Verdict: Act. Company 10/12. Role 14/15.
Contact Jeremy Epling, Chief Product Officer, this week. Cold LinkedIn. Hook on his own 20→90 number, not the job post.
The urgency here is a calendar. Vanta has said in public it will ship more than ninety agentic workflows by December 31. Today is August 1. The design seat that is supposed to sit over those launches has been vacant since late June. No company sits on that seat by choice this late into a stated shipping year.
Issue #5 nailed down Kawamoto's departure and left one thing hanging: whether Vanta is using the succession to quietly change the job. It is. The proof is sitting on Vanta's own resources page, and it comes down to who signed which post.
The sequence, by byline
Vanta sells evidence that auditors and compliance officers are willing to stand behind. Hold that in mind through the order below.
July 23, 2025 — Series D, $150M, $4.15B valuation, up 69% in a year. Confirmatory.
September 2025 — the design leader owns the AI story. Kawamoto, then VP Design, sits for a long interview: designers interviewing customers about AI, citations back to source, human approval steps, telling the user what the system is doing before it shows a result. Read the vocabulary. It is yours. Confirmatory, and the baseline every later event should be read against.
Undated, before March 2026 — Epling in seat. His start date is not recoverable from the public record. The March interview is the earliest confirmation of the reporting line, and in it he talks like an owner with mileage, not a new arrival. Work from that. He is not a CPO in his first ninety days placing a structural bet on design; he is patching a hole in a machine he already assembled. You do not have to sell him on design. You have to answer who enforces quality at 4.5x throughput. Trajectory-changing, and it moves your framing further than the job post does.
March 2026 — Epling defines the org. In a First Round profile he states that product, engineering and design all report to him, and that he manages the design leader directly. He calls himself the single accountable owner for execution and product quality. Trajectory-changing: settles the buyer question, and the ceiling question with it.
April 29, 2026 — $300M ARR, nine months after $200M, growth rate up four quarters running. Company-reported, so discount a notch. The shape matches the funding. Confirmatory.
May 20, 2026 — the number. Epling writes that Vanta shipped twenty agentic workflows over the prior year and is targeting more than ninety by the end of 2026. Call it 4.5x. Trajectory-changing.
June 3, 2026 — engineering publishes the quality regime. SVP Engineering Iccha Sethi posts how Vanta builds AI products: task-specific evaluations, compliance experts calibrating the evaluators, versioned offline datasets, tracing on every customer-facing AI call, a flat refusal to ship things that demo well and collapse in real workflows. No designer is named anywhere in it, and no design approval step appears in the pipeline. Trajectory-changing.
Late June 2026 — Kawamoto leaves. She says so herself on July 9: stepped away for health reasons, grew the team from five to forty-five, left a leadership bench, they are hiring her backfill. No new employer as of today.
Now — the Head of Design posting is live. Roughly forty people, director-level reports across GRC, Trust, Platform, Self-serve and AI, design system ownership, definitions of done, launch-review frameworks. $365K–$513K cash plus equity.
Three gaps in the record.
- Epling's start date does not exist in the public record.
- The posting states no reporting line. Epling's March interview is the sole source, and interviews age.
- Vanta's design careers page still lists Kawamoto as VP Design. So does her own site. Both are stale against her dated statement, which outranks them. Do not read that staleness as an open succession. It is closed.
At the pitch stage I read the $148K spread in the salary band as evidence of an unresolved internal level. Wrong. The posting attributes its ranges to location and benchmarking. Drop that read.
The pattern in the bylines
Over nine months, authorship of Vanta's AI quality story moved from design to engineering, and this posting is an attempt to buy it back.
September: the person explaining how Vanta earns trust in AI output is the design leader. June: same subject, deeper operational detail, named methodology, and the author runs engineering. Design does not appear in the loop. Vanta's published AI principles name an internal review committee staffed from product, engineering, security, compliance and legal. Design is not on that list either.
Now look at what the job description drops. It does not lead with growth, hiring, or team building, which is the thing Kawamoto was brought in to do. It leads with consistency, definitions of done, launch review. That is a control mandate, and it surfaces exactly as the volume of things requiring control is about to quintuple.
The launch-review language does not describe how Vanta runs today. It describes a seat Vanta wants occupied inside a system that currently runs fine without one. Succession in form, reorganization in substance.
Confidence: high on the authorship migration; it is documentary. Moderate on what that implies about design's real authority, which is inference from silence.
What happens next, and what would make me wrong
Vanta hires on quality-system credibility, not org-scaling credibility, and closes faster than a forty-person-org search normally closes. A public year-end number is the forcing function. Expect the shortlist to reward anyone who can describe how a review threshold gets enforced on an agentic release, and to quietly discount anyone whose best material is a headcount curve.
Second-order call, lower confidence: whoever takes this job will be told they own launch review and will find out around month three that engineering's evaluation pipeline holds the actual go/no-go. That is not cynicism; it is what the authorship record shows, and it is answerable in your first conversation.
What would make me wrong. If this goes to an internal promotion off the bench Kawamoto named on her way out (Conor Sheehan, Jerrod Larson, Braden Kowitz, Jenny Chang), then the mandate language is boilerplate and the external search is calibration. That risk is live, and it is the biggest single threat to your time here. Separately: if a design leader turns up as author or named owner on Vanta's next quality-methodology post before this role is filled, design already holds the gate and I have read silence as absence.
What to do differently
Rebuild the application around quality-system credibility. Retire organizational credibility from the lead.
Your reflex on a forty-person head-of-design role is to open with the shape of the org you ran and the function you built. Wrong instinct here, and worse than wrong: it parks you directly on top of the strongest internal candidate instead of beside them. Vanta already grew a design team. Then it watched authority over the AI quality story drift out of that team, and it is now buying the missing enforcement layer back under a familiar title.
So every artifact leads with how a review threshold gets specified and enforced. Org scale drops to a footer credential you never argue; it sits under Alibaba and answers itself. The five-to-forty-five story, the one that would have won you the previous version of this job, stays out of the first conversation entirely.
Rubric, argued
Company: 10/12
| Dimension | Score | Argument |
|---|---|---|
| AI centrality | 3 | The agentic trust platform is the positioning. 20→90 is the roadmap. |
| Stage and equity | 2 | $4.15B a year ago, $300M ARR now. Real. But no public tender, no disclosed employee liquidity program, nothing on secondary terms. Upside you diligence, not upside you assume. |
| Design influence | 2 | The posting claims 3. The byline record says 2. Score the record. |
| Trajectory | 3 | Growth rate accelerating, four quarters consecutive. |
Role: 14/15
| Dimension | Score | Argument |
|---|---|---|
| Comp | 3 | Clears your floor with room. |
| Scope expandability | 3 | Reports to a CPO who owns product, design and engineering. No layer above you inside design. |
| Craft depth | 2 | The posting names interaction quality and visual bar. Forty people and a director layer will eat that inside two quarters. Most fragile score on the card. |
| AI exposure | 3 | Release review at 4.5x throughput is the hard version of the problem. |
| Portfolio value | 3 | Publishable, and nobody has written it yet. |
Positioning
Lead with the Decision Gate. Not trust as a theme. Every candidate will say trust at a company named for it. Lead with the narrow claim: you design the moment authority changes hands between a system and a person, and you turn that moment into something specific: visibility, interruption, reversibility, signoff.
Then bolt it to their number. Ninety workflows is ninety transfers of authority. Every one of them contains a moment where a compliance professional either accepts machine output or overrides it. Specifying those moments consistently is what a launch-review framework actually is. Say that sentence out loud before you write a word.
The lead case changes by artifact, on purpose. Trust Is the New Interface leads in the message and on the resume; it is the framework, it is public, it is checkable in thirty seconds. Thermo Fisher leads in the cover letter, because Epling's team just published an operational quality regime, and in that room a shipped gate outranks a demo no matter how well built.
- Framework proof: Trust Is the New Interface (five handoffs, Watch → Verify → Delegate), plus Agentic Labs. Carrier IQ first: regulated domain, all five handoffs visible in a working build.
- Shipped proof: Thermo Fisher mySupply. Exception-first design across a pharma supply chain with one irreducible human gate at batch QA release. Same architecture Vanta needs.
- Scale anchor: Alibaba.com. $50B+ GMV, +20% daily transactions, trust signals designed into search and product pages. Answers the forty-person question so you never have to.
- TinyFish: one sentence. Current context and production currency. Never portfolio proof.
First contact with Epling on LinkedIn
Opening line / subject: Ninety workflows, ninety review moments
You've put the target at more than 90 agentic workflows by year end, up from 20. Your engineering team's post on how Vanta builds AI products reads like a control regime that already works: evals, calibration by compliance experts, tracing on every customer-facing call.
What I keep turning over is the half downstream of that. At ninety workflows, the binding constraint stops being whether output is correct and becomes whether a compliance professional can tell, in the moment, what to accept and what to override. Ninety of those moments designed inconsistently is a trust regression no eval suite catches.
I published a piece in June on that exact handoff, and I've enforced it once for real: a pharma supply-chain platform rebuilt around five continuous agents and exactly one surviving human gate at batch release. I'm Head of Product at TinyFish, an enterprise web agent platform, and returning to design for a vertical problem.
Twenty minutes on how you're thinking about launch review as workflow count scales?
No warm path surfaced. Before you send, check your own network for three or more relevant mutuals. If they're there, name them in the closing line. Never the opening.
Cover letter draft
Vanta shipped twenty agentic workflows last year and is targeting more than ninety by December. The hardest design problem inside that number sits one layer under the workflows themselves: each one creates a moment where a compliance professional has to decide whether to accept what the system produced, and ninety of those moments designed inconsistently is a trust failure that arrives after every eval has already passed.
I built the enterprise version of this at Thermo Fisher. The world's largest contract manufacturer was running six pharma partnerships on spreadsheets, and exceptions were surfacing at the delivery gate, three to five times costlier than catching them a week upstream. Thirty-two floor interviews, five failure modes, five modules: exception-first order management, at-risk batch flags, bidirectional KPIs, forecast auto-load, an eighteen-month capacity heatmap. $20M+ in annual margin recovered, 83% IRR, adoption by all six partners inside twelve months. When I redrew that architecture for agents, it resolved to five agents running continuously and one surviving human gate: batch QA release, where a regulatory signature can't be delegated. Deciding which gate survives automation is the judgment a launch-review framework has to encode, ninety times over.
I'm currently Head of Product at TinyFish, an enterprise web agent platform, where agent traces, attribution and reversibility in production deployments are weekly work rather than theory. I moved into product to build AI-natively from zero and shipped three products in the first three months. I'm returning to design because I want that depth pointed at one high-stakes vertical instead of a horizontal platform, and because I miss the craft.
The essay on the five handoffs in agentic systems, and the agentic work behind it, are at junochen.com. I'd welcome a conversation about what launch review means at Vanta's current shipping rate, and what authority the role carries when quality and ship date disagree.
Resume, ten minutes
- Lead: Agentic Labs and Trust Is the New Interface. AI exposure is the deciding rubric dimension. Put it above the fold.
- Second: Thermo Fisher. Pull the exception-first detail and the single-surviving-gate line up. That's the launch-review analogue, and it works harder than the headline metrics.
- Third: Alibaba. GMV and org scale stay intact.
- Trim: Red Cross and Equinox+ to one line each. Not mission-driven, not consumer.
- Header: "Head of Product, TinyFish (enterprise web agent platform, Series A)." One line in the body on three products in three months and daily work in agent traces. Stop there.
Who else is in the pipeline
- GRC-domain design leader. Moderate confidence as an archetype; speculative for this pipeline. Drata, OneTrust, some security vendor: someone who knows compliance workflows cold. Beats you on domain fluency. The objection is "she doesn't know compliance." Counter: Vanta already employs compliance experts to calibrate its evaluators. What it doesn't have is anyone who has designed the human override moment at scale.
- AI-native design leader. Moderate confidence. Frontier-lab pedigree, better brand recognition than yours. The objection is "she's coming from product, not design leadership." Preempt it in writing, the way the cover letter does, and let Alibaba sit underneath unargued.
- Internal director promotion. Speculative, structurally likely. Kawamoto named four leaders publicly on her way out. If one of them wants it, the external slate is calibration. This is why you send in days, not weeks.
In the room
Read Sethi's post before the call. Vanta has already published its AI quality methodology, and it came out of engineering. Don't walk in with a philosophy of AI quality. Walk in having read theirs, and ask the question it leaves open.
First fifteen minutes: Who holds go/no-go on an AI product launch today, and does this role inherit that authority or design a framework someone else applies?
Then the harder one: what happens when an AI quality regression and a ship date collide, and whose call is it? The posting names no six-month success measure. Epling's stated product outputs elsewhere are customer love, satisfaction by area, velocity, differentiation and per-product revenue. Neither list contains a quality-regression measure. If nobody in the loop can answer, you've just learned the real scope of the job.
Don't:
- Don't lead with five to forty-five. The posting deliberately drops team-building language. Leading with it tells them you read the last job, not this one.
- Don't critique the engineering methodology post. Read it, credit it, ask what it leaves open.
- Don't cite TinyFish product work as design proof. It isn't published, and at a company that audits evidence for a living, an unverifiable claim costs you more than a missing one.
Small discrepancy worth a question: Kawamoto says forty-five people. The posting says roughly forty. Ask what changed.
Warning signs
- You'd be hired as a gate during a velocity push. Gates get routed around by Q3 once the year-end number is public. Get the authority into the offer, not into a conversation.
- Design is absent from the named AI governance functions. That's the influence-ceiling risk, plainly stated, and it's why the company score is 10 and not 11.
- Equity is unpriced. The $4.15B mark is a year old and set at lower ARR. Tender history and 409A timing are offer-stage questions. Don't raise them earlier.
After you send
Growth-stage cadence. Three to five business days is a normal reply window. Past seven means the search is further along than the posting suggests, or an internal candidate is in play. One follow-up on day eight anchored to a new public signal, then redirect. If Epling comes back engaging the review-load problem instead of handing you to a recruiter, the hook landed: answer inside twenty-four hours and go deeper on the pharma gate architecture. A recruiter-only reply still counts as a landing; use the screen to reach him. Any request for a portfolio walkthrough gets built around the Decision Gate and nothing else.
Epling, LinkedIn, this week. The number is public, the seat has been empty since June, and the internal bench has had five weeks to decide whether it wants the job.
- Epling's own quality language: If you would rather open on the control regime than the workflow count, his first-person post promoting Trustcraft gives you his exact phrasing — real evals, full traceability, not just great demos — to mirror before introducing your own framing.
- Design credit on the agent launch: The one public instance of a Vanta designer being named on agentic work is the AI Agent launch credit from product lead Malvika Nagpal, which tells you design was in the room but says nothing about who held approval — worth a question, not a claim.
- Pricing the equity before the offer: The only public window into Vanta secondaries is a private-market listing page for potential access to Vanta shares, which discloses no executed transaction, price, or company-approved employee tender — treat the equity line as undiligenced until Epling's team gives you tender history.
- The parallel agent-oversight seat: If the internal bench takes Vanta, the closest structural match on the board is HackerOne's Director, AI Product Design, reporting to the CPO with a mandate covering agent-to-agent experiences and human-in-the-loop control in an adversarial domain.