No Priors

Jake Stauch on Why the Guardrails Are the Product

Jake Stauch· Co-founder and CEO of Serval at Serval
·~38 min·English·No Priors
AgentsAI SafetyAI InfrastructureBusiness Strategy
TL;DR

Serval's CEO argues that in the enterprise the guardrails (deterministic approvals, least-privilege permissions, audit trails) are the real product, and that they grow more valuable as the agents they wrap get smarter.

01Core thesis

The Guardrails Are the Product

Serval sells the boundary, not the brain: <strong>the guardrails are the product</strong>, and they get more valuable as the models they wrap get smarter.

Serval is the boundary. Serval is the guardrails. 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.

— Jake Stauch, No Priors
Key Insight
The real moat is not model quality, which Serval neither owns nor controls, but the governance layer wrapped around it; and that layer appreciates precisely as the models it wraps get more capable.
02Go-to-market

Attack the Giant, Don't Layer on It

Instead of layering on ServiceNow or going down-market, Serval <strong>attacked the giant head-on from day one</strong>, taking a harder path for a far bigger prize.

I like to say everything has gone exactly according to our very delusional plans. So we set out from the beginning to be the largest enterprise software company of all time.

— Jake Stauch, No Priors
Key Insight
Choosing the hardest target forced every early decision (hiring, product, growth) to be enterprise-grade from day one, which is itself the defensibility. Stauch frames the alternative, serving one market and hoping to expand later, as a fundamentally different starting position rather than a detour you can reverse.
03Design bet

Code Is the Right Abstraction

Drag-and-drop builders hit a ceiling for everyone; <strong>code is the right abstraction for automation</strong>, and AI finally made it writable by non-engineers.

Code is a better way of building automation.

— Jake Stauch, No Priors
Key Insight
Code-gen agents did not invent a new abstraction; they removed the one barrier, literacy, that kept the oldest abstraction (code) out of reach for most of the workforce. The winning move was not a better builder, it was deleting the need for a builder at all.
04The product

Coding Agents for the Rest of Us

Serval wraps code-gen agents in the engineer's own safety net of dev, staging, approval, and audit trails, so <strong>any department can build without breaking production</strong>.

under the hood it's coding agents, but then you have all these abstractions and guardrails and controls, so things go through approval, they go through a staging and approval cycle and audit trail and all of that.

— Jake Stauch, No Priors
Key Insight
Serval's wedge is not a new agent capability; it is importing a decades-old engineering discipline, the SDLC, into departments that never had one. The novelty is organizational, not technical, which is exactly why incumbents shipping raw agents keep tripping over it.
05Counterintuitive insight

The Least-Privilege Paradox

The paradox: as agents get more capable, enterprises will <strong>shrink everyone to least privilege</strong> and route the exceptions through a control layer Serval wants to own.

what I'm actually going to want to do if I'm a large enterprise is shrink everyone's permissions down to truly least privilege. Like you cannot access any document and you cannot take any action unless you absolutely need to do it for your job.

— Jake Stauch, No Priors
Key Insight
This inverts the usual worry: the danger of a capable agent is less the model than the standing access a human already holds and the agent inherits. Shrinking people to least privilege makes the escalation layer, not the model, the control point, and Serval intends to own it.
06The frontier

The Autonomy-Safety Tension

Background agents are 'the ultimate moonshot,' but <strong>full autonomy is an unbounded security risk nobody will take yet</strong>, so Serval climbs by approving one capability at a time.

They could theoretically just start writing code that could take action on your systems and that is an unbounded security risk that I don't think anyone's going to be able to take anytime soon.

— Jake Stauch, No Priors
Key Insight
Incremental escalation is a trust-building protocol, not a safety guarantee: it caps blast radius one approved tool at a time but never closes the gap to true hands-off autonomy, which Stauch concedes is still unbuilt. The honest position is that safety here is a ceiling you approach, not a box you check.
07Economics

Measure the Work, Not the Tokens

The token-maxing era is pulling back and customers now demand ROI proof, so Serval <strong>prices on workflows run, not tokens</strong>, because tokens measure verbosity, not work.

They realize that there's going to be a short-term pullback, but they're still planning on increasing spend. What they need though is more proof of ROI.

— Jake Stauch, No Priors
Key Insight
Pricing on workflows run turns Serval's own invoice into the ROI proof Stauch says customers are demanding: a workflow is a unit of completed work, where a raw token count is not. It ties what Serval charges to what the customer got done, not to how verbosely the model happened to get there.