Source: Go Hard Early — Jake Stauch on Serval
Source: Go Hard Early — Jake Stauch (Serval), First Round Capital, 1:23:02. YouTube: fSl5zdnM_HE. Transcript: auto-generated captions — verify names/figures against a second source before external use.
What It Covers
A long-form founder interview (host appears to be Brett Berson^[inferred]) covering how Jake Stauch is building Serval, an AI-native platform for IT/ITSM. It is the strongest source for Stauch’s strategy reasoning — what he carried from Verkada (his prior company), why he chose to build a full platform on day one, the code-based automation architecture, the forward-deployed-engineering org model, and how he frames the competitive window against incumbents. Light on cap-table specifics (covered better by the Peel interview); heavy on product, GTM, and hiring philosophy. Host intro pegs the company’s Series A at $47M led by Redpoint, with customers cited as Perplexity, Verkada, Mercor^[ambiguous], and Together.
Caption note: proper nouns are auto-transcribed. Corrected silently throughout — e.g. “Vicata/Ricotta/Forata” = Verkada, “Serbal/servil/servo” = Serval, “Moroi” = Meraki, “Octa” = Okta, “Alassian” = Atlassian, “cloud code” = Claude Code.
Key Claims (with timestamps)
Existing-category playbook from Verkada (2:11)
“One of the big ones is going after an existing market and an existing category of spend… there’s something really powerful about going in and just building a better product in a category where people buy that product and want that product.”
Verkada “owned the entire platform” — you sell a better camera, then ride that to sell the AI/video-software capabilities. Stauch carried the question to Serval: what platform with existing spend can we sell, while packaging something new inside it? The product “has to be 10 times better… undeniable” because customers “build their businesses around these systems” (3:50).
”Go hard early” — the title thesis (7:41)
“Verkada gave me this confidence of just just do the hard things. Actually seek out the hard things that are harder for other people to do that unlock a lot of customer value. And then once you’ve built them, it’s so much harder for somebody to come in after you… because you decided to like go hard early.”
Verkada’s hard choice was building its own cameras rather than software on top of others’. Serval’s analog: building a full ticketing system, on-prem deployment, roles/permissions, and enterprise integrations rather than a thin point solution.
Day-one enterprise readiness (9:51)
“Verkada waited several years to support things like SCIM and SSO and SOC 2… We did all of those things from day one because we knew that that was going to be a blocker to sell into the large accounts.”
Unlike Verkada (whose small-headcount/large-footprint customers were atypical), Serval’s spend scales with headcount, so it had to be enterprise-ready early — user management, permissions, flexible/on-prem deployment all built up front.
Serval’s origin — “reboot as a service” that died (10:24)
The first kernel was auto-resolving device issues at Verkada customers (“reboot as a service” for offline cameras/switches). Customer discovery killed it — too Verkada-specific. But the same customers volunteered the real pain: “I’ve got access requests. I’ve got password resets. I’ve got people just asking me the same question they could look up in the knowledge base.” That pivot became Serval (12:04–12:37).
The “if you could hire someone” reframe (15:18)
“One of the questions I started asking… if you could hire somebody today to just sit next to you and do work for you, what would you have them do?”
This non-judgmental framing surfaced what direct pain-point questions could not. Crucially: “they never identified ticketing as a pain point. Nobody said they wanted a new ITSM… They just said they wanted somebody to go and handle a bunch of these requests for them and build cool automations” (15:52).
Core automation insight (19:46)
“Automation doesn’t actually work for this IT use case unless it’s faster to automate something forever than to do it manually once. You have to take away the trade-off.”
Why existing tools (Okta Workflows, Zapier) went unused despite being a crowded category: building a single workflow (e.g. a password reset with enterprise rules) takes days/weeks for unclear ROI. Serval’s bet is to flip the trade-off so automating is always cheaper than doing it once.
The workflow-builder demo — Serval’s “SMS moment” (26:24)
Asked what Serval’s equivalent of the famous Verkada “text-the-camera-link” demo is, Stauch answers: the workflow builder. You describe a complex automation in natural language (offboarding, onboarding, “take a web hook from Rippling, Slack the manager”), hit enter, and “the workflow just generates before your eyes… it actually gets written out in code… It’s not some kind of LLM black box.”
“It’s basically vibe coding for IT automation, or vibe coding on rails.” (21:25)
Configurability, not simplicity — positioning vs incumbents (28:04)
“We’ve done that… not say we’ve got a simpler, more opinionated version of Service Now, but actually a more powerful version of Service Now because now you can write underlying code that does anything. You’re not even limited by Service Now’s domain-specific language.”
Critique of prior startups attacking ERP/CRM/ITSM with “a simpler, more opinionated” product: enterprises want to adapt the software to their process, not vice versa. Workday/Salesforce/Service Now win on that adaptability; Serval aims to beat them on it via code-based configuration.
Mid-market + enterprise straddle via code-based architecture (29:46)
Conventional wisdom is a narrow ICP; Serval instead straddles ~200- to 5,000+-employee customers. It works because everything is code-based: supporting a new app (e.g. Microsoft vs Rippling) is “as simple as giving our system context on the Microsoft API” rather than rebuilding hardcoded actions (31:58). AI boardroom pressure also made enterprises “ready for early-stage products in a way that hasn’t been true in a long time” (30:52).
Forward-deployed engineering as PM/SE/CS replacement (49:07)
A central org thesis. Serval treats forward-deployed engineers (FDEs) as full software engineers who spend “20% or more of their time talking to customers” and ship product (not one-off customizations) on a tight loop.
“I think the need for a lot of product managers is probably going to go down… you’re going to need a lot less or maybe no solutions engineers… a very limited set of customer success.” (54:37)
Framing: FDEs replace the implementation consultants that Workday/Service Now/Salesforce relied on, and recreate “early co-founder energy” — a CTO hearing feedback and just fixing the product. “We have a call at 10 a.m. and the feature they ask for is shipped at 4 p.m.” (56:51). PM hiring is to be delayed; PMs are envisioned as GMs of a business unit (“Verkada was a thousand people and had two PMs,” 57:57).
Hiring bar — “who are they better than” + recruit-the-next (1:07:49)
Two distinctive practices: a storytelling/life-decisions interview to gauge agency and raw intelligence; and forcing the interview panel to name “who are they better than at the company” to prevent the talent bar from silently sinking.
“We evaluate every candidate on their ability to recruit the next candidate.” (1:11:04)
Bias toward “really smart generalists” who are technical/quantitative over domain experience, because “the job’s going to change.” Referrals have been formalized — some team members “are probably making more money from referrals than from their salary” (1:13:14).
Land-grab framing — fear of being #2 (1:15:26)
“There’s going to be an early-stage competitor to Service Now. Somebody is going to build a next-generation ITSM and we’ve got a short amount of time to position ourselves as the ones to do it.”
Justifies contrarian moves: hiring a “fully baked” GTM team before market traction proved out, raising/going up-market fast. Board/exec pressure to adopt AI means enterprises bet on early-stage startups now — “you could get boxed out of a lot of these accounts if you’re not moving quickly.”
Innovator’s-dilemma argument (1:18:45)
Incumbents (Workday, Service Now, Salesforce) won on configurability, which pre-LLM was hard to deliver alongside simplicity. Startups historically attacked on simplicity (e.g. “what linear does to Jira”) and missed on configurability. AI lets a startup win on both, while incumbents are slowed by backwards-compatibility and revenue-protection. “We built ticketing in months and that used to take years.”
Closing synthesis — Filip vs Hans (1:20:21)
Stauch credits two Verkada leaders: CEO Filip (Kaliszan)^[inferred] — first-principles, “what would we build today” — and chairman Hans (Robertson)^[inferred] — build a better version of an existing market. Serval deliberately fuses both: a better ITSM (existing TAM) plus AI agents that build automations and resolve help-desk requests (never existed before).
Limitations
- Auto-generated captions. Proper nouns and figures are unreliable; corrections applied per the cheat-sheet note above, but verify names/numbers against a second source before any external use. Items flagged
^[ambiguous](e.g. customer “Mercor”) could not be fully resolved. - Thin on cap-table mechanics. The $47M Series A / Redpoint lead comes from the host’s intro, not Stauch; the Sequoia Series B preempt, exact FTE counts, and total raised are better sourced from the Peel interview.
- No hard product metrics — no ARR, customer counts, pricing, or seat economics. Claims are qualitative/strategic.
- Host name not stated in-transcript (addressed as “Brett” in the closing line).