Trigger: Three staff-level provider design roles posted August 3–4: Onboarding, CRM & Prospect Management, Group Practices. The CRM posting names an AI front desk and scheduling assistant where uncertainty or incorrect behavior reaches a real patient. Confidence: high on the cluster, moderate on the coherence inference below.
Window: 19 days old. Headway's product org hired 25 people in the four months before these postings. Three simultaneous staff roles under one director means the pipeline is funded and moving. Act this week. After Labor Day the competing-candidate pool thickens.
What this probably means for the design org
Roland Tiangco is carving the provider journey into owned domains. He joined late 2025 to lead provider design (per Ellen Dong's year-end team post), and these three roles are his first major staffing push — each covering an adjacent surface with its own AI-assisted workflows.
None of the postings say who owns the shared pattern for how those AI assistants handle consequential actions across the provider experience. The CRM designer is the likely answer, because the front desk agent is where AI first acts on a patient without per-action provider review.
Rubric
| Score | Note | |
|---|---|---|
| AI centrality | 2 | AI is a primary bet (Scribe, AI assistant) but Headway is a marketplace platform |
| Stage / equity | 3 | Series D, $100M, $2.3B valuation (Jul 2024). Pre-IPO |
| Design influence | 2 | Head of Design (Ellen Dong) exists; design reports into product under VP Product |
| Trajectory | 3 | 25 product hires in 4 months, 3 simultaneous staff design roles, Medicare/Medicaid expansion |
| Company | 10 | |
| Comp potential | 2 | $212K–$265K + equity. Solid cash; equity upside depends on undisclosed terms |
| Scope expandability | 3 | Staff IC setting the AI delegation pattern across provider experience |
| Craft depth | 3 | Hands-on IC designing the hardest AI problem in the provider org |
| AI exposure | 3 | Delegation boundaries, patient-consequence design, trust architecture |
| Portfolio value | 3 | Three-party delegation model in healthcare — differentiated case |
| Role | 14 |
Verdict: Act.
Narrative fit: Thermo Fisher's regulated multi-party supply chain is the closest published analog. Pharma partners delegated procurement decisions affecting downstream patients with no per-action review possible. The Trust essay's Delegate tier provides the framework for the CRM role's central question: how to design bounded autonomy when the affected party is absent from the delegation decision.
Your angle
Lead with Thermo Fisher mySupply (Product Design Director at BCG DV).
Target the CRM role out of the three. Headway already has a working authority model for consequential AI: Scribe. Every AI-generated clinical note gets provider review, editing, and sign-off before it enters the record. Six million notes since September 2025. That works because the provider is present at the moment of consequence.
The front desk agent can't work that way. A scheduling assistant that pauses for provider approval before every patient interaction defeats the purpose. The provider delegates, the agent acts, the patient experiences the consequence — and the party who bears the risk has no control surface. The CRM designer will need to build a different authority model: bounded autonomy where the agent acts within defined limits while the provider retains accountability for outcomes they didn't individually review.
Thermo Fisher is the direct match. Six pharma partners delegated procurement decisions across nine sites. Downstream patient access to medication depended on system actions no partner reviewed individually. You designed the authorization boundaries that made bounded autonomy trustworthy in a regulated, multi-party workflow. $20M+ in recovered margin, 100% partner adoption, 12 months from zero.
The Trust essay's Delegate tier extends this into the CRM role's specific territory — what the delegation boundary looks like when the affected party has no seat at the table.
The Human-Agent System Design domain in your forward-looking artifacts maps precisely to this three-party delegation problem. If an artifact in that domain is published or near-complete at junochen.com, lead with it alongside Thermo Fisher. If not, move now with Thermo Fisher + Trust essay — the window is 19 days old. Flag this build as high priority: having it live before a portfolio walkthrough changes how Dong and Tiangco evaluate your AI-native depth.
Secondary: Allē (Product Design Director at BCG DV). Dual-surface design where provider-side decisions directly shaped the consumer experience. 30M members, 3.2× redemption. Structural parallel to Headway's provider-patient split.
Who to contact
Primary — Roland Tiangco, Director of Provider Product Design
- Public fingerprint: Effectively silent for six months. No LinkedIn posts, conference talks, or Headway blog bylines found. His most recent public signal is from November 2024 when he joined Faire — marketplace discovery language, irrelevant to his current mandate. Prior roles: Director of Product Design at Faire, Senior Manager at Shopify, plus eBay and Grailed.
- Warm path: No verified co-instructor, ADPList, Dive Club, Designer Fund, BCG DV, or Alibaba network overlap. No public evidence of 3+ mutual LinkedIn connections or shared design-community groups. Assume cold unless your authenticated LinkedIn graph shows mutual connections — if it does, use them in the close, not the opening.
- Channel: LinkedIn cold DM. Since Tiangco has been publicly silent, the CRM posting's patient-consequence language is your hook rather than his own words. If the hook doesn't land, that gap may be why — calibrate accordingly.
Secondary — Jake Poses, VP of Product
- Public fingerprint: Active on LinkedIn. Describes the provider org as building "telehealth, calendar, and AI-note tools" to reduce administrative burden. Favors candidates from top growth and pre-IPO startups. His own interview process included a PRD exercise and 10+ back-channel references.
- Channel: LinkedIn cold DM. Poses is the product sponsor, not the scope-specific buyer. Contact him only if Tiangco doesn't respond within 5 business days, or if you discover a warm path to Poses first.
Likely evaluator — Ellen Dong, Head of Design
- Dong leads the full design function (Product Design, Creative, Research) and has been at Headway since January 2022 as the first designer. She won't be your initial outreach target, but expect her in the evaluation loop. Her hiring process includes an app critique and whiteboarding exercise, and she says candidates' questions reveal the depth of their thinking. She values "taste and craft." Prepare the junochen.com walkthrough for her lens — she'll be evaluating design judgment, not domain fit.
Outreach messages
Subject line: Delegation boundaries when the patient bears the risk
Cold message to Roland Tiangco:
Your CRM posting asks how an AI front desk should behave "with a real patient on the other end." That's the design problem I keep returning to — not the review surface (Scribe already handles that) but the delegation problem where the provider delegates, the agent acts, and the patient bears the consequence before anyone reviews.
At BCG Digital Ventures, I designed Thermo Fisher's mySupply — a regulated multi-party supply chain where pharma partners delegated procurement decisions affecting downstream patients. Six partners, $20M+ recovered margin, 100% adoption. That work and a framework for designing bounded autonomy in agentic systems are published at junochen.com.
20 minutes on delegation boundaries vs. review surfaces?
Subject line: The front desk agent's authority model
Warm message (adapt if a mutual connection surfaces):
[Name] mentioned you're building out the provider design team at Headway. The CRM role's AI assistant problem — what happens when the agent reaches the patient before the provider reviews — maps directly to work I led at BCG Digital Ventures on Thermo Fisher's regulated multi-party supply chain ($20M+ margin, 100% partner adoption). That case and a framework for delegation boundaries in agentic systems are at junochen.com.
Would 20 minutes work to discuss how you're thinking about the front desk agent's authority model?
Subject line: Authority boundaries beyond per-output review
Cold message to Jake Poses (backup, day 8):
Your description of the provider org — telehealth, calendar, AI-note tools reducing administrative burden — names a design problem that gets harder when the AI acts on the patient before the provider can review. That's the specific challenge in the CRM role, and it's structurally identical to work I led at BCG Digital Ventures on Thermo Fisher's regulated multi-party supply chain (six pharma partners, $20M+ recovered margin, 100% adoption in 12 months).
More at junochen.com. 20 minutes?
What not to say:
- Don't call the AI assistant a chatbot. The posting frames it as a front desk — a delegation of provider authority.
- Don't lead with Alibaba. Marketplace scale is the wrong signal for a healthcare provider-experience role.
Competitive landscape
Healthcare design specialist (EHR/clinical workflow background): Deeper provider workflow familiarity, but the CRM role is scheduling and prospect management, not clinical care. The design problem is delegation boundaries. Thermo Fisher is healthcare-adjacent and structurally identical. Weak on AI-native production proof. Confidence: moderate.
AI-native design leader (Anthropic/startup pedigree): Strong AI brand, unlikely to have regulated multi-party workflow experience. Will probably frame this as two-party human-in-the-loop and miss the three-party structure. Confidence: moderate.
Enterprise design director (Salesforce/Adobe pedigree): Scale credentials, weak on 0→1 speed and AI-native credibility. Won't have the Trust essay or Agentic Labs production proof. Confidence: moderate.
Your differentiator is the combination: regulated multi-party workflow proof, AI production experience, and a published framework for delegation boundaries. No single competing archetype brings all three.
Your vulnerability: no direct healthcare product design experience. If the hiring committee weights domain over design-problem match, the healthcare specialist wins.
The non-obvious thing
Headway published its authority model for consequential AI ten days ago: "Scribe drafts; it does not decide." That's the precedent the hiring committee already understands. Your job in the conversation is to name why the front desk requires a different pattern — and show you've designed that pattern before. Competing candidates will almost certainly reference Scribe as validation of the company's AI maturity. You should reference it as the precedent that reveals what the CRM role actually needs to solve: an authority model for situations where per-output review isn't available.
Warning signs
- Design inside product. Poses's VP Product role encompasses product design, product management, and product marketing. Ellen Dong leads design but the reporting line runs through product. At Staff IC level this may not limit craft autonomy, but design's strategic influence depends on Poses's advocacy.
- Director still establishing. Tiangco joined roughly nine months ago. He's building his team and his organizational authority simultaneously. A Staff designer reporting to a director who is still establishing can mean high autonomy or unclear mandate — hard to tell from outside.
- Revenue concentration risk. Provider-sourced patients account for ~70% of new patients. If the AI assistant underperforms, the revenue impact is immediate. That's an opportunity for portfolio-grade work and a risk if the org isn't ready to support the design investment the problem requires.
None of these are disqualifying — they're questions for the first conversation.
Questions for the first conversation
- Are you hiring for healthcare domain expertise or delegation-boundary design expertise? (Tests whether your vulnerability is disqualifying or irrelevant.)
- How much organizational authority does Tiangco have over cross-role design patterns — does the CRM designer set the AI delegation model, or inherit one? (Tests whether the scope-expandability score holds.)
- What's the relationship between the Scribe authority model and the front desk agent? Does the team see them as the same pattern or different problems? (Tests whether the hiring committee has already identified the gap you're naming.)
Cover letter
The front desk and scheduling assistant in your Provider CRM role faces a problem Scribe doesn't: the agent reaches the patient before the provider can review. Scribe solved consequential AI for clinical notes by requiring sign-off on every draft. A scheduling assistant that pauses for provider approval before every patient interaction defeats the purpose. The provider delegates, the agent acts, and the patient experiences the consequence with no control surface of their own.
As Product Design Director at BCG Digital Ventures, I led Thermo Fisher's mySupply platform from 0 to 1 — a regulated multi-party supply chain where six pharma partners delegated procurement decisions affecting downstream patient access to medication across nine sites and three continents. The delegating party couldn't review every action, and the affected party had no control surface. The system needed to earn trust through bounded autonomy rather than per-action approval. $20M+ in recovered margin, 100% partner adoption in twelve months. That work is published at junochen.com, alongside Allē — a dual-surface platform (30M members, 40K+ provider practices, Product Design Director at BCG DV) where provider-side decisions directly shaped the consumer experience, paralleling Headway's provider-patient relationship.
Most recently as Head of Product at TinyFish, I shipped an enterprise web agent platform from 0 to 1 within three months, working daily with agent traces, auditability, and governance challenges in production. I returned to design because I want to apply that technical depth to a specific, high-stakes vertical domain — and because the design problems at the delegation boundary are where I do my strongest work.
I'd welcome 20 minutes to discuss how delegation boundaries differ from review surfaces when the affected party is absent. More at junochen.com.
Resume emphasis
Craft depth 3, AI exposure 3.
- Lead: Thermo Fisher mySupply — regulated multi-party workflow, 0→1, $20M+ margin, 100% adoption. Frame as delegation-boundary design. Product Design Director at BCG DV.
- Follow: Agentic Labs + Trust essay — Carrier IQ (InsurTech, regulated, all five handoffs) establishes AI-native credibility. Trust essay's Delegate tier is the conceptual framework for the CRM role's authority model.
- Support: Allē — dual-surface (provider + consumer), 30M members, 3.2× redemption. Structural parallel to Headway's provider-patient split. Product Design Director at BCG DV.
- Keep but don't feature: Alibaba — $50B+ GMV, Head of Design and Research. Enterprise scale credibility. Secondary here.
- Trim: Red Cross — mission-driven but not the domain match. Present in resume, not featured.
- Header: TinyFish — "Most recently, Head of Product at TinyFish (enterprise web agent platform, Series A). Shipped agentic platform from 0 to 1 within 3 months. Agent traces, auditability, governance in production." One line. Past tense.
Post-outreach signals
Growth-stage platform archetype:
- Response within 3–5 business days = normal.
- Silence past 7 business days = likely in process with other candidates. One follow-up at day 8 referencing the Scribe launch as a new signal hook: "Your Scribe update last week clarifies the review model for clinical notes — curious how you're thinking about authority boundaries for the front desk agent, where per-output review isn't possible."
- Recruiter screen scheduled on first contact = fast-moving search. Prioritize prep immediately. Ellen Dong's hiring process includes an app critique and whiteboarding exercise — prepare the junochen.com walkthrough framed around delegation boundaries and the Thermo Fisher case.
- If portfolio walkthrough is requested: Lead with Thermo Fisher's multi-party authorization model, then Carrier IQ's five handoffs, then the Trust essay's Delegate tier. Frame every artifact through the three-party lens: who delegates, who acts, who bears the consequence.
- No reply after day 8 follow-up: Close the loop internally. Redirect energy. Monitor for a posting refresh.
- Scribe's audit claim: Headway says AI-assisted templates are three times more likely to pass an audit, which is company-reported and unverified — worth asking in conversation how they measure that and whether the CRM role's AI assistant will face similar compliance scrutiny.
- No AI assistant launched yet: The CRM posting describes an AI front desk and scheduling assistant, but no public product announcement for that feature has appeared through August 22 — probe whether this is a pre-launch build mandate or an iteration on something already in production.
- Dong's hiring format: Ellen Dong's documented interview process includes an app critique and whiteboarding exercise where she evaluates the depth of candidates' questions — prepare the junochen.com walkthrough specifically for that format before any screen is scheduled.
- Equity terms undisclosed: Headway's $100M Series D at $2.3B was over two years ago with no subsequent priced round reported — ask about current equity structure and liquidity preferences early in the process so the comp picture is clear before you invest in later rounds.