The world is too loud. Read what matters.

No Priors

The Smarter AI Gets, the Tighter Enterprises Lock Permissions

Jake Stauch argues that as models grow more powerful, the damage from permission abuse scales up—so enterprises shrink default permissions to absolute minimum and rely on approval workflows to grant exceptions. That's what Serval sells.

Enterprise softwareCoding agentsServiceNowPermission controlBackground agentsSaaS
A counterintuitive yet logical design pattern for builders of enterprise-grade agents or permission systems, plus insight into how founders choose to fight giants head-on rather than hunt for niches.

The argument · tap a timestamp to hear it

2:00

ServiceNow's root problem: automation platforms cannot keep pace with business

ServiceNow is fundamentally a workflow automation platform—powerful enough, but every new requirement means reimplementation, often requiring consultants. This means enterprise solutions always ship six to twelve months behind what the business actually needs. Acceptable for companies whose requirements hold steady for five to ten years, but for enterprises chasing continuous automation, ServiceNow lags while the bill climbs.

— Jake Stauch
5:06

Personal agents are nearly useless inside large enterprises

Personal agent products work well for consumers and mid-market, but large enterprises are a different game: employee permissions are already so restricted—many can't even read their own email—that agents inherit those same narrow constraints. Serval's approach is different: use service accounts and admin-level permissions so employees can temporarily borrow higher privileges, but each use requires approval and a full audit trail. Fundamentally, it's still a work order system underneath.

— Jake Stauch
8:12

Day one: commit to defeating ServiceNow

Most enterprise software startups either build a thin layer atop giants or go downmarket—both paths have never truly succeeded. Serval went a different direction. Two co-founders in a living room decided on day one to list every large company, circle ServiceNow, and declare they would beat it head-on and become a bigger company, rather than finding a niche and growing gradually.

— Jake Stauch
11:16

The bet: deterministic scripts will be AI's first conquest

Two and a half years ago, coding ability was still autocomplete-only, nowhere near complex software generation. Serval's bet wasn't ‘AI can code anything’ but rather that password resets, permission approvals, and similar workflows are provably deterministic—verifiable scripts where AI would improve fastest. The team started hand-feeding case libraries to the model, but it got good so quickly that they can barely find workflows it cannot automate.

— Jake Stauch
13:17

The smarter AI gets, the tighter enterprises lock permissions

Jake expects stronger models will not unlock enterprise permissions—the opposite. Once everyone can invoke powerful agents, even slightly oversized permissions can scale the damage, a risk already visible in cyber security. Result: large enterprises will shrink default permissions to absolute minimum while needing many temporary exceptions. Whoever controls that approval process, leaves the audit trail, and builds the workflow is what Serval actually sells.

— Jake Stauch
23:38

Internal vibe coding cannot replace system-of-record software

Against ‘SaaS is dying’ arguments, Jake's view is split: companies will absorb hundreds of scattered point SaaS tools into core platforms, and middleware will vanish. But ERP, HR, ITSM, CRM—the categories of system-of-record software that document enterprise state—won't be replaced by internal vibe coding. He's found that enterprises discover homegrown enterprise software is expensive to build and maintain; unless requirements are exotic, buying the product still pencils out. His forecast: these categories will see new leaders, not disappear entirely.

— Jake Stauch
28:46

Background agents are the ultimate moonshot—no one has solved it

Jake calls self-discovering, self-solving backend agents the enterprise software equivalent of the moonshot—tremendously powerful if achieved, but autonomy and enterprise security requirements are natural opposites. His middle ground: let agents only gradually escalate capability. Each new tool they want to call requires an approval request first. After approval, they can use it. Over time the toolbox grows, but true hands-off operation never arrives.

— Jake Stauch
36:53

No product managers—engineers live inside customer Slack channels

Serval has no separate product function. Engineers spend time daily in customer Slack, and most have at least one customer call per day. They stay immersed enough to catch real-time shifts in customer preference, so product decisions get made at the front line rather than bubbling up through layers. It also means the team must be willing to frequently overturn whatever they were confident about yesterday. Jake sees that capacity for self-reinvention as the core skill for keeping pace with how fast the field moves.

— Jake Stauch

In their own words · checked verbatim

So it's always kind of running six to 12 months behind where you want it to be.

Jake Stauch2:00

even if you are 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.

Jake Stauch5:06

So we set out from the beginning to be the largest enterprise software company of all time.

Jake Stauch8:12

We circled ServiceNow and we said we're going to beat them and we're going to build something much bigger.

Jake Stauch8:12

the bet was that coding agents would be would get very good

Jake Stauch11:16

So one of our largest customers is Fox. Fox Corporation adopted Serval company-wide. And the person who led that migration on the ground level actually got nominated to be the MVP of the entire company and got an award for his work there.

Jake Stauch15:17

it's quite expensive to build and maintain enterprise software.

Jake Stauch23:38

I think what's so unique about Serval and the category we're in and the moment we're in is the TAM is unbounded.

Jake Stauch33:50

Figures

Company founded2.5 years ago8:12
Current employees20032:50
Employees (January 2026)2032:50

Glossary

TAM
Total Addressable Market: the maximum market size a product could theoretically capture.
SDLC
Software Development Lifecycle: the multi-stage process (dev, staging, prod) code must pass before going live.
vibe coding
Describing intent to AI without detailed specification, then using the generated application as-is.
ITSM
IT Service Management: the software category for enterprise IT ticketing, asset management, and process automation; ServiceNow is the category leader.

How to listen

Who it's for

Founders and product managers building enterprise software, permission architectures, or agent products who want to understand how large enterprises actually land coding agents.

Skip

From 32:23, the discussion of hiring pace and startup ambition is subjective and low-density—you can skip it.