Phase 1 — Gusto Printed the Problem, Then Went Shopping
On June 4, Gusto's Chief Design Officer, Amy Thibodeau, published the company's AI principles and left one problem unsolved in print: the handoff back to a human. The moment the system stops and asks a person to take over, without that stop reading as a failure. Forty-eight days later Gusto posted a requisition that hands the same problem to one hire, with the decision of when not to send AI output written into the requirements. Both documents are public and dated (sourced). The line I've drawn between them is my read, not a company statement.
The real mandate: own the readiness-to-send threshold on Gusto's service platform. The judgment layer between what the machine proposes and what a small-business operator lets out the door to an employee, a tax agency, a customer. Not the automation underneath it.
Urgency verdict: Act Now. (Posted July 22. Today is day 10. Prime window closes day 21, Aug 12. Listing live on Gusto's Greenhouse board as of Aug 1. Provenance: July 22 is the original post date, corroborated outside Greenhouse, and I found no trace of this requisition running earlier under different language, so every deadline below rests on that one date. The record was touched again July 30, which tells you the post changed and nothing about where the funnel sits. Funnel state unknown. No public evidence of active interviewing either way.)
Correction to Issue #5. The Issue #5 Design Roles Board put Gusto at Company 9, Role 12 and filed it Watch. Wrong, and wrong for a specific reason: I scored the company off what its product looks like from the outside instead of what the posting's requirement language actually asks somebody to own. Company moves +2 on AI centrality and design influence ceiling. Role moves +3 on AI exposure quality, craft depth (player-coach, explicit in the posting), and scope expandability (the design seat in a Product–Engineering–Data–Design quartet). Rescored: Company 11/12, Role 15/15. Act. The Company Battlecard holds the org shape and the small-team question. None of that repeats here. Its timing lines are superseded.
Phase 2 — Two Bars, Three Proof Modes
Intelligence-layer check: pass. Gusto's service platform puts machine-generated proposals in front of an operator who is not a payroll expert, a tax expert, or a compliance officer, and asks that person to accept, edit, or kill. Intelligence layer, not autonomous agent. The distinction fixes your entry point: the second between the proposal and the send, not the pipeline that produced the proposal.
The posting sets two separate bars. Most candidates will read them as one, and that gap is where your opening sits.
Bar 1 — AI product judgment. Shipped or led production experiences where a model sits between the user and the outcome, with fluency on uncertainty, error states, graceful failure, human override (sourced: explicit in requirements).
Bar 2 — AI-enabled design practice. Personal working use of agentic coding tools, meaning assistants that write and run code from natural-language instruction, plus ownership of the team's fluency. This is not aspirational filler. Gusto published its own account on June 10: designers trained into the codebase and shipping pull requests, a design system built to be readable by their coding agents, a sandbox running production code, an internal June 1 deadline to stop working out of static Figma files. One designer on the AI product logged roughly 150 pull requests in eleven weeks (sourced). Most Director-level candidates clear Bar 1 and fail Bar 2.
Prepare three proof modes: published AI-product judgment, published organizational leadership under consequence, current production practice.
| Asset | Bar it clears | Where it lands | What it cannot do |
|---|---|---|---|
| Brand Pulse, live, plus "Trust Is the New Interface" | Bar 1 | Sponsor conversation, portfolio screen, AI-product interview | Solo-built; carries no leadership weight on its own |
| Agentic Labs as a set — Brand Pulse, Retail Velocity, Carrier IQ: three products, solo-built, shipped, live | Bar 2 | Portfolio screen, and the "how do you work" question | Not team-practice proof; that part stays verbal |
| American Red Cross disaster relief platform | The organizational mode: six legacy systems into one shared record, three roles acting on the same case for the first time, FEMA audit trails, 10× surge with zero retraining, $847K disbursed and 1,689 cases in two weeks | Portfolio review, leadership panel | Says nothing about current tooling practice |
| TinyFish, context only | Backs Bar 2 verbally: tracing agent runs in LangSmith, pulling signal from session replays and product analytics straight into design calls | Résumé line, recruiter screen, sponsor conversation | Not a case study. Never design proof, never design metrics |
Lead with Brand Pulse and the trust ladder from the essay: Watch, then Verify, then Delegate, autonomy handed over in stages instead of granted at once. You already published a written answer to the question Gusto published without answering.
Bring the Labs forward as a set the moment Bar 2 surfaces. Three shipped, running products built alone is the only evidence you own that speaks straight to a mandate about designers who build. Give Carrier IQ one line: agentic workflow automation in a regulated, form-heavy vertical sits nearer payroll and tax filing than brand sentiment does.
Bring Red Cross next, and name the mechanism, not the domain. Three roles reading and acting on one case record, each with different authority over it, is structurally the same problem as a service platform where an agent, a support rep, and a business owner all touch one payroll event. The analogy is shared authority over a single record, not disaster relief.
Hold Equinox+ and Allē back completely. Consumer behavior design is not what this buyer is solving, and surfacing it blurs a sharp match.
The one objection. You're Head of Product right now. When did you last manage designers? It's hard because the posting asks for 12+ years of design experience including 5+ managing designers, and because the application collects formal-management status and team size in a form field before any human reads narrative (sourced). A checkbox answers this question, not you. Your answer is continuity, not return: Head of Design and Research for North America at Alibaba.com, where you named a structural gap nobody had named, built the research case, secured the executive mandate, and ran three cross-functional sprints across homepage, search, and PDP. Then Product Design Director at Thermo Fisher. Then Product Design Director and GM on the Red Cross build. TinyFish stacked product and commercial scope on top of that line. It didn't interrupt it.
Confirm total years formally managing designers, and largest direct-report count. Neither is derivable from your published record. Both go in the résumé summary and the application field. Both need to be exact.
Phase 3 — Sponsor First, Then Submit
Route strategy is the whole game here. Whether that form field auto-filters isn't publicly knowable (inferred), but the risk only runs one direction: submit cold, and a thin checkbox answer can end the candidacy before anyone opens a portfolio.
Don't submit first. Send the sponsor message, wait two business days, then submit, so a named person is looking for your file instead of finding it in a queue.
Route A (primary): Amy Thibodeau, Chief Design Officer. Bylined on the AI principles post, and credited in Gusto's own account with driving the practice change. Attribution is confirmed on both pages, so skip verification. Channel: LinkedIn DM. The posting names no hiring manager and no reporting line, so her role in this particular search is inferred.
Route B (fallback, day 5 if no reply): Katie Kovalcin, who wrote the designers-to-builders post and built much of the AI product's frontend. Channel: her personal site, kovalc.in, or LinkedIn. She's a practitioner contact, not a router. Ask a craft question, reference her published account, don't ask her to move your application. Title calibration: her site says design lead across product and systems, Gusto's byline says product designer, and a more senior "Distinguished Designer" title shows up only in secondary circulation, unverified against either first-party source.
First contact message
Subject: The escalation question from your AI principles post
In your June piece on how Gusto builds with AI, you flagged one problem as still open: designing the moment work gets handed back to a human without that handoff feeling like a failure. I've been working the same problem from the other side of it.
I published an essay in June arguing that people anchor to a system's worst outcome rather than its average accuracy, and proposing a trust ladder, Watch to Verify to Delegate, as a way to hand over autonomy in stages instead of all at once. Before that I led a national disaster-relief platform at the American Red Cross where three roles shared a single case record for the first time, and the entire design problem was who could act on what, and when. I'm currently Head of Product at TinyFish, an enterprise web agent platform, so my days go to agent traces, attribution, and reversibility in production deployments.
The Unified Service Platform role reads like it owns that threshold. The question I'd most want to ask you: does design own the readiness criteria for AI behavior at Gusto, or inherit model behavior and design the interaction layer around it? Would you have 20 minutes?
Résumé framing note
Open the summary with function, tenure, headcount, in that order: design leader, 13 years, X years leading design teams, N direct reports at peak. That line has to clear the same bar the form field checks, alone, with no narrative propping it up. Surface the Red Cross consolidation numbers first (six systems to one, national deployment, 1,689 cases in two weeks), then Alibaba's $50B GMV and +20% transactions as scale proof. Cut Equinox+ and Allē to a line each, or drop them. TinyFish appears as current role and production-agent currency only, and "Head of Product" must never be the first identity a reader meets. Delete "transitioning back to design" everywhere it appears in writing. You're a design leader who took a product mandate to build AI-natively from zero. Say that out loud in a room, with confidence. Don't bury it as a caveat in a document.
Cover letter hook
In June, Gusto published its AI principles and named intelligent escalation, the handoff back to a human that shouldn't feel like a failure, as a question it hadn't answered yet. This is the first design mandate I've read that puts the decision not to send inside the job description, and that line is why I'm writing.
Phase 4 — Window Summary
| Action | Deadline | What degrades without it |
|---|---|---|
| Confirm management years and headcount; send Thibodeau message | Mon Aug 3 | Sponsor route needs two business days of lead time; after Aug 5 you're submitting cold into a form field |
| Submit with reframed résumé summary | Wed Aug 5 | Prime window closes day 21 on Aug 12; later applications land against a slate already in loops |
| Kovalcin fallback contact, then verify status with recruiting or retire the target | Wed Aug 19 (day 28) | Past four weeks the funnel is likely closed; the listing exits the board at six weeks (Sept 2) |
September 2, 2026.
- Bar 2 in Gusto's own words: The internal account of the code-first transformation, including the Workbench design-system MCP, the production Sandbox, and roughly 150 pull requests across an eleven-week build, is effectively the scoring rubric for the practice half of this role — reread the designers-to-builders post the night before any portfolio screen.
- Fieldguide as the cleanest domain parallel: Regulated audit workflows, multi-party accountability, and practitioner oversight of a long-horizon multi-agent system make the Staff Product Designer role the closest published match to Thermo Fisher mySupply, and it arrived on the back of a $75M Series C led by Goldman Sachs at a $700M valuation.
- Vanta's 40-person stretch mandate: The Head of Design role covers GRC, Trust, Platform, self-serve, and AI at VP scope, and the company's technical write-up on giving its agent a computer tells you exactly which supervision surfaces that org will be designing next — treat organizational scale as the evidence gap and reporting line as the decision gap.
- Where the rubric and the money disagree: Amplitude scores 15/15 on role with a $317K–$476K San Francisco base yet lands on Watch purely on company stage, so read the Head of Product Design posting against your own equity-window tolerance before dismissing it on score alone.