ExpertsFMTaking on ServiceNow: Coding Agents the Rest of Us with Serval Co-Founder and CEO Jake Stauch
No Priors: Artificial Intelligence | Technology | Startups · October 29, 2026
5 Selected Lessons:
framework · 3 heuristic · 1 case · 1
Guardrails-as-the-product for enterprise agent autonomy
Source
No Priors: Artificial Intelligence | Technology | Startups — Taking on ServiceNow: Coding Agents the Rest of Us with Serval Co-Founder and CEO Jake Stauch (October 29, 2026) · at 0:00 ▸ · at 12:21 ▸
When this applies
when introducing any powerful autonomous tool into an organization with approval chains, compliance, and political risk
The takeaway
When the core technology is a commodity getting better every year, build your moat on the trust/control layer that lets buyers safely adopt it, not on the raw capability itself
Why it works
Autonomy without bounded permissions creates unbounded downside risk, so the commercial value shifts from raw intelligence to the control layer that makes autonomy safe to grant
One level up
In categories racing toward capability abundance, differentiate on governance and trust, since capability itself will be commoditized.
Where it stops holding
This stops working when the buyer doesn't face real compliance/political/security risk (e.g., small, informal teams) and just wants raw capability with no oversight overhead.
In their words
We've always believed fundamentally that we are not building the LLMs ourselves. We are building the guardrails. We are building the boundaries. And the boundaries are the product for us because we are the way that a large enterprise with a lot of rules and approvals and process and politics can adopt these tools safely and comfortably at scale.
Import software engineering's SDLC gates into AI-driven enterprise automation
Source
No Priors: Artificial Intelligence | Technology | Startups — Taking on ServiceNow: Coding Agents the Rest of Us with Serval Co-Founder and CEO Jake Stauch (October 29, 2026) · at 17:25 ▸
When this applies
when deploying any AI system that takes real-world actions inside an organization with compliance and risk requirements
The takeaway
To get risk-averse institutions to adopt a new powerful technology, embed it inside the institution's existing trusted risk-management ritual rather than asking them to adopt a new one
Why it works
Enterprises already trust a proven pattern (dev/staging/prod gates, code review, UAT) for managing risk of change; mapping new AI capability onto that existing trusted process lowers the adoption barrier more than any amount of raw capability would
One level up
Diffusion of powerful new technology into institutions is gated by process fit, not capability — wrap the new in the familiar.
Where it stops holding
This stops working when deploying into organizations or teams with no such existing change-management discipline, where SDLC gates would just slow down otherwise welcome experimentation.
In their words
This idea of, hey, you have to actually follow an SDLC process. If you're a large enterprise, a large financial institution is not just going to ship new agent capabilities without a review and user acceptance testing and all of that.
Treat non-engineer power users as adults: expose full capability with optional abstraction
Source
No Priors: Artificial Intelligence | Technology | Startups — Taking on ServiceNow: Coding Agents the Rest of Us with Serval Co-Founder and CEO Jake Stauch (October 29, 2026) · at 25:27 ▸
When this applies
when building a tool for a skilled-but-non-credentialed user group (e.g., ops staff, analysts) who are often given dumbed-down tools by competitors
The takeaway
Don't infantilize a user segment based on their job title; give them full access to the real mechanism with an optional simplified layer on top, and let each user choose their depth
Why it works
Competitors assume non-engineers need simplified, opinionated, stripped-down tools, but power users in adjacent roles actually understand and want the full complexity; giving them the underlying code with an optional abstraction layer serves both the technical and non-technical members of that group without alienating either
One level up
Respecting the latent sophistication of an underserved user segment is itself a product differentiator against incumbents who stereotype that segment.
Where it stops holding
This stops working when the target user segment genuinely lacks the conceptual background to engage with the underlying mechanism even as an option, making full exposure purely confusing.
In their words
I think a lot of the products that have been out there have actually kind of been pedantic to IT of like, oh, you don't actually understand this stuff. So you need this very stripped down, more opinionated, simpler version. Actually, IT appreciates the complexity and appreciates the power of these platforms.
Personal agents fail in enterprise due to narrow personal permissions
Source
No Priors: Artificial Intelligence | Technology | Startups — Taking on ServiceNow: Coding Agents the Rest of Us with Serval Co-Founder and CEO Jake Stauch (October 29, 2026) · at 3:00 ▸
When this applies
when porting a consumer product pattern into an enterprise context that has strict individual permission boundaries
The takeaway
A tool's usefulness in a constrained environment is bounded by the access rights of its operator, not by the tool's intelligence
Why it works
An agent can only act within the permission set of the person deploying it; in large enterprises individual permissions are deliberately narrow, so agent usefulness is capped by that narrowness regardless of model capability
One level up
Before assuming a capability will transfer across contexts, check whether the structural constraints of the new context (permissions, regulation, politics) will bottleneck it.
Where it stops holding
This stops working when the organization is small/informal/SMB where individual permissions are already broad, or when agents are given elevated service-account access instead of personal permissions.
In their words
The permission sets of any individual are so limited that even if you were allowed to bring your agent to work, which a lot of cases you're not, what it can do is so incredibly limited, it can barely even read your emails.
Going after the giant incumbent instead of the niche layer or down-market play
Source
No Priors: Artificial Intelligence | Technology | Startups — Taking on ServiceNow: Coding Agents the Rest of Us with Serval Co-Founder and CEO Jake Stauch (October 29, 2026) · at 6:29 ▸ · at 9:53 ▸
When this applies
when a new entrant chooses a strategic posture against a dominant incumbent — attack directly, build a thin layer on top, or go down-market
The takeaway
Choosing the hardest, biggest target from the outset changes every subsequent decision in a company's favor toward building something categorically larger, even though the near-term difficulty is higher
Why it works
Building for the largest, hardest market from day one forces fundamentally different product, team, and growth decisions than starting small and expanding later, and most 'layer on top' or 'down-market' strategies cap the ceiling or get absorbed/undercut
One level up
The size of the prize you aim for from day one shapes the architecture of the company you become, not just its eventual scale.
Where it stops holding
This stops working when the startup lacks the capital, patience, or technological unlock needed to survive the much longer and harder path to displacing an entrenched giant.
In their words
We set out from the beginning to be the largest enterprise software company of all time. We actually made a list of all the companies. We circled ServiceNow and we said we're going to beat them and we're going to build something much bigger.