What Defines This Tier
Headway, Gusto, Ramp, and their peers form a tier because design quality, for all of them, is judged by system behavior under failure. A billing error that reaches a provider's income statement. A payroll run that shorts an employee. A payment authorization that freezes a business's operations on a Friday afternoon.
Most design portfolios showcase success states. Clean dashboards, smooth onboarding, delightful moments. This tier's evaluation room is scanning for something else: evidence you've designed for the exception state, in a context where the exception has a paper trail and a dollar amount.
Your portfolio has that evidence. The problem is sequencing. Lead wrong and you sound like every other senior candidate with a strong background. Lead right and you sound like someone who has already solved the problem they're hiring for.
Reading the Postings Through Headway
I read job postings the way a radiologist reads scans. What's present matters. What's absent matters more. Headway's two Director-level design postings are the clearest window into how this entire tier evaluates.
The postings reveal the evaluation DNA: provider workflows, insurance complexity, design systems, AI-native features inside consequence-bearing contexts, and an explicit player-coach expectation that includes prototyping with Claude and Cursor. That last detail matters. They're asking whether you build with these tools, today, in a domain where a wrong output reaches a provider's income statement.
Headway's EHR announcement corroborates the posting signals. It describes AI-assisted progress notes that are "compliant, clawback-protected, insurance-ready." That compound adjective tells you the failure mode they're designing against: a note an insurer rejects, triggering a clawback that costs the provider money already spent. The design problem is making the provider confident enough to sign the note, knowing the system got it right. The postings sit inside a product organization led by Jake Poses, VP of Product, who oversees product, product design, and product marketing. Speculative, based on unverified org-chart data: Ellen Dong holds a Head of Design title above the Director level. If confirmed, this means you'd be leading a vertical within an established function, with air cover above you.
Now the corroborating signals.
Gusto's Senior Product Design Manager, Payroll posting is the most explicit about the evaluation question. It asks for "experience leading AI or automation initiatives where uncertainty, graceful error surfacing, and user agency matter when automation gets things wrong in high-stakes contexts." They are asking whether you have designed for the moment AI is wrong, when being wrong reaches someone's paycheck. Their careers page reinforces: "accuracy, trust, and real consequences."
Ramp's Director, Product Design posting (published June 8) describes its problems as "high-stakes, data-dense, and unforgiving." Ramp automates over $200B in annualized spend. The posting describes designers who "ship PRs" and build agents. High confidence this is the most AI-forward company in the tier, and the Claude/Cursor prototyping expectation appearing at both Headway and Ramp confirms it's a tier-wide signal, not an outlier.
The shared evaluation question across all three, high confidence: When the system produces a wrong output in a consequence-bearing context, what did you design to happen next?
Your Lead Evidence, by Company
Same background, different lead edge based on what each buyer is afraid of.
For Headway: Lead with Allē, Back with Thermo Fisher
Allē maps directly to Headway's structural reality. Two user populations with different needs, connected by a system that has to serve both.
Headway has providers and patients. Allē had 30M+ members and 40K+ provider practices. The consumer app redesign turned a points wallet into a treatment timeline. The provider backoffice redesign turned a billing tool into a patient relationship dashboard with tier visibility, inactivity signals, and one-click outreach. The connection between the two surfaces drove the outcomes: 3.2x redemption, 47% reactivation of lapsed members, $42 CAC down from $92.
Emphasize the provider backoffice. That is the surface Headway's evaluators identify with. They build tools for professionals managing complexity on behalf of someone else. A provider dashboard that surfaced which patients were lapsed, at risk, or near milestones is the same design problem as a clinical documentation tool that surfaces which notes are at risk of insurer rejection before the provider signs.
The Allē dual-surface work also answers the design-systems signal in Headway's postings. Designing systematic patterns across two distinct user populations with different mental models, different task frequencies, and different error tolerances is design-systems work at the architectural level, well beyond component libraries. Name it that way.
Back the Allē case with Thermo Fisher's exception-first architecture to prove you can design for the moment the system is wrong. The old system showed 197 orders in a flat list. Your redesign defaulted managers to Exceptions. That single decision tells the evaluator where your instincts go: the primary view should be the failure state.
One note on Allē: do not overstate it as compliance or HIPAA proof. The case demonstrates dual-surface healthcare-adjacent design, not deep regulatory navigation. Let Red Cross and Thermo Fisher carry the compliance weight.
For Ramp: Lead with Thermo Fisher
Thermo Fisher mySupply is your strongest opening here. Your research traced five failure modes across 32 interviews and translated them into five modules: exception-first order management, batch status transparency, bidirectional KPIs, forecast collaboration, capacity planning. The failure modes drove the architecture. Every module exists because a specific failure mode demanded it.
Lead with this sentence:
"Partners were discovering batch problems at the delivery gate, when rescheduling cost 3-5x more than prevention."
This tells the evaluator you understand that the cost of a system error is the cascade it triggers when caught too late. Ramp's evaluators live inside that cascade every day with $200B in automated spend.
For Ramp specifically, pull the agentic thread further than you would elsewhere. The Thermo Fisher agentic vision, where five modules become five agents while batch QA release remains a human gate, maps directly to Ramp's operating reality. One structural signal worth noting: Ramp's Head of Brand posting reports to a VP of Design, confirming executive-level design representation. Positive.
When they ask about your current work, frame TinyFish as the practice where you're building the design methodology for agentic systems in production. Keep it verbal. Keep it philosophical. "I'm working on the operating contract between human judgment and autonomous systems" positions your current thinking without pointing to case studies you can't share.
For Gusto: Lead with Red Cross
Red Cross answers the question Gusto is actually asking. Six months to national deployment. Six legacy systems replaced. Caseworkers on slow networks with untrained volunteers during active disasters. Federal oversight from FEMA. The design had to work when everything around it was failing.
What matters for Gusto: FEMA documentation requirements weren't a compliance review step layered on top of the workflow. They were embedded in the workflow itself. The constraint dissolved into the product so it never became a separate gate. That is exactly what payroll compliance needs to look like. Gusto's evaluators know that payroll runs have exceptions because payroll has exceptions. They need evidence you can build the regulatory requirement into the flow rather than bolting it on as a checkpoint.
Red Cross also answers the unspoken concern that exception-first design means slow, cautious, process-heavy design. Federal oversight, national deployment, six months. Fast and right under consequence.
Use their language back to them. Their posting says "graceful error surfacing" and "user agency." Your Red Cross case demonstrates both under conditions more extreme than a payroll run.
Tier Deviations Worth Knowing
Not every company in your target list fits this tier cleanly.
Gusto's posting is Senior Product Design Manager, not Director+. Their public board showed no Director-level product design title at the time of checking. The role is real leadership (managing designers, setting UX direction for Payroll, leading AI initiatives), but the title and scope may signal a narrower mandate than what Headway and Ramp are offering. Calibrate expectations for function-building authority accordingly.
Ramp is the most AI-forward and least traditionally regulated. Finance automation carries consequence, but Ramp's culture self-identifies around speed and builder identity ("hands off doesn't exist here"). The consequence reframe still works, but lead with the agentic architecture story, not the compliance story. If you open with regulatory language at Ramp, you risk reading as risk-averse in a culture that prizes velocity.
Headway is the deepest in regulated territory. Insurance complexity, clinical documentation, HIPAA-compliant telehealth. Of the three, Headway is where the consequence frame needs the least translation. It's also where the design-systems expectation is most explicit, which plays to your Allē dual-surface evidence.
Pin this section. These framings sound senior. They fail with this tier's evaluators.
What Gets You Killed
"I elevate craft quality across the organization." This tier has a consequence problem. Craft language signals you will focus on visual and interaction polish while the system produces wrong outputs that cost users money. Lead with system integrity.
"I've designed AI-native products." Everyone has. Gusto's posting explicitly asks whether you've designed for when automation gets things wrong. Saying you've designed AI features without addressing AI failure modes is the equivalent of showing a success-state portfolio to an evaluator scanning for exception states.
"I bring a consumer-grade experience to enterprise workflows." This implies simplification. These workflows are complex because the domain is complex. A payroll run has exceptions because payroll has exceptions. A clinical note needs specific language because insurers require specific language. The evaluator hears "consumer-grade" and worries you'll sand off edges that exist for regulatory or operational reasons.
"I build design culture." These companies need someone who can ship a provider workflow that handles insurance exceptions correctly under time pressure. Leading with culture-building sounds like you're solving a problem they don't have yet while ignoring the one keeping them up tonight. Culture follows from doing the work well. Lead with the work.
"My consulting background means I can ramp quickly." Do not open with consulting framing. If they ask about BCG DV, answer with deployment evidence. "Every engagement shipped into production with real users and real operational consequences" makes the employer logo irrelevant. Do not get drawn into a definitional argument about what counts as consulting.
The Player-Coach Diagnostic
Every company in this tier describes the role as player-coach. You need to know which version they mean before you invest further.
The real version: senior judgment exercised close to the artifact during early ambiguity, transitioning to team standards and systems as the product stabilizes. You stay hands-on because the artifact is how strategy gets tested. You build the team because the system needs to scale beyond your hands.
The dangerous version: under-resourced execution disguised as a leadership expectation. They need someone doing the work of three designers while also attending leadership meetings, and they're calling that "player-coach" because it sounds like a philosophy rather than a staffing gap.
Ask this in the first conversation:
"When the last significant design decision was made on this team, walk me through who was in the room, what the inputs were, and how the final call happened."
This reveals whether design decisions are made by designers or ratified by PMs. It reveals whether "player-coach" means the leader is in the room making calls or in Figma making screens. And it surfaces the actual decision-making altitude of the role without asking the direct question, which always gets an aspirational answer.
If the answer describes a designer presenting options to a product or engineering leader who chose, the role is execution with a leadership title. Pass.
Reading the Company Back
You are also evaluating them.
Signals the mandate is real:
- Design reports to someone who has managed design before. At Headway, Jake Poses led a 50-person product team at Thumbtack that included design. High confidence he understands what a functional design org needs.
- A named design leader exists above the Director level. Speculative for Headway based on a single unverified org-chart source. If confirmed, you're leading a vertical within an established function, with someone above you who understands what you need. Verify this in your first conversation.
- The posting describes outcomes, not activities. Ramp says "define product direction" and "ship alongside designers, PMs, and engineers." Compare that to postings listing "create wireframes and prototypes" at the Director level.
- The company's product announcements use design vocabulary that matches the posting vocabulary. Headway's EHR announcement and its postings both speak in terms of provider confidence, insurance compliance, and documentation quality. Alignment between external narrative and internal role definition is a positive signal.
Red flags worth walking from:
- No design leader above the Director role, and the Director reports to a GM or engineering lead who has never managed design. You are the air cover with no one above you who understands what you need.
- The posting emphasizes "cross-functional collaboration" repeatedly but never mentions design's role in decision-making. Design has no seat at the table. They're hoping you'll fight for one.
- Player-coach language combined with a team of two or fewer. That's a senior IC who also does 1:1s.
- The company's public product announcements use design language ("intuitive," "seamless") but the job posting uses product-management language ("drive alignment," "manage stakeholders," "build consensus"). That gap between how they talk about design externally and how they describe the role internally predicts whether you'll be empowered or managed.
The Throughline
When you walk into any conversation with this tier, the story is one sentence long:
"You have spent your career designing systems where being wrong means a financial event, a clinical consequence, or a federal compliance failure."
Your portfolio shows what you designed to happen when the system doesn't work. That is what they're evaluating for. Lead with it, and sequence everything else behind it.
- Gusto's AI-failure language: Gusto's Senior Product Design Manager posting explicitly asks for experience with "graceful error surfacing" and "user agency" when automation gets things wrong, which is the most direct articulation of exception-state evaluation in any current posting across this tier.
- AI trust as governance reality: NIST's Generative AI Profile frames trustworthiness as a lifecycle concern spanning design, deployment, and incident disclosure, which gives you external vocabulary if an evaluator asks how you think about AI risk beyond the product surface.
- Design maturity as two-directional filter: NN/g's UX maturity research identifies hidden weaknesses like unsupportive leaders and development processes that exclude iterative design even in organizations that appear structurally mature, which reinforces why the player-coach diagnostic and red-flag inventory matter before you commit.
- Reference calls are expanding: Riviera Partners and other retained search firms are treating AI leadership hiring as a distinct executive-search category in 2026, which means your positioning needs to survive compression by a search partner who may not understand design leadership nuance.