PHASE 1 — THE MOMENT
Altana is a supply-chain intelligence platform. They map global trade relationships using a knowledge graph built from customs records, shipping data, and corporate filings, then surface risk signals and compliance classifications for enterprise and government buyers. On July 21 (45 days ago), they acquired Cervo AI, a customs-brokerage automation company. That acquisition changed the product's design problem. Altana used to surface intelligence for humans to act on. Now its agents can draft customs entries, classify tariffs, propose compliance decisions — and the product needs a trust layer for when they should and shouldn't.
The Head of Product Design posting went live June 1 (95 days ago), but its ATS metadata shows a July 17 update — four days before the Cervo announcement. The revised language now names interruption, consent, task previews, confidence, provenance, and human-in-the-loop review. A prior req from March 2025 under a different Greenhouse ID either failed or was scoped differently. The posting was rewritten to match the post-acquisition mandate.
The real mandate: Design the interface between AI-inferred supply-chain data and high-stakes human decisions in trade compliance, customs filing, and government enforcement — now that agents can act, not just inform.
The posting says this role sits as peer to CEO, CPO, CTO, and Head of Engineering. Altana's leadership page lists no design executive. A former Head of Product Design appears to have departed; a Director of Product Design may still be in place (moderate confidence — LinkedIn lists him, Altana's site does not). A Chief People Officer joined June 15 (81 days ago). They are building organizational infrastructure at the same time they are running this search.
One gap to close early: the posting refers to "the CPO" without naming one, and no Chief Product Officer appears on the leadership page. Whether that title exists, is being hired, or is aspirational language changes the reporting line entirely. Ask in the first conversation.
Company score: 10. AI centrality 3 — knowledge graph, inference engine, agentic workflows; AI is the product. Stage/equity 2 — Series C at $1B in July 2024, $322M raised; meaningful pre-IPO equity but the round is 26 months old and current valuation is unconfirmed. Design influence 2 — posting promises C-suite parity; nothing on the leadership page validates it yet. Trajectory 3 — Cervo acquisition, $7.7M DHS/CBP contract obligated July 1, Unusual Machines defense partnership announced September 3.
Role score: 14. Comp 2 ($240K–$300K advertised plus equity; equity value depends on unconfirmed current valuation). Scope 3 (function-building, C-suite peer). Craft 3 (player-coach). AI exposure 3 (trust, supervision, human-agent handoffs are the explicit design territory). Portfolio value 3 (supply-chain AI for government and enterprise — differentiated, publishable).
A 95-day search with a prior req that either failed or was scoped differently, for a role this specific, means the candidate pool is thin on people who understand both enterprise trust architecture and agentic system design. The Cervo acquisition raised the bar. You are in a small field.
PHASE 2 — PORTFOLIO MAPPING
Intelligence layer check: yes. Altana's core product surfaces inferred intelligence for human decision-making. Their product documentation distinguishes observed information (bills of lading, customs filings) from inferred information (resolved identity, upstream suppliers, sourcing shares) and documents confidence indicators with plain-language explanations of inputs. The entry point is the moment between what AI surfaces and what a human decides to trust and act on. The Trust essay's five handoffs are the through-line.
Lead with the delegation envelope. The posting's vocabulary — interruption, consent, task previews, human-in-the-loop review — describes when the system can claim the user's attention and what it must show before acting. In a customs-brokerage workflow where agents prepare entries, these are delegation problems: which actions need approval, from whom, under what conditions, with what evidence.
-
Delegation envelope → Altana's agent permission model. The confidence/authority separation maps directly. An agent can be highly confident in a tariff classification but still lack authority to file it. Altana's docs describe exactly this: permissions governing tools, readable/writable data, and actions requiring approval. The delegation envelope — separating what the system knows from what it's allowed to do — is what this product needs. Prepare a visual walkthrough for the first conversation.
-
CarrierIQ's provenance layer → Altana's confidence indicators. Production proof that this thinking ships. Both trace how an inference was derived and present the evidence chain to a human reviewer. CarrierIQ shows the data sources behind a quote confidence score. Altana's docs describe the same architecture for supply-chain inferences.
-
CarrierIQ's structured intake → Altana's agent SOPs. Altana documents configurable standard operating procedures for agents: steps, required evidence, when to request input, when to stop. CarrierIQ's structured intake is a production implementation of Intent-Setting — specifying what the system should do before it does it.
-
CarrierIQ's approval gate → Altana's propose-for-acceptance default. Altana's customs brokerage surface says the system proposes actions for acceptance, editing, or rejection — "Altana never submits on its own." CarrierIQ's approval gate is the same pattern: Decision Gate handoff, built and published.
Support with Alibaba and Thermo Fisher. Alibaba proves enterprise-scale trust design in a B2B platform with complex multi-stakeholder workflows ($50B+ GMV, +20% transactions, +2.2pt NPS). Thermo Fisher proves 0→1 in a regulated supply chain with high-stakes compliance decisions ($20M+ margin, 6 pharma partners, 100% adoption). The domain parallel to trade compliance is the strongest in your portfolio.
TinyFish framing for this audience: Most recently as Head of Product at TinyFish, you shipped an enterprise agent platform 0→1 in three months — agent traces, auditability, governance in production. This is the technical credibility that makes the delegation envelope and CarrierIQ more than conceptual. Past tense. Position as deliberate AI immersion that grounds the design work.
The objection: "Her most recent title is Head of Product, not design." This is a screening risk. It can kill candidacy before a human reads the portfolio. The answer: the Trust essay, CarrierIQ, Brand Pulse, and Retail Velocity are published design work at junochen.com, produced during and after TinyFish. The Agentic Labs apps are live production design artifacts. The product title gave you the technical depth to design for agentic systems with practitioner authority — which is what this posting asks for.
PHASE 3 — THE OUTREACH PACKAGE
Buyer-adjacent contact: Evan Smith, co-founder and CEO. High confidence he is involved — the posting names the CEO as a peer to this role.
Hook source: Smith's July 8 article in SCMR (58 days ago) argues that supply-chain risk products fail when they present AI-derived visibility without a way to verify it through trusted data.
First Contact Message
Evan — your July piece in SCMR named a failure mode I've been designing against: AI-derived visibility without a way to verify it through trusted data. That gap between what an AI surfaces and what a human can trust enough to act on is the design problem I've focused my recent work around.
I published a framework for this — five handoffs between AI systems and human operators — along with production design artifacts that implement them: provenance layers that trace how an inference was derived, confidence indicators that separate observed from inferred, approval gates that let operators review before the system acts. The Cervo acquisition makes these patterns urgent across Altana's product surface.
Before this work, I led design for Alibaba.com's B2B platform — enterprise trust architecture at $50B+ GMV — and shipped a pharma supply-chain platform at Thermo Fisher from 0→1 with 100% partner adoption as Product Design Director at BCG Digital Ventures. Most recently I shipped an enterprise agent platform as Head of Product at TinyFish, working through the governance and auditability problems your customs-brokerage workflow will require.
I'd welcome 20 minutes to discuss how I'd approach the trust layer for Altana's agentic workflows.
Resume Framing Note
Lead the summary with "trust architecture for AI-native products." Surface Alibaba metrics first ($50B+ GMV, +20% transactions) — this buyer is building an enterprise platform and needs proof you've operated at that scale. Follow immediately with Thermo Fisher ($20M+ margin, regulated supply chain, 100% adoption) — the domain parallel to trade compliance is direct. TinyFish line: "Head of Product, TinyFish — shipped enterprise AI agent platform 0→1 in 3 months; agent traces, auditability, and governance in production." Past tense, one line, positioned as technical depth that informs the design work. Subordinate Equinox+ and Allē — consumer behavior design is not what this buyer is solving for. Avoid the phrase "design system." Do not use "design-led" — Altana isn't design-led yet. Claiming it as a value signals you haven't read the room.
Cover Letter Hook
Your Cervo acquisition turned a supply-chain intelligence platform into an agentic one, and the posting's emphasis on consent, task previews, and human-in-the-loop review tells me you know the trust layer is now the core design problem. I've been designing for exactly this — a published framework and production design artifacts for the five handoffs between AI systems and human operators, from intent-setting through decision gates.
PHASE 4 — WINDOW SUMMARY
| Action | Deadline | What degrades without it |
|---|---|---|
| Verify posting is live on Greenhouse | Before any outreach | 95-day posting with a prior req that either failed or was scoped differently; the funnel may already have finalists — confirm the req is active before investing effort |
| Send first contact to Evan Smith | September 9, 2026 | The Unusual Machines partnership announced September 3 gives a 7-day freshness window for referencing Altana's expanding defense footprint as a secondary hook if needed |
| Prepare delegation-envelope visual walkthrough | September 12, 2026 | Altana's agent permission model maps to the confidence/authority separation; a prepared walkthrough differentiates from candidates who can only discuss trust abstractly |
September 12, 2026.
- Altana's "fully agentic" tension: The Cervo acquisition announcement describes the combined product as "fully agentic" while the operating documentation describes propose-for-acceptance defaults and "Altana never submits on its own" — ask which version reflects the current deployment model.
- The missing CPO question: Altana's leadership page lists no Chief Product Officer despite the posting referencing "the CPO" as a peer — whether that role is filled, being hired, or aspirational determines the actual reporting line and design authority ceiling.
- Fable 5.1's progress-without-reasoning pattern: Anthropic's latest model emits user-readable progress messages while keeping underlying reasoning hidden, which maps to Altana's need for task previews and in-progress visibility without overwhelming customs brokers with inference traces.
- Enterprise safeguards as reviewer burden: Anthropic's Enterprise Frontier Safeguards shift monitoring to customer-designated reviewers without quantifying the operational workload — relevant to how Altana designs its own human-in-the-loop review surfaces for compliance teams.