:28:44.320
So I think the base starts at just identity
309
00:28:44.560 --> 00:28:46.320
and tracing of identity
310
00:28:46.720 --> 00:28:52.975
and being able to say, okay, this is something that was generated through an AI. This was generated through a human.
311
00:28:54.575 --> 00:29:00.415
I am increasingly beginning to believe that we are actually going to go from human in the loop to human on the loop.
312
00:29:02.255 --> 00:29:03.295
And what
313
00:29:03.295 --> 00:29:04.575
AI is able to do,
314
00:29:05.160 --> 00:29:07.320
humans are just going to be the blocker
315
00:29:07.720 --> 00:29:08.440
to
316
00:29:08.680 --> 00:29:14.600
AI going into production and leveraging the scale. Like, again, you know, we have
317
00:29:14.840 --> 00:29:17.160
AI within a week doing
318
00:29:17.895 --> 00:29:19.095
context generation
319
00:29:19.095 --> 00:29:32.775
for millions of assets. Like there's no way a human is going to be able to like go approve, reject all of this. Right? That's just not possible anymore in the real world. So one of my mental models has become how do you have AI do all of the work
320
00:29:33.920 --> 00:29:38.320
and bring it to humans for governance and decision making. So for example,
321
00:29:38.640 --> 00:29:43.279
one of our most popular agents is an agent we call a metrics conflict resolution agent.
322
00:29:43.840 --> 00:29:52.715
And what it does is it basically goes, reads through all your code and your SQL and all this context that you have. And then it'll basically say, okay, hey, these are the metrics
323
00:29:52.715 --> 00:29:56.474
that you're literally defining differently in two parts of your organization.
324
00:29:57.115 --> 00:29:58.475
And that's
325
00:29:58.475 --> 00:29:59.195
a human decision.
326
00:29:59.880 --> 00:30:01.000
You know, like AI
327
00:30:01.320 --> 00:30:08.600
cannot make that decision for you. And some human in the company needs to decide that when we are reporting to the streets
328
00:30:08.680 --> 00:30:09.480
today,
329
00:30:09.800 --> 00:30:16.275
we decide that we are going to report this versus this, Right? And so that's a human decision and that needs human governance.
330
00:30:16.755 --> 00:30:17.315
But
331
00:30:17.795 --> 00:30:20.195
is this the most accurate description
332
00:30:20.195 --> 00:30:38.890
of this? Or is this how we're defining the metric? Is this the most accurate definition of it? I think AI can actually do a really good job in the right way. Again, like obviously not if you go and just throw this into Claude and expect it to give you a result. But if you build the right kind of harness and the right kind of guardrails and checkpoints and the right kind of AI systems to do it, I think AI can do it pretty well.
333
00:30:39.995 --> 00:30:42.795
The second thing is traceability is really important.
334
00:30:43.435 --> 00:30:49.435
So one of the reasons actually for building the lake house in many ways was actually that versioning,
335
00:30:49.675 --> 00:30:51.035
lifecycle management,
336
00:30:51.115 --> 00:30:59.159
all of these things that we've actually dealt with from a data world perspective will now start becoming very important from a context world perspective. So for example, agent
337
00:30:59.480 --> 00:31:00.679
made a decision,
338
00:31:01.240 --> 00:31:03.240
and it was a wrong decision.
339
00:31:03.880 --> 00:31:11.434
You will want to reverse state. Right? You will want to go back to, okay, which version of the context that did the agent use to make this decision.
340
00:31:11.835 --> 00:31:18.155
And so how do you build these same kinds of systems that we've had in the data management world,
341
00:31:18.315 --> 00:31:27.549
in the context management world, and how do you bring back a lot of those paradigms in the right way, I think will become even more critical. Again, it's a discipline.
342
00:31:27.870 --> 00:31:38.285
And how do we build the disciplines around who's allowed to approve what inside what organization becomes really critical. And we've been talking about data systems, but give you an example in,
343
00:31:38.285 --> 00:31:45.325
you know, a lot of our customers now use our Context Platform, not just for data agents, but also for operational agents.
344
00:31:45.965 --> 00:31:50.285
And so one common example that we'll hear as they manage knowledge and skills is
345
00:31:50.605 --> 00:31:51.245
brand voice.
346
00:31:52.290 --> 00:31:58.050
And the brand voice agent is so important in all these downstream agents like customer support agent,
347
00:31:58.530 --> 00:32:01.250
you know, the SDR agent that's on your website
348
00:32:01.330 --> 00:32:02.290
and so on.
349
00:32:02.850 --> 00:32:04.050
And typically,
350
00:32:04.050 --> 00:32:17.505
there's only one person in the company who's allowed to make a change or approve the Brandvoice skill and the changes that happen in that Brandvoice skill. Right? And so how do you think about these human governance loops inside organizations,
351
00:32:17.505 --> 00:32:18.865
ownership, maintenance,
352
00:32:18.945 --> 00:32:19.825
propagation?
353
00:32:20.240 --> 00:32:28.799
If something changes upstream, how does it propagate downstream? Those are all challenges that context management will solve over the next couple of years.
354
00:32:29.840 --> 00:32:33.360
And as I was preparing for this conversation,
355
00:32:33.360 --> 00:32:39.565
I was looking through some of the materials that you and your team have been publishing, and you make a differentiation
356
00:32:39.644 --> 00:32:40.445
between
357
00:32:40.605 --> 00:32:49.004
the idea of a context layer and the idea of a semantic layer. And I'm wondering if you could unpack some of the different gradations
358
00:32:49.004 --> 00:32:49.644
of
359
00:32:49.810 --> 00:33:06.184
what it means to go from, I have a metadata catalog to I have a semantic layer to I have a context layer. And especially when you're dealing with AI systems where the model itself needs to be something that is reflected in terms of the overall
360
00:33:06.505 --> 00:33:17.179
data catalog and also the fact that a lot of these AI systems maybe require specialized data stores, particularly in the form of vector indices and just how that
361
00:33:17.420 --> 00:33:23.100
layers on different technical elements of what needs to be present to have a fully blown context layer.
362
00:33:23.500 --> 00:33:24.220
Absolutely.
363
00:33:24.460 --> 00:33:25.179
Yeah.
364
00:33:25.420 --> 00:33:26.059
So
365
00:33:26.780 --> 00:33:28.300
the way I like to layer this,
366
00:33:29.315 --> 00:33:34.355
a metadata catalog in its true form was about data context.
367
00:33:34.915 --> 00:33:43.795
So it was saying, okay, my data needs to have context associated with it. So I need to be able to describe my table. I need to be able to describe my columns.
368
00:33:44.320 --> 00:33:48.240
I need to have lineage that's associated with these data objects.
369
00:33:48.960 --> 00:33:51.200
And I need to be able to understand how
370
00:33:51.440 --> 00:33:57.200
my data itself or my data context ecosystem works. And AmeriData Catalog is the foundation for
371
00:33:57.200 --> 00:33:58.640
being able to do that really well.
372
00:34:00.865 --> 00:34:05.184
The layer on top of that is what we think of as the semantic
373
00:34:05.505 --> 00:34:09.825
layer. And the semantic layer is bringing one layer of meaning
374
00:34:09.985 --> 00:34:10.385
to
375
00:34:10.960 --> 00:34:15.680
the data itself. Right? And so this is where we say, okay, well, how do we define customer?
376
00:34:15.840 --> 00:34:17.120
How do we define
377
00:34:17.440 --> 00:34:18.560
new revenue?
378
00:34:19.120 --> 00:34:20.720
And that's both the
379
00:34:20.960 --> 00:34:21.920
conceptual
380
00:34:22.000 --> 00:34:22.560
definition.
381
00:34:23.015 --> 00:34:29.335
This is where, you know, you will hear people use words like business glossary or metrics catalog and so on.
382
00:34:29.735 --> 00:34:35.255
And second, it's the actual, how do I actually calculate this and encode this in semantically?
383
00:34:35.335 --> 00:34:37.655
And eventually, how do I execute this?
384
00:34:37.975 --> 00:34:41.510
So, you know, what is my execution query paradigm?
385
00:34:41.510 --> 00:34:46.150
This is where you'll see Snowflake talks about a concept that they call the Semantic View.
386
00:34:46.390 --> 00:34:52.390
You have some very specific query providers like Kube, for example, that will say, hey, run the query across
387
00:34:52.875 --> 00:34:59.515
five or six of my different tools, my BI tools, my warehouse and so on, right? And so you would see that as the semantic layer.
388
00:35:01.755 --> 00:35:04.715
On top of this, you need to actually add when you think about context.
389
00:35:05.700 --> 00:35:06.580
Context
390
00:35:06.740 --> 00:35:07.460
is
391
00:35:07.700 --> 00:35:10.740
you do need data context and you do need semantic context,
392
00:35:10.900 --> 00:35:13.460
but you also need knowledge context
393
00:35:13.540 --> 00:35:15.540
and you need procedural context.
394
00:35:15.540 --> 00:35:19.540
This is where, you know, your your institutional knowledge,
395
00:35:19.540 --> 00:35:20.180
your skills
396
00:35:20.875 --> 00:35:23.675
associated with it actually start layering on.
397
00:35:23.995 --> 00:35:28.635
And that's when you build a true context layer, where you bring data context,
398
00:35:28.795 --> 00:35:31.275
semantic context, knowledge context,
399
00:35:31.355 --> 00:35:32.235
procedural
400
00:35:32.235 --> 00:35:32.955
context
401
00:35:33.140 --> 00:35:34.020
together
402
00:35:34.180 --> 00:35:36.900
into that one core overall ecosystem
403
00:35:36.900 --> 00:35:38.660
from a context perspective.
404
00:35:38.900 --> 00:35:41.300
This is also where, and you alluded to this,
405
00:35:42.100 --> 00:35:44.180
the primary consumer
406
00:35:44.500 --> 00:35:45.300
of
407
00:35:45.300 --> 00:35:46.820
the context layer
408
00:35:46.820 --> 00:35:48.820
is an AI agent.
409
00:35:50.055 --> 00:35:51.495
And so technically,
410
00:35:51.895 --> 00:35:54.295
this means, well, how does AI traverse
411
00:35:54.855 --> 00:35:56.775
context inside my organization?
412
00:35:56.935 --> 00:36:00.535
This is where, you know, how do I build hybrid search capabilities?
413
00:36:00.535 --> 00:36:02.940
How do I ensure we have vectorization
414
00:36:02.940 --> 00:36:07.340
that's associated with it? How do I build the right kind of frameworks?
415
00:36:07.340 --> 00:36:14.300
At what point? And what's the right protocol, right? This is where protocols like MCP or agent to agent, A2A.
416
00:36:15.005 --> 00:36:22.045
At what point should AI just actually use SQL as an interface and etcetera? So how do you think about these interfaces, the protocols,
417
00:36:22.445 --> 00:36:24.765
the traversal, the graph relationships?
418
00:36:24.765 --> 00:36:28.045
All of these start becoming really critical from a context layer perspective.
419
00:36:29.010 --> 00:36:33.810
Does that help? I think it dubs the mental model of thinking through the, all the layers.
420
00:36:34.290 --> 00:36:39.970
Yeah, it absolutely does. And I think it's also worth exploring a little bit more some of the
421
00:36:40.210 --> 00:36:42.610
details of the maybe
422
00:36:43.065 --> 00:36:47.785
data modeling and semantic modeling that needs to be factored in when you do incorporate
423
00:36:47.785 --> 00:37:10.280
the LLMs and the ML models themselves into the overall catalog and the organizational context to say, okay. Well, I have this model that is being deployed in the context of this agent. This agent is using these data sources to produce these decisions and how that maybe changes some of the ways that the catalog itself needs to be able to store and represent the
424
00:37:10.935 --> 00:37:17.975
primitives of the information that we would typically have just relied on, what are the tables and how are they linked together.
425
00:37:19.335 --> 00:37:21.095
Yeah. Absolutely. And,
426
00:37:21.495 --> 00:37:36.360
I mean, I don't know if this is an evolution of the catalog alog or Semantic Layer or if it's just the new, like, you know, I like to think of the context. Like sometimes I tell my customers, I'm like, if you're using a traditional Catalog or something like just, we can just connect it and we can bring it in into the
427
00:37:36.760 --> 00:37:51.275
context layer. And the context layer becomes your living compounding system for your AI agent. Your catalog just becomes something your humans are continuing to use for as long as humans continue to be the primary user for data analysis purposes, right?
428
00:37:51.835 --> 00:37:53.275
And so I don't know if
429
00:37:53.770 --> 00:37:56.410
the right question is how much
430
00:37:56.890 --> 00:37:58.890
of this needs to go back to the catalog.
431
00:37:59.050 --> 00:38:07.450
I think the right question that I like to think of now is how do you manage context as a living, breathing asset inside your organization?
432
00:38:08.484 --> 00:38:09.445
And there,
433
00:38:09.685 --> 00:38:14.245
I think these concepts of traces and people start using the word decision traces now
434
00:38:14.805 --> 00:38:21.925
are really critical to bring back into the observability loop. So like, I like to think of it as
435
00:38:21.339 --> 00:38:24.220
there's a context repo that is the center of it.
436
00:38:24.780 --> 00:38:25.740
That's the
437
00:38:25.900 --> 00:38:29.180
central point for this AI agent,
438
00:38:29.500 --> 00:38:33.900
which brings knowledge, skills, and tools that that AI agent needs
439
00:38:34.075 --> 00:38:37.115
to do its job really well. Outside of the context,
440
00:38:37.275 --> 00:38:41.755
all the agent needs is it's the model and the harness. These are the three things that make an agent.
441
00:38:42.795 --> 00:38:46.235
And then this context is a living compounding thing.
442
00:38:46.555 --> 00:38:48.235
So it has a maintainer.
443
00:38:49.290 --> 00:38:52.490
There are probably governance loops that are associated with this context.
444
00:38:52.810 --> 00:39:02.410
The context is changing both from the left because your business is not static. So your business is changing. And the context is also improving over time because with traces and observability,
445
00:39:02.715 --> 00:39:08.075
you're able to say, okay, how do I improve this context? So I improve the level of autonomy
446
00:39:08.155 --> 00:39:10.635
as I move towards an autonomous system.
447
00:39:10.955 --> 00:39:12.075
And that's
448
00:39:12.075 --> 00:39:15.755
really the unit of the discipline that needs to get built
449
00:39:16.230 --> 00:39:18.230
as humans go from
450
00:39:19.030 --> 00:39:23.030
being the people who do the job to people who context engineer
451
00:39:23.270 --> 00:39:24.390
these systems.
452
00:39:25.110 --> 00:39:27.430
So these systems can do this, the job
453
00:39:27.670 --> 00:39:29.750
as well as a human can maybe better.
454
00:39:30.935 --> 00:39:32.935
And as you have been
455
00:39:33.175 --> 00:39:37.975
working with your customers, and especially as you have been going through this evolution
456
00:39:37.975 --> 00:39:39.495
from the
457
00:39:39.575 --> 00:39:42.135
first version of the ATLAN product,
458
00:39:42.215 --> 00:39:49.250
we are going to focus on being the source of truth for all of the data that exists in an organizational context
459
00:39:49.250 --> 00:39:55.410
through this disruptive time of AI and agentic capabilities to being this
460
00:39:55.490 --> 00:39:59.835
context layer that also has all of that core
461
00:39:59.915 --> 00:40:02.235
data context and data semantics?
462
00:40:02.235 --> 00:40:07.675
What are some of the most interesting or innovative or unexpected ways that you're seeing teams manage the
463
00:40:07.915 --> 00:40:08.795
creation
464
00:40:08.955 --> 00:40:16.609
and application of the this agentic context and data context and to empower their businesses?
465
00:40:16.849 --> 00:40:20.369
The best use cases we've seen are where
466
00:40:21.329 --> 00:40:22.130
companies
467
00:40:22.130 --> 00:40:24.450
are starting to think about their organizations
468
00:40:24.450 --> 00:40:25.170
as
469
00:40:26.130 --> 00:40:28.930
how do I build platforms for autonomous
470
00:40:28.845 --> 00:40:30.925
organizations or compounding organizations?
471
00:40:31.085 --> 00:40:31.565
So
472
00:40:31.965 --> 00:40:34.205
the way I like to think of it is,
473
00:40:35.565 --> 00:40:38.285
are agents in your company doing
474
00:40:38.365 --> 00:40:43.870
most of the work and asking humans for help? And how do you get there? And
475
00:40:44.190 --> 00:40:53.710
if you think about the paradigm to get there, organizations are at different slopes of majority, right? The starting point is humans are using AI for improving their productivity.
476
00:40:54.474 --> 00:40:55.435
And then
477
00:40:55.994 --> 00:41:03.835
you start having your first few truly agentic use cases go to production, right? And you have a customer support agent that's
478
00:41:03.994 --> 00:41:05.275
doing more work
479
00:41:05.435 --> 00:41:09.355
than the humans that are responding to customers or so on.
480
00:41:10.020 --> 00:41:13.140
And then you start moving to like a truly compounding system
481
00:41:13.300 --> 00:41:17.060
where, you know, you have, you know, tens and hundreds of agents
482
00:41:17.060 --> 00:41:21.940
that are doing work and then coming to humans for help. The most exciting
483
00:41:22.805 --> 00:41:26.085
customers that we have been working with have been really
484
00:41:26.805 --> 00:41:28.165
building the
485
00:41:28.805 --> 00:41:29.765
foundation
486
00:41:29.765 --> 00:41:30.805
of this.
487
00:41:31.445 --> 00:41:33.605
You can call it a frontier company,
488
00:41:33.605 --> 00:41:34.885
autonomous company.
489
00:41:35.950 --> 00:41:39.230
And almost all of them are building this primitive that
490
00:41:39.870 --> 00:41:45.470
I like to think of as the company brain. And what I'm most excited about this company brain
491
00:41:45.630 --> 00:41:48.350
is that it might allow us
492
00:41:48.984 --> 00:41:53.865
to move past the human inefficiencies that existed inside organizations.
493
00:41:53.865 --> 00:41:58.505
I was talking to a large telco customer of ours the other day and
494
00:41:58.505 --> 00:41:59.465
they said
495
00:42:00.105 --> 00:42:01.945
customer
496
00:42:01.490 --> 00:42:02.369
support,
497
00:42:02.530 --> 00:42:05.330
if I really want to solve my resolution time,
498
00:42:05.490 --> 00:42:07.250
I need to solve my network problem.
499
00:42:07.650 --> 00:42:09.410
And these are typically disconnected
500
00:42:09.410 --> 00:42:21.075
problems inside companies, right? Like typically you have resolution that's dealt with in contact center, you have network that's dealt with somewhere else. You have, you know, discounts handled somewhere else.
501
00:42:21.795 --> 00:42:24.994
And most of this is because of the way organizations are created.
502
00:42:25.234 --> 00:42:26.674
And organizations
503
00:42:26.674 --> 00:42:33.250
have these different silos and siloed ecosystems that are built. AI can actually allow us to break the cycle.
504
00:42:34.210 --> 00:42:35.810
If you learn something
505
00:42:35.970 --> 00:42:39.730
from the ground, from a customer perspective, AI can hop
506
00:42:39.970 --> 00:42:40.530
from
507
00:42:40.770 --> 00:42:41.490
the
508
00:42:42.210 --> 00:42:48.285
individual customer use case all the way up to the root cause of that problem and solve that problem.
509
00:42:49.005 --> 00:42:51.085
And, you know, do you even need
510
00:42:51.725 --> 00:43:06.930
do you even need somebody to make a call anymore to report a problem? Can you actually proactively fix the call? And there is a world in which you can see that world happening and not too far away. And I think that can just dramatically change how organizations function
511
00:43:07.250 --> 00:43:14.755
and learn. And so those have been the most exciting use cases where companies are really rethinking their foundational infrastructure how and they operate,
512
00:43:14.995 --> 00:43:15.875
and not just
513
00:43:16.435 --> 00:43:17.875
slapping on AI
514
00:43:18.115 --> 00:43:20.995
on top of existing processes and
515
00:43:21.635 --> 00:43:27.075
work that they're already doing. And in your own work of building and evolving
516
00:43:27.670 --> 00:43:29.270
the Atlin product
517
00:43:29.270 --> 00:43:30.790
and understanding
518
00:43:30.790 --> 00:43:33.430
the needs of the ecosystem
519
00:43:33.430 --> 00:43:37.590
as it continues to shift and grow and change directions,
520
00:43:37.590 --> 00:43:41.430
what are some of the most interesting or unexpected or challenging lessons that you've learned in the process?
521
00:43:43.015 --> 00:43:44.055
The most
522
00:43:44.215 --> 00:43:46.695
interesting and useful one has been
523
00:43:47.175 --> 00:43:50.055
internally transforming into an AI native company
524
00:43:51.015 --> 00:43:54.455
and really pushing the boundary of
525
00:43:54.710 --> 00:43:56.310
the true problems
526
00:43:56.390 --> 00:43:57.030
that
527
00:43:57.750 --> 00:44:02.230
companies face as they work towards building that frontier organization.
528
00:44:02.390 --> 00:44:17.495
Our roots at Atlin has been we were a data team ourselves. We never meant to build a product for anybody else. We tried to buy a product quite honestly, and we couldn't find something that solved our own problems. So we eventually were kind of forced into building Athlean for ourselves.
529
00:44:18.214 --> 00:44:39.865
But that DNA is really important to us because we've always been very close to the problem. One of the reasons why I think we've, we've grown as fast as we have, and we're, you know, the fast growing company in our space and all these other things, there's these accolades that we get. They're all a function of just, we were closer to the problem and we cared very deeply about the problem. And so when AI happened, it was really important for us to go back to
530
00:44:40.185 --> 00:44:48.905
being really close to the problem. And thankfully, we are a company that is at scale. And so we have had the ability to have a front row seat
531
00:44:48.985 --> 00:44:53.970
in transforming how our company operates, right? You know, engineers at ATLAN
532
00:44:54.210 --> 00:44:55.890
don't code anymore.
533
00:44:56.130 --> 00:45:08.825
They only teach AI how to code. Marketers at ATLAN don't run campaigns anymore. They're only allowed to teach AI how to market. And so I think we've been very on the frontier with that. And that has helped us really, I'd say,
534
00:45:09.705 --> 00:45:10.585
grounded
535
00:45:10.585 --> 00:45:12.425
in what are real problems.
536
00:45:12.505 --> 00:45:30.150
How is context management? We started talking about this context thing now, well over a year and a half ago for how, you know, AI systems are going to need context management systems. And the reason we were able to get there ahead of the curve of where the market is, is largely just because we face the problems of
537
00:45:30.390 --> 00:45:34.835
the pain that it takes to productionalize these systems. So that has been, I'd say,
538
00:45:36.194 --> 00:45:40.435
coolest part of this journey. Also seeing how people can become superhumans,
539
00:45:41.075 --> 00:46:06.175
I would say, when people unlock themselves with AI systems. There's so many there's so many stories that we have in this world right now around AI, around, you know, like, what is it going to do? Is it going to take away human jobs? You know, blah, blah, blah. And I can't, you know, I can't prophesize on what will happen in the world in ten years. But what I do know now is that there are enough problems to be solved in the world that we have not found a cure for in humanity.
540
00:46:06.655 --> 00:46:22.180
And humans that allow find a way to turn themselves into superhumans with AI will be key to being able to do that. So just seeing humans unlock themselves has been beautiful to see. Seeing our customers have their own ClaudeCode moments when
541
00:46:22.900 --> 00:46:30.855
they turn on context agents and they tell us things like this is I had a customer last week who said, this is a 404000%
542
00:46:30.855 --> 00:46:33.095
increase to what we did last year.
543
00:46:33.495 --> 00:46:35.335
And so, you just see the
544
00:46:35.735 --> 00:46:36.535
possibilities
545
00:46:36.535 --> 00:46:41.335
of what can happen and you see people come alive. I think that has been really cool.
546
00:46:42.490 --> 00:46:45.130
And the last one, the most unexpected one
547
00:46:45.290 --> 00:46:46.970
has been that
548
00:46:47.530 --> 00:46:50.330
there's so much education to be done right now.
549
00:46:50.970 --> 00:46:53.930
In fact, technology can scale exponentially.
550
00:46:54.250 --> 00:46:56.410
They say organizations scale logarithmically.
551
00:46:57.165 --> 00:46:58.685
But humans,
552
00:46:59.085 --> 00:47:04.445
our slope is not the same. And right now there's so much market noise. There's so much hype.
553
00:47:04.925 --> 00:47:08.285
And so how do you make sure that people are
554
00:47:09.300 --> 00:47:10.260
educated,
555
00:47:10.660 --> 00:47:11.540
authentic,
556
00:47:11.780 --> 00:47:15.060
are able to trust like, you know, I honestly think it's I
557
00:47:15.300 --> 00:47:18.900
would hate to be a buyer of technology right now. Would it
558
00:47:20.580 --> 00:47:24.595
sucks, You know, like everybody is trying to market everything
559
00:47:24.595 --> 00:47:27.955
and it's very hard to like really tell the difference between,
560
00:47:28.115 --> 00:47:30.035
you know, reality and hype.
561
00:47:30.275 --> 00:47:33.315
And my favorite push to
562
00:47:33.315 --> 00:47:35.474
customers has been just like, are you building?
563
00:47:36.890 --> 00:47:39.050
Whoever you are, you could be like
564
00:47:39.050 --> 00:47:41.770
CTO, CIO, CDO, are you building?
565
00:47:42.410 --> 00:47:43.930
Did you build on the weekend?
566
00:47:44.330 --> 00:47:44.810
And
567
00:47:45.210 --> 00:47:46.490
that just changes.
568
00:47:46.570 --> 00:47:57.285
You don't need to believe anyone. You just know yourself on what it takes. And so seeing that also come alive and having a part to play in that has been just amazing.
569
00:47:57.605 --> 00:48:00.005
And I think too that while
570
00:48:00.405 --> 00:48:06.085
the pace of change and the pace of new products hitting the market has been
571
00:48:06.560 --> 00:48:11.520
massively accelerated because of the fact that agentic coding speeds your time to delivery.
572
00:48:11.840 --> 00:48:12.640
It also,
573
00:48:12.800 --> 00:48:14.720
I think, makes it more
574
00:48:15.040 --> 00:48:17.280
tenable to do that evaluation
575
00:48:17.280 --> 00:48:20.720
of technology selection. So for instance, I was recently
576
00:48:21.295 --> 00:48:25.455
looking at, do I want to use Apache Doris or StarRocks,
577
00:48:25.455 --> 00:48:29.615
which originated from the same code base but have diverged in terms of their overall
578
00:48:29.855 --> 00:48:44.150
project goals. And so you can just say, okay. Well, here, Copilot or Claude code. Here are the two code bases. Do an analysis of the actual code that's there, the GitHub issues, the pace of change, the types of projects and pull requests that are being opened,
579
00:48:44.309 --> 00:48:46.630
and give me a comparative analysis
580
00:48:46.630 --> 00:48:54.355
of what these two projects are doing since they have diverged and which one I should be focusing on for this particular use case. And
581
00:48:54.835 --> 00:48:58.515
two years ago, I would never even even bothered really doing that. I would just maybe
582
00:48:58.835 --> 00:49:02.994
read through some of the documentation and make a best guess and hope that I made the right choice.
583
00:49:03.714 --> 00:49:07.555
Yeah. Absolutely. I mean, it's interesting how much
584
00:49:08.500 --> 00:49:10.740
agents might become the primary
585
00:49:11.460 --> 00:49:17.860
buyer of most things in the world. I think we're already seeing this, like if you look at the data from some database companies
586
00:49:18.340 --> 00:49:21.285
and you look at, I'd say, the
587
00:49:21.285 --> 00:49:28.724
percentage of revenue that is starting to get generated directly from just Cloud Code setting up and using databases.
588
00:49:28.724 --> 00:49:33.285
And I'd say that's the first place you would typically see this change is in the developer ecosystem, right?
589
00:49:33.970 --> 00:49:36.850
And I think it will start following in other parts of
590
00:49:37.090 --> 00:49:45.010
the world. And in some ways, I think it would be a it's a positive change, right? Because it allows you to the best product wins,
591
00:49:46.224 --> 00:49:48.945
the most useful thing to the consumer wins.
592
00:49:49.025 --> 00:49:54.705
And I am really excited about by all the new paradigms that that will create in
593
00:49:55.345 --> 00:49:56.785
just the way businesses are on.
594
00:49:58.720 --> 00:50:06.400
And so for people who are interested in figuring out how best to leverage the data that they have,
595
00:50:06.800 --> 00:50:12.320
figure out what data they have, what are the cases where you would argue against this very
596
00:50:12.965 --> 00:50:14.245
agent native
597
00:50:14.245 --> 00:50:23.205
context layer and instead focus on building out the maybe more traditional data catalog, either as a first step or exclusively
598
00:50:23.205 --> 00:50:26.165
maybe for a particular level of scale or complexity?
599
00:50:30.180 --> 00:50:31.380
I would say
600
00:50:32.100 --> 00:50:34.580
if you are a company where you have
601
00:50:35.380 --> 00:50:38.100
or an organization where you have not rolled out
602
00:50:38.980 --> 00:50:40.020
an Ecopilot
603
00:50:41.234 --> 00:50:42.835
to your team
604
00:50:43.234 --> 00:50:44.994
and you're predominantly
605
00:50:44.994 --> 00:50:47.155
in sort of this era of
606
00:50:47.315 --> 00:50:49.795
humans are doing most of the work and
607
00:50:51.394 --> 00:50:53.395
at the best case I want to drive
608
00:50:53.555 --> 00:50:54.595
human productivity
609
00:50:55.490 --> 00:50:58.930
from from this, then I would say focus on
610
00:50:59.569 --> 00:51:01.570
maybe getting your foundational
611
00:51:02.369 --> 00:51:03.650
data catalog
612
00:51:03.890 --> 00:51:14.895
right and stay there. I would still say you should not be doing human data stewardship and you should be using an agent tech way of being able to build it, but maybe just the consumer being
613
00:51:15.055 --> 00:51:24.359
a human is maybe the way I would think about it. However, as I say this, I will also vehemently say, I do not believe that any organization
614
00:51:24.359 --> 00:51:25.960
should choose this strategy.
615
00:51:26.359 --> 00:51:27.000
Because
616
00:51:28.040 --> 00:51:29.240
foundationally,
617
00:51:30.200 --> 00:51:31.640
the bet of data
618
00:51:32.440 --> 00:51:33.800
has always been
619
00:51:34.119 --> 00:51:36.440
that you want to democratize data
620
00:51:36.599 --> 00:51:38.200
to help business.
621
00:51:39.295 --> 00:51:44.095
And this has been the dream, like for the last maybe twenty, thirty years, this has been the dream.
622
00:51:44.575 --> 00:51:47.215
And the challenge that you've always faced
623
00:51:47.775 --> 00:51:49.855
is that in the old world,
624
00:51:50.015 --> 00:51:52.655
you needed to teach business the language of data.
625
00:51:53.900 --> 00:51:58.700
And this was, you know, usually to these data literacy programs and all of these things.
626
00:51:59.260 --> 00:52:03.340
And for the first time with foundation models, you don't need to do that.
627
00:52:03.740 --> 00:52:04.940
You can actually
628
00:52:05.100 --> 00:52:06.940
teach data the language of business.
629
00:52:08.095 --> 00:52:19.135
And so if you think about yourself in the shoes of the business, what does the business care about? The business doesn't actually care about data. The business cares about running the business and doing that really well.
630
00:52:19.695 --> 00:52:20.735
And finally,
631
00:52:20.735 --> 00:52:22.095
from a technology
632
00:52:22.175 --> 00:52:23.455
perspective, we are there.
633
00:52:24.230 --> 00:52:28.230
And I believe that is the dream of any data practitioner
634
00:52:28.309 --> 00:52:35.270
as to why they do what they do. And we are finally at that frontier. And so if you're investing in any technology right now,
635
00:52:36.230 --> 00:52:39.670
there is no downside in just starting at
636
00:52:40.665 --> 00:52:53.225
and leapfrogging the generations. You don't need to go gen one, gen two, gen three. Think you can just leapfrog to gen three and miss the chaos and the trauma that gen one and gen two brought to you and just move directly to gen three and
637
00:52:53.785 --> 00:52:55.385
roll this out in a way where
638
00:52:55.839 --> 00:53:00.400
it will drive a true AI native kind of adoption
639
00:53:00.640 --> 00:53:02.000
inside the organization.
640
00:53:02.240 --> 00:53:11.744
And as you continue to build and iterate on your product, what are some of the predictions that you have for the next set of architectural
641
00:53:11.744 --> 00:53:18.065
shifts that we're going to be seeing in the data ecosystem that are driven by the pressures of AI powered systems?
642
00:53:18.464 --> 00:53:21.585
The first prediction I have is there'll be a lot of change.
643
00:53:21.984 --> 00:53:22.465
And
644
00:53:23.505 --> 00:53:26.510
it'll be almost impossible to predict the change. Change.
645
00:53:28.030 --> 00:53:28.590
But
646
00:53:29.550 --> 00:53:30.910
change is inevitable
647
00:53:31.630 --> 00:53:32.670
and so
648
00:53:33.950 --> 00:53:34.910
organizations
649
00:53:35.390 --> 00:53:36.510
that build
650
00:53:37.470 --> 00:53:38.270
open
651
00:53:38.270 --> 00:53:41.395
non lock in ecosystems that allow
652
00:53:41.555 --> 00:53:42.755
them to evolve
653
00:53:43.075 --> 00:53:47.395
very quickly will be very valuable. Like just let's talk about skills.
654
00:53:48.035 --> 00:53:51.555
Skills as a paradigm did not exist six months ago.
655
00:53:53.270 --> 00:54:01.590
Six months ago. MCP as a paradigm did not exist a year ago, you know. And so we're just dealing with the rapid
656
00:54:01.910 --> 00:54:03.830
pace of change and evolution
657
00:54:03.830 --> 00:54:05.270
harnesses
658
00:54:06.295 --> 00:54:09.975
what I thought was a best in class AI system in
659
00:54:10.935 --> 00:54:13.335
October of last year versus what
660
00:54:13.655 --> 00:54:17.895
now is old school way of building AI systems. It's been six months.
661
00:54:18.990 --> 00:54:19.630
And
662
00:54:19.950 --> 00:54:21.870
while you could feel
663
00:54:21.950 --> 00:54:28.590
some cost bias and say, hey, you know what, like, I just spent all this time building this thing and now I have to rebuild the whole thing.
664
00:54:29.230 --> 00:54:36.125
Or you could say, change is the only reality and I'm going to keep rebuilding and changing along with times.
665
00:54:36.925 --> 00:54:40.045
And so the first prediction I have is that teams that
666
00:54:40.765 --> 00:54:42.045
change fast
667
00:54:42.205 --> 00:54:43.965
and build a culture
668
00:54:44.205 --> 00:54:44.845
that
669
00:54:45.309 --> 00:54:46.109
allows
670
00:54:46.109 --> 00:54:51.150
teams to change quickly and evolve quickly will be teams that win.
671
00:54:51.710 --> 00:54:52.109
And
672
00:54:52.750 --> 00:54:53.550
the second
673
00:54:54.750 --> 00:54:56.510
prediction I have is
674
00:54:56.829 --> 00:55:01.575
there will be a lot of heterogeneity that will get created in the agent tech stack.
675
00:55:02.775 --> 00:55:04.375
There will be, you know,
676
00:55:04.935 --> 00:55:08.935
dozens of agent tech platforms, hundreds of agents inside organizations,
677
00:55:09.335 --> 00:55:11.655
and it will be a new kind of heterogeneity
678
00:55:11.655 --> 00:55:13.415
than we've seen before.
679
00:55:14.650 --> 00:55:17.370
In that world, what I like to tell customers is
680
00:55:17.530 --> 00:55:18.250
context
681
00:55:18.650 --> 00:55:20.250
is your business IP.
682
00:55:20.970 --> 00:55:22.010
Keep it your own.
683
00:55:22.650 --> 00:55:28.330
Like don't lock it in into any individual agent. Don't lock it in into any individual system.
684
00:55:29.115 --> 00:55:32.155
Make sure that it's open, it's interoperable.
685
00:55:32.555 --> 00:55:37.275
We are starting to already see customers tell us about how they've context engineered
686
00:55:37.275 --> 00:55:41.515
these two different agentic platforms differently. Each of these systems have memory.
687
00:55:41.755 --> 00:55:45.570
So now they're all speaking a slightly different version of the truth.
688
00:55:45.970 --> 00:56:00.905
This was a classic data problem. People used to say, you ask sales and finance a revenue number, you'll get two different numbers. This is just happening at far larger scale for truly autonomous kinds of systems already. We're already seeing this happen. And so what you want to ensure is that you have
689
00:56:01.065 --> 00:56:07.385
an open context ecosystem where you're not locked into a single agentic platform or a single model ecosystem.
690
00:56:07.705 --> 00:56:11.320
Yeah, I'd say those are my two biggest predictions for the next,
691
00:56:11.720 --> 00:56:15.000
I don't know, six months. Is that is that a long enough horizon?
692
00:56:15.319 --> 00:56:19.000
I I think that's about as long as anyone can hope to look forward. And
693
00:56:19.960 --> 00:56:28.095
so are there any other aspects of the work that you're doing at ATLAN or this overall space of context layers and the
694
00:56:28.335 --> 00:56:34.575
shared substrate for agentic and human workloads that we didn't discuss yet that you'd like to cover before we close out the show?
695
00:56:35.055 --> 00:56:42.150
Maybe the one last thing is that I am really excited about the future of the data practitioner.
696
00:56:42.470 --> 00:56:43.030
Data
697
00:56:43.270 --> 00:56:46.390
people are the only people inside organizations
698
00:56:46.869 --> 00:56:49.510
that have dealt with non determinism before.
699
00:56:50.150 --> 00:56:50.550
And
700
00:56:51.349 --> 00:56:55.270
I don't think I realized what that means until I truly productionalized
701
00:56:55.270 --> 00:57:01.625
so much of AI. And I realized that, you know, like software engineering, you think in binary, right? Like it's zero one.
702
00:57:02.265 --> 00:57:06.984
And, you know, if you've worked on a machine learning system, you've worked on a data engineering platform,
703
00:57:07.065 --> 00:57:09.545
you have dealt with non determinism.
704
00:57:10.105 --> 00:57:12.870
And there is a foundation
705
00:57:13.030 --> 00:57:13.750
where
706
00:57:14.309 --> 00:57:16.950
data practitioners have an extremely
707
00:57:16.950 --> 00:57:19.430
important role to play in
708
00:57:19.670 --> 00:57:20.870
the AI world.
709
00:57:21.510 --> 00:57:23.510
It will need change, it will need learning,
710
00:57:24.075 --> 00:57:32.395
It will need, you know, there's open questions like who owns AI platforms inside companies? Who owns context platforms inside companies? How does this play out?
711
00:57:32.635 --> 00:57:37.675
But I think there's a lot to learn from what we've just gone through in the last decade of data.
712
00:57:38.330 --> 00:57:38.890
And
713
00:57:39.130 --> 00:57:57.765
I'm really excited about just the role that data practitioners can play in this new world. All right. Well, anybody who wants to get in touch with you and follow along with the work that you're doing, I'll have you add your preferred contact information to the show notes. And as the final question, I'd like to get your perspective on what you see as being the biggest gap in the tooling or technology
714
00:57:57.765 --> 00:57:58.484
or,
715
00:57:58.964 --> 00:58:02.005
I guess, knowledge that's available for data management today.
716
00:58:02.325 --> 00:58:05.125
I'd say the biggest gap is the confusion
717
00:58:05.890 --> 00:58:07.410
more than anything else.
718
00:58:07.970 --> 00:58:09.650
There is so much confusion
719
00:58:09.810 --> 00:58:10.690
right now
720
00:58:11.010 --> 00:58:12.530
in the space.
721
00:58:13.090 --> 00:58:14.290
And I mean,
722
00:58:14.530 --> 00:58:18.369
not to say that there are not technology gaps. Our technology will evolve pretty dramatically,
723
00:58:19.005 --> 00:58:25.165
you know, in the ecosystem over the next couple of years, I'm sure. But I think the biggest gap really is
724
00:58:25.484 --> 00:58:28.765
how do you get a trusted verified
725
00:58:29.164 --> 00:58:30.845
source of context
726
00:58:31.005 --> 00:58:34.365
about what to believe and what to not believe in the ecosystem.
727
00:58:34.900 --> 00:58:37.060
And how do you
728
00:58:37.700 --> 00:58:39.220
make the right decisions
729
00:58:39.380 --> 00:58:40.020
about
730
00:58:40.740 --> 00:58:42.740
tooling technology investments.
731
00:58:43.619 --> 00:58:49.495
So for example, one of the things we've started doing is we've actually been running a WTF is the
732
00:58:49.975 --> 00:58:53.255
context layer series, and we're trying to just demystify
733
00:58:53.255 --> 00:59:00.135
some of what you and I just did, right? Like, you know, what is the difference between a semantic layer and a knowledge graph and a graph database?
734
00:59:00.135 --> 00:59:00.535
And
735
00:59:01.415 --> 00:59:02.855
these are pretty deep
736
00:59:03.750 --> 00:59:04.550
concepts
737
00:59:04.550 --> 00:59:06.230
to go deeper into.
738
00:59:06.310 --> 00:59:07.110
And so
739
00:59:07.350 --> 00:59:11.750
my hope for the ecosystem is over the next couple of years,
740
00:59:11.990 --> 00:59:15.030
we end up seeing a lot more real life implementations
741
00:59:15.510 --> 00:59:16.950
that
742
00:59:16.445 --> 00:59:18.925
create trusted spaces for the community.
743
00:59:19.405 --> 00:59:46.295
All right. Well, thank you very much for taking the time today to join me and share your thoughts and experience on building these shared context layers for enabling agents and humans to collaborate on organizational data problems. It's definitely one of the perennial challenges and definitely one of the faster moving spaces that we have to deal with today. So appreciate all the time and energy that you're putting into making that more attractable, and I hope you enjoy the rest of your day. Thank you.
744
00:59:53.895 --> 00:59:58.150
Thank you for listening, and don't forget to check out our other shows. Podcast.net
745
00:59:58.150 --> 01:00:07.350
covers the Python language, its community, and the innovative ways it is being used. And the AI engineering podcast is your guide to the fast moving world of building AI systems.
746
01:00:07.910 --> 01:00:17.865
Visit the site to subscribe to the show, sign up for the mailing list, and read the show notes. And if you've learned something or tried out a project from the show, then tell us about it. Email hosts@dataengineeringpodcast.com
747
01:00:17.865 --> 01:00:24.105
with your story. Just to help other people find the show, please leave a review on Apple Podcasts and tell your friends and coworkers.