AI Engineer World's Fair 2026
Knowledge Systems: The New GTM Stack — Jeffrey Wang, Exa
Read the talk
Knowledge Systems: The New GTM Stack
Jeffrey Wang explains how Exa turns market discovery, customer activity, company knowledge, and past decisions into a go-to-market system that people and agents can use—then examines the interfaces, permissions, and organizational roles needed to operate it.
From a talk by Jeffrey Wang
At a glance
Ideas worth remembering
Go-to-market can be modeled as a data system connecting product knowledge, internal customer behavior, and changing external market information.
Exa pairs stable interfaces with flexible agents: the ICP dashboard maps potential customers, Request Lens exposes behavioral signals, and Slack agents investigate accounts and build demos.
Jeffbot treats personal replication as three problems: writing style from 760 emails, judgment calibration from hundreds of past decisions, and controlled access to company systems.
Agent-first systems need programmatic data access, but stable graphical interfaces remain useful for repeated tasks that benefit from a familiar user experience.
Buy versus build is better evaluated through customizability: Exa retains Salesforce’s existing sales model while giving agents programmatic access through MCP.
Agent permissions should depend on the caller. Wang’s Jeffbot invocation can read and write broadly, while other employees receive drafting capabilities and a reduced tool set.
FDEs can shorten the loop between customer work and internal tooling, but Wang expects specialization may become necessary as Exa grows beyond its current model.
Treat go-to-market as a live data problem
Engineers naturally want to keep improving the product. Wang says Exa followed that instinct early on and consequently did too little marketing and sales. His correction is blunt: a company must both build something good and get it into people’s hands. AI makes the second job more approachable to engineers because research, account preparation, and other go-to-market work can now be designed as software systems rather than handled only through manual processes.
Customer research, target identification, contact discovery, and proofs of concept look like separate sales activities, but they share a dependency: the team needs information about what the product does and which prospective customers might need it. Wang therefore frames go-to-market as a data problem. The goal is to maintain a live model of the company’s world that agents can act on, rather than repeatedly assembling account context from scratch.
That model combines two broad classes of information:
- Internal data: customer records, employee knowledge, and product-usage behavior.
- External data: companies, people, news, and other events occurring outside the business.
Wang cites more than 60 million companies worldwide and more than a billion people on LinkedIn to convey the scale of the external side. These figures motivate broad discovery infrastructure; they do not establish that Exa’s internal model covers every company or person.
Exa began approaching its own go-to-market operation from an agent-first perspective after launching in the middle of 2023. Wang organizes the resulting system around two interfaces and two kinds of agents. The interfaces make recurring information easy to inspect; the agents handle less predictable investigation and production work.
Customer knowledge, company information, and product usage.
Internal and external information feed a live view of the market and customers. Interfaces support repeated inspection, while agents answer flexible questions and help produce account-specific work.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Map the addressable market with the ICP dashboard
The first interface is Exa’s internal ICP dashboard. It answers a foundational question: which customers and use cases make up the company’s world? Exa uses its search system to classify almost every company it identifies within its total addressable market. Wang names model providers, AI coding platforms, and go-to-market intelligence tools as example segments.
The dashboard supports both breadth and depth. At the top level, it groups potential customers into market segments. At the company level, it exposes metadata and an estimate of anticipated annual spend; Wang uses SpaceX as the example. He intentionally hides some category revenue information. The talk also supplies no calculation method, error rate, refresh schedule, or coverage audit, so anticipated spend should be understood as a planning estimate rather than a verified forecast.
The discovery mechanism begins with Exa crawling the web and training embeddings for web search. Wang’s shorthand is “embeddings over the internet”: web information is represented so the system can retrieve and group it by semantic meaning rather than exact keyword overlap alone. Exa uses that semantic filtering to generate a large candidate list for the dashboard. The talk explains this retrieval layer but does not specify the downstream classifier, validation process, or rules that decide whether a company belongs in the market.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Turn customer changes into signals, then investigate in Slack
A market map changes slowly; customer behavior can change by the minute. Request Lens handles that second timescale. It alerts the team when someone signs up, runs many searches, stops searching, or appears from an account Exa especially cares about. These events turn raw activity into prompts for attention. Wang says the team can act on them, but he does not describe the significance thresholds or claim that an alert automatically triggers outreach.
The go-to-market team then uses perhaps a dozen agents inside Slack to investigate accounts and produce customer-specific work. These agents can access substantial internal data. Account executives use them to build demos, and the wider team relies on them when it needs a deeper account analysis. The important operating choice is placement: the tools live in the collaboration environment the team already uses, rather than requiring every salesperson to operate a separate developer interface.
Wang describes agent spending by the go-to-market team as unusually high, even relative to engineering, but gives neither a dollar amount nor a measured return. The stronger evidence in the talk is behavioral: agents are routinely used for account questions and demo construction. Productivity is asserted rather than quantified.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Separate a personal agent’s voice, judgment, and access
Wang’s most personal example is Jeffbot, an attempt to build a digital version of himself during a week off in Mexico. The system has three distinct ingredients: a writing profile, a decision framework, and access to company systems. Keeping those ingredients separate matters because sounding like someone, deciding like them, and acting with their privileges are different engineering problems.
For voice, Wang analyzed 760 of his emails. The resulting profile included concrete habits: his emails averaged 18 words, and he tended to close with “Best” rather than “Sincerely.” This replaces a vague instruction to imitate his tone with observations derived from his actual writing.
For judgment, he analyzed hundreds of previous decisions and converted them into evaluations. Those examples became a calibration target: given comparable situations, the system should produce decisions resembling his past choices. The source material came from ordinary work records rather than a formal decision journal; in the Q&A, Wang identifies Slack and email as the artifacts containing those decisions.
The talk does not report evaluation scores, holdout methodology, grader design, or failure categories. Jeffbot therefore demonstrates a useful construction pattern—derive style from communications and behavioral tests from past decisions—but does not establish how faithfully it reproduces Wang’s judgment, especially in new situations unlike those records.
The final ingredient is tool and data access. Wang can invoke Jeffbot with broad read and write access to systems he can personally use. Other employees use it to create drafts of Slack messages, answers, decisions, and emails. That distinction becomes crucial when the audience later asks whether sharing the assistant also shares the cofounder’s privileges.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Build programmatic access without turning everything into chat
Wang extracts three design principles from these systems. First, agent-first requires API-first. Dashboards and chat agents can only use internal and external information if that information has a programmatic interface. He treats MCP, a CLI, and an API as possible implementations of the same requirement: agents need a callable route to the data and operations involved in the work.
Second, agent-first does not mean chatbot-only. A model can generate a fresh interface for a one-off request, but repeated tasks benefit from a stable interface whose layout and behavior people can learn. Exa’s design keeps both:
- Crystallized interfaces for recurring views such as the ICP dashboard and Request Lens.
- Flexible chat agents for account questions and tasks the interface did not anticipate.
The shared programmatic data layer supports both forms rather than forcing the team to choose one.
Third, Wang reframes buy versus build around customizability. Software built internally can be changed directly, but purchased software can also work if it exposes enough of its data and behavior for the company’s agents to adapt it. The criterion is whether the system can participate in the company’s workflows, not whether the company wrote every screen itself.
Salesforce is his example. Exa keeps the database and sales conventions Salesforce already provides instead of recreating those choices, then lets agents access it through MCP. Wang says the team uses that connection daily. This is a pragmatic middle path: buy an established system of record, then build agent-driven behavior around its programmatic interface. It does not mean every SaaS product is infinitely adaptable; the approach depends on what the purchased system actually exposes.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Give tool users training—and give agents caller-dependent permissions
The Q&A clarifies that Exa does not expect every salesperson to build these systems. Account executives run deals, SDRs support demand generation, and a separate forward deployed engineering organization builds and maintains much of the tooling. Non-FDE staff generally use rather than create the interfaces, with training sessions helping them become proficient with the available AI tools.
Forward deployed engineers connect both sides of the loop. They run or support deals, encounter friction in the sales process, and then add features or maintain the systems used to do that work. The people closest to customer-specific failures can improve the shared tooling rather than sending every problem through a distant product queue.
An audience member then identifies the dangerous implication of Jeffbot’s broad access: if everyone can call an assistant carrying Wang’s permissions, has the company effectively collapsed into one security level? Wang’s answer is that permissions depend on the caller. When he invokes Jeffbot, it can read and write across many systems. When another employee invokes it, the system can draft messages but does not receive all the MCP and tool permissions available in Wang’s own use.
This establishes two separate controls: tool availability and allowed actions. Restricting both is stronger than merely telling the model not to use a capability. The talk remains at the policy level, however; it does not explain identity verification, authorization enforcement, audit logging, or which data may enter an employee-visible draft.
The invocation can access many systems and perform reads and writes.
The shared assistant does not run with one universal permission set. Wang describes broader capabilities for his own calls and a drafting-only path with fewer tools for other employees.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Combine deal support and system building—until scale forces specialization
The final question asks how Exa’s FDE organization emerged. Wang contrasts the role with the older separation among solutions engineers, sales engineers, and account executives. His hypothesis is that AI lets a technical person supporting revenue also build the tools that make that support easier. Work that previously resembled two jobs—helping a deal and improving the sales system—can temporarily fit into one role.
That arrangement creates a short feedback loop: support a customer, observe repetitive work, build an improvement, and make the next deal easier for both the FDE and the account executive. It also explains why Exa assigns maintenance of its go-to-market AI systems to people participating directly in customer work.
Wang closes with an important limit. Exa has about eight or nine FDEs in a company of about 115 people, and he doubts the model will scale indefinitely with everyone doing everything. At the company’s current size, he considers the combined role a productive way to get far. Growth may eventually require more specialization. The organizational lesson is therefore a working experiment, not a universal end state.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Resources
From the talk
The recording, corrected transcript, chapter links, and official reading version of Wang’s presentation.
- Jeffrey Wang on XReference
Wang’s supplied public profile, identifying him as an Exa cofounder.
Further reading
- AI EngineerReference
Conference information and the broader catalog of talks on agents, MCP, evaluation, interfaces, and AI engineering.
Related talks
- Agents for Everything Else — swyx
A complementary account of using coding agents for design implementation, data synchronization, research, and other knowledge work outside conventional software development.
- Forward Deployed Engineering 101
Expands the organizational model Wang introduces at the end, including how FDE work differs from bespoke services and conventional sales engineering.
- AI in GTM at Notion — Flora Liu
Offers another go-to-market architecture built around shared customer context, behavioral signals, agent drafting, human approval, and outcome feedback.
Read the complete timestamped transcript
- 0:01
[music]
- 0:12
>> Hey everybody, I'm Jeff. I guess I was
- 0:13
introduced, but I'm the co-founder of
- 0:15
Exa, and today going to give a talk on
- 0:19
turning go-to-market into an AI
- 0:20
engineering problem in the spirit of
- 0:21
this
- 0:22
AI engineering fair. And just a quick
- 0:24
show of hands just to like understand
- 0:26
the audience, like raise your hand if
- 0:27
you're a technical.
- 0:30
Okay, great. Okay, so I kind of
- 0:32
oriented this talk around like
- 0:34
go-to-market
- 0:36
as presented to to engineers. So, happy
- 0:39
that I did that.
- 0:40
Cool. So, first just to like ground the
- 0:43
ground like what what Exa is cuz it's
- 0:44
sort of relevant inside of this
- 0:46
presentation. Exa is a Exa is a search
- 0:48
engine for agents.
- 0:49
Think like agents are really smart, but
- 0:52
they don't have access to the web. We're
- 0:53
like this web MCP web tool that agents
- 0:55
can access. We power Cursor, we power
- 0:56
Cognition, we power a lot of the AI
- 0:58
ecosystem at this point.
- 1:00
And
- 1:01
before we start, I also just want to
- 1:02
like talk about, you know, especially to
- 1:03
the technical audience, like why should
- 1:05
you even care?
- 1:06
Like why should you care about
- 1:07
go-to-market? I guess this audience
- 1:08
cares about go-to-market cuz you chose
- 1:09
to go to this
- 1:11
go-to-market talk, but I think there's
- 1:12
this like funny narrative right now,
- 1:14
which is like people are like, "Oh, like
- 1:15
product is the only thing that matters."
- 1:17
Or "Distribution is the only thing that
- 1:19
matters." And there's all sort of like
- 1:20
all sorts of like Twitter flame wars
- 1:22
like like oh, is Glean going to succeed
- 1:24
because they're really good at
- 1:25
distribution, but they're like what what
- 1:26
the heck is their product? And then and
- 1:28
other people are like, "Oh, the like the
- 1:29
the product needs to be super good cuz
- 1:31
agents
- 1:32
you know, agents shop for the product,
- 1:33
so they'll shop the for the best
- 1:34
product." And so, my view and my
- 1:36
experience in the last few years is that
- 1:39
you just kind of have to do both. Like I
- 1:41
think you have to get product right and
- 1:42
you have to get go-to-market right. Like
- 1:44
you got to build this thing, it's got to
- 1:45
be good, and then you got to get it into
- 1:47
people's hands. If you don't do both
- 1:49
things, then you don't have a company.
- 1:51
So, that's kind of my view on the
- 1:53
matter. And I I say like a really funny
- 1:56
thing also is like as a technical person
- 1:58
when you start a company or you start
- 2:00
some some sort of project, like very
- 2:02
much so the bias is like, "Hey, I'm
- 2:04
going to just build the thing. I'm going
- 2:06
to make it really really freaking good,
- 2:08
right?" Like that's kind of like the
- 2:09
bias you have as like an engineer.
- 2:10
That's the bias we had when we started
- 2:12
X.ai and we were like honestly pretty
- 2:14
bad at go-to-market. Like we
- 2:15
>> [laughter]
- 2:16
>> we were not doing enough marketing, we
- 2:17
were not doing enough sales.
- 2:18
Um but I think the cool thing about
- 2:21
about um go-to-market particularly in
- 2:23
2026 is you can treat go-to-market like
- 2:26
an engineering problem and particularly
- 2:28
an AI engineering problem. And so I
- 2:30
think that's like a super exciting
- 2:31
thing. Like it's like more fun for
- 2:33
engineers than ever to do go-to-market
- 2:36
cuz you can automate things. You can you
- 2:38
can do so much as one person and uh
- 2:41
etc. Also um I want to make this
- 2:44
interactive. If if anybody has questions
- 2:45
at any point, please please ask cuz I'm
- 2:47
aware there's a lot of talks and
- 2:49
I don't want to bore you.
- 2:51
Cool. So um cool. So so the hypothesis I
- 2:54
have is if you're an engineer or or if
- 2:56
you're anyone, you can treat
- 2:57
go-to-market like an engineering
- 2:58
problem. So first, I guess like what
- 2:59
does what do go-to-market teams do? So I
- 3:01
have like a laundry list of things here
- 3:04
of things that go-to-market teams do,
- 3:05
but here are a few. Like one is you got
- 3:06
to research
- 3:08
like your customer, right? You got to
- 3:09
research your targets. You have to find
- 3:11
out information about your about
- 3:13
targets. You have to find the right
- 3:15
people at particular companies. You have
- 3:17
to build POCs. Um there's like a just a
- 3:19
ton of stuff you have to do, right? Um
- 3:22
so you know, I'm not going to I'm not
- 3:24
going to list everything here, but like
- 3:25
what is the grand unifying theme? Well,
- 3:27
go-to-market is a data problem, right?
- 3:30
So you have all you have this like
- 3:31
entire world of uh of of what your
- 3:35
product does and then and this entire
- 3:37
world of like all your potential
- 3:38
customers and you're just going to like
- 3:40
learn and figure out what your world
- 3:44
looks like. And so this is my this is my
- 3:46
uh proposal. It's a data problem and we
- 3:48
have to solve from a data perspective.
- 3:51
Cool. So okay, so what is data that is
- 3:53
relevant? Uh I propose that you need
- 3:56
basically a live model of your world
- 3:58
that agents can act on. And so, what
- 4:01
does that mean? Okay, well, one is you
- 4:02
have a ton of internal data, right?
- 4:04
There's all this information that you
- 4:05
know about your customers, about people
- 4:08
that are at your company,
- 4:11
uh
- 4:12
data about how people use the product.
- 4:14
That's like internal data that you know.
- 4:16
And then there's all sorts of external
- 4:17
data, right? Like there's over 60
- 4:19
million companies in the world, and
- 4:21
there's like billions of people, like
- 4:23
over a billion that are on LinkedIn, for
- 4:24
example, and all sorts of stuff are is
- 4:27
is like happening every day, right? Like
- 4:28
there's all this news. And so, when
- 4:31
you're building like this data
- 4:33
go-to-market system, um it's important
- 4:35
to keep in mind just like all the
- 4:37
different sources that exist and and and
- 4:39
uh
- 4:40
and are available to your agents.
- 4:42
And cool. So, I'm going to like go
- 4:44
through, hopefully pretty fast, just all
- 4:47
the different components of what we've
- 4:49
built at Exa. And just for like context,
- 4:53
uh I've been really passionate about
- 4:54
this for a long time. So, like Exa was
- 4:56
launched in
- 4:58
uh the middle of 2023, and so we were
- 5:00
post-GPT-4. And
- 5:02
GPT-4 was really incredible, cuz it
- 5:04
could actually, even then, even though
- 5:06
it's way worse than like Fable or
- 5:08
whatever, like it could actually just
- 5:09
automate entire
- 5:11
parts of go-to-market. And so, from the
- 5:12
beginning, I've been thinking about our
- 5:13
go-to-market from
- 5:14
from a very, very uh AI agent-first
- 5:16
perspective. And so, we're going to go
- 5:18
over two interfaces that we have that
- 5:19
help us, and then two agents.
- 5:22
Cool.
- 5:26
Cool. Okay, the first is what we call
- 5:28
our ICP dashboard. And the ICP dashboard
- 5:32
is a product that we have internally
- 5:33
that answers the question, like what is
- 5:35
our world? Like what is the world of
- 5:38
customers and use cases that we care
- 5:40
about? And what we actually do is we go
- 5:44
ahead and use Exa, and again, Exa is
- 5:47
this like uh
- 5:49
arbitrarily powerful search engine for
- 5:50
AIs essentially. And we just classify
- 5:53
basically like every possible company
- 5:54
that is inside of our total addressable
- 5:56
market. And I kind of blurred out some
- 5:59
of the details on like how much money we
- 6:00
make from each category and stuff like
- 6:03
that. But yeah, we have like categories
- 6:04
like model providers, AI coding
- 6:06
platforms like say Cursor, go-to-market
- 6:07
intelligence tools.
- 6:09
And this makes up our TAM and we have an
- 6:11
understanding of literally like
- 6:13
almost every company within those
- 6:14
segments.
- 6:16
And then for each of those companies we
- 6:18
can deep dive, right? So here's the
- 6:19
example of SpaceX. We can see how much
- 6:22
annual spend we could anticipate them to
- 6:23
have then all this like metadata about
- 6:25
the company. So we have a list of all
- 6:26
the companies and then a ton of data
- 6:28
about each company.
- 6:29
How do we do this? Again, we're able to
- 6:32
do this because Exa is this search
- 6:33
engine. We take the internet, we crawl
- 6:35
it, we train we train embeddings to do
- 6:38
web search really well. And so basically
- 6:40
from a technical perspective you can
- 6:41
think about Exa as like embeddings over
- 6:43
the internet. And when you have
- 6:45
embeddings over the internet you have
- 6:46
this like arbitrarily powerful semantic
- 6:49
filtering and slicing and dicing of any
- 6:51
type of data that you want. And so we
- 6:53
use that to generate this this like
- 6:54
gigantic list of potential ICPs.
- 6:59
Cool. Next, we have a tool we call
- 7:02
Request Lens. Request Lens, what is
- 7:04
Request Lens? Well, it's basically a
- 7:05
system where anytime something
- 7:07
significant happens with any of our
- 7:08
customers, we're alerted.
- 7:10
Someone signed up, someone used a ton of
- 7:13
searches, someone stopped using
- 7:14
searches, someone showed up that we
- 7:16
really really care about. All these
- 7:18
things are signals that we are notified
- 7:20
about and that our team can act on.
- 7:28
Cool.
- 7:29
So those are the two interfaces that we
- 7:31
have and then I'll go over two types of
- 7:33
agents that we have. So
- 7:34
one is coding agents. So our
- 7:37
go-to-market team
- 7:38
is crazy crazy crazy deep on agents. So
- 7:42
like like our our our engineering team
- 7:44
uses a lot of agents, but our
- 7:45
go-to-market team is like like you could
- 7:46
look you could look at some of their
- 7:47
like devin spend and like other agent
- 7:49
spend. It's really freaking high. And
- 7:50
that's because everybody on our
- 7:52
go-to-market team is constantly asking
- 7:54
agents about our customers.
- 7:57
Uh we have like
- 7:58
like account executives that build demos
- 8:01
for our customers. Like it's just this
- 8:03
crazy ecosystem where we have like maybe
- 8:05
a dozen different agents inside of our
- 8:07
Slack and anybody can use any of them.
- 8:09
They all have access to tons and tons of
- 8:10
our internal data. And uh yeah. Anytime
- 8:13
we want to dig deeper on account,
- 8:15
anytime we want to make a demo, etc., we
- 8:17
depend heavily on agents.
- 8:23
Cool. And then I want to talk about
- 8:24
another really cool agent that I'm
- 8:25
pretty proud of. We call it Jeff Bots.
- 8:27
Uh or I call it Jeff Bots. Uh Jeff Bots
- 8:29
is an AI clone of myself. Uh as much as
- 8:32
possible. So, what is it? Well,
- 8:34
basically
- 8:35
this winter break uh
- 8:38
I'm sure a lot of you spent that break
- 8:39
playing with Opus 4.5. And I was no
- 8:42
different. So, I was in Puerto No, I was
- 8:44
in Mexico. I was in Mexico and I would I
- 8:46
had a week off. And so, my goal with
- 8:49
that week and with Opus 4.5 was to uh
- 8:52
just try to make a digital clone of
- 8:53
myself. And so, I did things like
- 8:54
analyze
- 8:55
like 760 of my emails to figure out what
- 8:59
my email voice is. Like, oh, I use 18
- 9:01
words on average per email and I like to
- 9:03
end emails with best and not sincerely.
- 9:05
Like all all that type of stuff, right?
- 9:06
So, I made like a like a voice for
- 9:09
myself.
- 9:10
And then I also made a decision-making
- 9:13
framework. So, I made like a
- 9:14
decision-making framework where I
- 9:16
analyzed hundreds of decisions I've made
- 9:17
in the past.
- 9:19
And I I analyzed them and I created
- 9:21
evals. So, I actually created evals from
- 9:23
those decisions and calibrated this
- 9:25
agent system to behave like myself.
- 9:29
And then finally I gave it like read and
- 9:30
write access to all the data that I
- 9:32
personally have. And there's a cool
- 9:34
advantage to this because like I
- 9:35
basically have access to every single
- 9:37
system at the company
- 9:39
uh cuz I'm in the the the nice seat
- 9:41
of of having that and so like um yeah,
- 9:45
this thing has access to like
- 9:46
everything. And basically what happens
- 9:48
is anybody at the company can use
- 9:49
Jeffbot to create drafts of Slack
- 9:52
messages that are basically like answers
- 9:54
or decisions that are made. And this is
- 9:56
a huge great thing. Like our
- 9:57
go-to-market team uses it to like draft
- 9:59
emails, for example.
- 10:03
Cool. Um
- 10:05
all right, so those those are the
- 10:06
systems that uh
- 10:08
that we have at at Exa. It works pretty
- 10:10
well. Our go-to-market team is very
- 10:11
lean, but very productive.
- 10:14
Um and so yeah, I just want to cover
- 10:15
like
- 10:16
lastly just a few principles
- 10:19
um
- 10:20
principles I have
- 10:22
around what it means to be an
- 10:23
agent-first company.
- 10:25
So firstly, to be agent-first you must
- 10:27
be API-first, right? So like all these
- 10:29
systems that we built, whether they were
- 10:30
those whether it was those agents or
- 10:32
whether it was those GUIs that we have,
- 10:35
like if there did not exist really good
- 10:37
APIs on top of any internal and external
- 10:39
data
- 10:40
we'd be we'd be out of luck, right? Like
- 10:42
you need to create really good APIs. If
- 10:44
you don't have really good APIs your
- 10:46
agents are not going to be able to have
- 10:48
data access. So you can think about this
- 10:50
as MCP, CLI, whatever, right? Like it
- 10:52
doesn't really matter. Uh you just need
- 10:53
some interface that's programmatic.
- 10:57
Secondly is like I I think there's like
- 10:58
still this mistake in
- 11:00
the agent world which is made that's
- 11:02
like hey, does everything
- 11:04
need to be a chatbot?
- 11:06
Uh
- 11:07
I think the answer is no. Like I think
- 11:09
I think both GUIs and chatbots are are
- 11:12
both super useful and have their own
- 11:16
benefits. Like uh I don't know how many
- 11:19
people in this room have thought about
- 11:19
dynamic user interfaces
- 11:21
but like yes, dynamic user interfaces
- 11:24
are amazing. Like yes, technically AI
- 11:26
can just produce a new UI for any use
- 11:28
case that you have. Like
- 11:30
just to answer a question, it could
- 11:31
produce like an HTML markdown file,
- 11:32
right? But I think there is something
- 11:34
really nice about being able to visit
- 11:36
the same consistent UX for the same use
- 11:39
cases over time so that you can like
- 11:41
learn how to use some tool. Um so yeah,
- 11:43
I think like having crystallized UIs and
- 11:45
then also arbitrarily powerful flexible
- 11:47
chat agents are both important
- 11:49
components of being agent first.
- 11:53
And then finally,
- 11:55
uh
- 11:56
you know, there's this question like,
- 11:57
"Hey, should you like shop for like
- 11:59
Salesforce
- 12:00
or should you like build your own CRM or
- 12:03
something, right?" I actually think this
- 12:05
is like a false dichotomy. It's like
- 12:08
like there it's not a choice. Like we
- 12:10
don't live in a world where the choice
- 12:11
is between purchasing SaaS and building
- 12:13
things yourself. Like the way I like to
- 12:15
think about it is like
- 12:18
you should just be using something that
- 12:20
is arbitrarily customizable, right? Like
- 12:22
whether you like obviously if you build
- 12:24
something yourself, then it's
- 12:25
arbitrarily customizable cuz you can
- 12:26
write code and make it better at any
- 12:28
given point.
- 12:29
But also if you procure SaaS, um if you
- 12:33
can make that SaaS work on your behalf
- 12:35
and
- 12:36
be arbitrarily customizable, then that
- 12:38
works too, right? Like you don't need to
- 12:39
build this like
- 12:41
GUI and like have a proactive roadmap as
- 12:44
to like what features would make really
- 12:45
great sense inside of some system. Like
- 12:47
if you can arbitrarily customize the
- 12:49
system, even if it's a system you've
- 12:50
purchased, then you're like pretty good,
- 12:52
right? So like for example, we use
- 12:53
Salesforce. Like we use Salesforce at
- 12:55
X.ai and uh it's great because uh it's a
- 12:58
really good good database. It's made a
- 13:00
lot of amazing choices around what sales
- 13:03
should look like, choices that we don't
- 13:04
want to make ourselves. And then it
- 13:06
exposes MCP. So all of our agents have
- 13:08
access to Salesforce MCP. Works really
- 13:10
well. Our team uses it every day.
- 13:11
And so yeah, I think infinite
- 13:13
customizability um is is really the
- 13:15
highest order bit.
- 13:19
Cool. Um
- 13:21
that's that's all I had.
- 13:23
Uh
- 13:24
Yeah, does anyone have any questions?
- 13:26
>> Oh, we have time for a few questions.
- 13:28
Okay, coming.
- 13:36
>> Hey, um so you said you uh took all your
- 13:39
past decisions. Can you elaborate a bit
- 13:41
about that?
- 13:43
What What artifacts are those?
- 13:45
Usually people don't save like their
- 13:47
decisions. Is it Slack? Is it email? Is
- 13:49
it other other artifacts?
- 13:52
>> Yeah, good question. I looked at
- 13:53
decisions I made within Slack and email.
- 13:56
I mean, a surprisingly large amount of
- 13:59
everything that goes on a company is
- 14:01
is is on Slack, right? So like if you
- 14:03
just read like a ton of Slack history,
- 14:04
like you can definitely find hundreds of
- 14:07
decisions that you made in the past.
- 14:09
>> Awesome. Quick question uh over here. Uh
- 14:12
so your go-to-market team, what's the
- 14:13
split between uh are they just all like
- 14:17
AI cracked or do they also have like the
- 14:19
domain expertise, too? What's the split
- 14:21
between
- 14:22
technical and non-technical? Because
- 14:24
obviously you need them to like know how
- 14:26
to do marketing, sales, etc. But then do
- 14:28
they also are they also upskilling in
- 14:30
terms of using AI systems? Are you
- 14:31
handing them tools or they building
- 14:33
their own?
- 14:35
>> That's a very good question. So our
- 14:37
go-to-market team
- 14:39
is comprised of like there's there's
- 14:42
account executives which like run the
- 14:44
deals.
- 14:45
There are like sales like SDRs that help
- 14:49
with uh demand generation.
- 14:51
And then there are separately they're
- 14:54
separate like a FDE org. So forward
- 14:56
deployed engineering organization.
- 14:58
And
- 14:59
what I'll say is that like
- 15:02
everyone that's Okay, everyone that's uh
- 15:04
not not in FDE
- 15:06
is like
- 15:08
has learned how to use AI really well.
- 15:11
So like
- 15:12
the answer is like they're not vibe
- 15:13
coding. They're not generally with you
- 15:16
know, in some there's some exceptions.
- 15:17
They're not generally vibe coding these
- 15:18
interfaces that we have. Um but they're
- 15:20
using the tools really really well. And
- 15:22
like we make sure that we have training
- 15:24
sessions and like just make sure that
- 15:25
people
- 15:26
really understand how to use these
- 15:28
tools. And then this funny we have this
- 15:30
funny thing which is like our four
- 15:31
deployed engineering organization is
- 15:33
actually the one that like does a lot of
- 15:35
the maintenance and feature building
- 15:37
uh of these AI systems. And so they're
- 15:39
both running deals and like like
- 15:42
supporting deals
- 15:43
um but then also making
- 15:46
everything smoother by like
- 15:48
doing sales but then also building the
- 15:51
sales system. Like it's it's kind of
- 15:54
it's kind of a funky thing we have going
- 15:55
on. Yeah.
- 15:58
>> How do you think about uh
- 16:00
different security?
- 16:02
>> Oh.
- 16:02
>> Hey. How do you think about different uh
- 16:04
security boundaries within your
- 16:05
enterprise? What do you What you said
- 16:07
suggested that you've got Jeffbot which
- 16:09
had runs with all of your full
- 16:11
privileges and then it's available to
- 16:13
everybody which suggests that there's
- 16:14
one security level and everyone can see
- 16:16
everything all the time. Is that what
- 16:18
you're going with or
- 16:19
is there some uh other guardrails in
- 16:21
place?
- 16:22
>> Yeah, that's a good question. We we pay
- 16:24
pretty special we we pay pretty careful
- 16:26
attention to guardrails. So for example
- 16:29
um in the case of Jeffbot
- 16:31
um
- 16:32
when I use Jeffbot and I call Jeffbot
- 16:34
has access to
- 16:36
a ton of systems and it can for example
- 16:38
do reads and writes. However, when
- 16:40
anybody else calls Jeffbot all can do is
- 16:42
draft messages and also I don't give
- 16:45
Jeffbot
- 16:46
permissions to
- 16:48
all of our MCPs and tools in the case
- 16:50
where other people call it. And so in
- 16:52
short it's like pretty
- 16:54
it's pretty well defined or we we we do
- 16:56
pay some care to the security.
- 16:58
Yeah.
- 16:59
>> Okay, last question.
- 17:04
>> Um, can you can you share the origin
- 17:07
story of the FDE team? Did that just
- 17:10
happen organically or did you
- 17:11
intentionally do it? I'm just really
- 17:13
curious like how that came to exist.
- 17:16
>> Yeah, for sure. I mean,
- 17:18
uh
- 17:19
my
- 17:20
my philos- my my hypothesis on this is
- 17:22
like, once upon a time the FDE role
- 17:24
didn't really exist. Like, Palantir
- 17:26
started calling some people FDEs, but it
- 17:28
that was really it. And what tech
- 17:30
companies had was like solutions and
- 17:32
sales engineers.
- 17:34
And then, like account executives.
- 17:37
>> I was a solutions engineer.
- 17:38
>> Got it. Yeah, yeah. The thing The thing
- 17:40
that I think has changed is that um
- 17:42
because of AI, as like
- 17:46
en- as a technical person that is
- 17:48
supporting revenue generation,
- 17:51
you can actually not only support the
- 17:52
revenue generation, but then very easily
- 17:54
build the tooling
- 17:55
to smooth everything over.
- 17:58
And make your own life easier, make the
- 18:00
lives of AEs easier. Like, because of
- 18:01
AI, this is just possible now. Like,
- 18:03
that's like two
- 18:04
Before that was like two jobs, and now
- 18:05
it's like one job.
- 18:07
Um in theory. Like, now when our team
- 18:09
grows, like right now it's about eight
- 18:11
or nine FDEs, like what will will it
- 18:13
scale such that everyone does
- 18:14
everything? Probably not. But, at least
- 18:16
right now that's what we have, and I
- 18:17
think that's a really good working model
- 18:18
to get pretty far.
- 18:20
>> Eight out of how many?
- 18:22
>> Uh eight Oh, eight of like how big is
- 18:23
our go-to-market org?
- 18:24
>> Or eight You have eight FDEs, and the
- 18:26
size of the company right now is how
- 18:28
many?
- 18:28
>> Oh, the We're about 115 people.
- 18:31
>> Okay.
- 18:31
>> Yeah.