Fourteen Companies Worth an Alert, Zero Worth an Application

Status Changes

Downgraded to Ignore-until-trigger:

Fieldguide — Staff Product Designer hit day 43 on August 22. No repost, no scope change.
Hightouch — Staff Product Designer, AI Creative Tools original route dates to June 26. The August update didn't reset the public clock.
Vanta — Head of Design hit day 47; Staff Design Systems at day 57. Both stale. A Senior Systems Designer, EPD appeared August 21 but doesn't qualify as a product-design leadership search. Specific posting URLs for the decayed roles are no longer directly linkable; the public board reflects current inventory only.
Amplitude — Head of Product Design hit day 46. No refresh. The public board reflects current inventory only.

Retired from monitoring:

Brex — Retired in the prior cycle. The Staff Product Designer, AI posting remains live and Brex-branded, but the pre-IPO equity premise disappeared with the Capital One acquisition. If the AI design mandate interests you on craft alone, treat it as a Capital One subsidiary role.
Freshworks — The CPTO appointment on July 28 didn't name design in the reporting scope. The qualifying roles — Senior Director, Product Design and Staff Product Designer, AI Agent Studio — are both India-based. Geographic hard filter.

No expiring signals this cycle

All entries below are Ignore-until-trigger.

How to Read This List

You cannot apply to any of these companies today for a qualifying role. That's the point. When a trigger fires — a fresh requisition, a design leadership search, a repost with changed scope — you already know the angle, the portfolio lead, and the priority. The account research is done. You go straight to outreach.

Entries within each cluster are ordered by likelihood of the trigger firing in the next 90 days. Companies that already posted a qualifying role and let it decay are closer to re-opening than companies that haven't posted one yet. In the funded-but-no-design-seat cluster, recency of funding is the proxy.

Stale Posting, Live Thesis

Five companies where the posting decayed past threshold but the company still clears the rubric. Each already showed intent to hire for a qualifying role. The trigger is straightforward: a fresh requisition ID, a repost with changed scope, or a new qualifying mandate.

1. Harvey. Legal AI platform for research, drafting, and professional legal workflows used by elite law firms.

Stage / Size / Capital: Series D ($100M, December 2024), ~300 employees, Sequoia-backed.
Offices: San Francisco, New York.
Value chain: Downstream application layer. Lawyers interact directly with AI-generated legal research and drafts. Design owns the surface where model output meets professional judgment — the moment a lawyer decides whether to rely on what the AI produced for a client matter.
Best at: Practice-area-specific depth for AmLaw 100 firms. The legal-domain training goes deeper than generic document AI.
Signals: Staff Product Designer routes from September and December 2025 are deeply stale (specific posting URLs no longer directly linkable; public board reflects current inventory). A Product Designer, Design Systems posted August 10–11 across multiple locations — one role with location variants, not multiple seats. Below Staff level, but it signals active design-system investment.
Tier: AI-native.
Rubric scores: Company 11/12 — AI centrality 3, Stage/equity 3, Design influence 2, Trajectory 3. Role: unscoreable from current postings.
Portfolio match: Harvey's core design problem is the review layer between AI-generated legal research and professional reliance — how much of the AI's process does the lawyer need to see before signing off? The five-handoff framework from the Trust essay decomposes this: the lawyer watches the AI's research process, verifies citations against source material, delegates routine research while retaining judgment on novel questions, reviews the draft before it reaches a client. The forward-looking artifact leads here because Harvey's evaluators will want someone who has broken "human reviews AI output" into its component design surfaces rather than treating it as a single interaction. Red Cross provides the production proof for the system-unification challenge the design-systems posting reveals: six separate systems consolidated into one national deployment, where the design system served caseworkers making disbursement decisions ($847K) across fundamentally different workflows. Harvey's practice-area variation is the same structural problem.
Verdict + action: Ignore-until-trigger. Trigger: Fresh Staff/Senior Staff product-design route, or two distinct IC product-design mandates posted within two weeks.

2. Fieldguide. AI platform for audit, tax, and advisory firms automating workpaper preparation, evidence gathering, and review workflows.

Stage / Size / Capital: Series C (February 2026), ~150 employees, Goldman Sachs-led.
Offices: San Francisco. US-remote for the Staff role.
Value chain: Midstream workflow layer between accounting firms, their clients' source systems, and the audit opinion. The critical surface is where CPAs decide whether AI-gathered evidence is sufficient to sign off.
Best at: Embedding AI into the procedural steps of professional audit — workpaper-level automation with firm-specific methodology.
Signals: Staff Product Designer posted July 10, now at day 43. No repost or scope change.
Tier: AI-native.
Rubric scores: Company 10/12 — AI centrality 3, Stage/equity 2, Design influence 2, Trajectory 3. Role 13/15 (scored from decayed posting) — Total comp 2, Scope 2, Craft depth 3, AI exposure 3, Portfolio value 3. This role cleared the Act threshold of 12+ before it decayed. Watch closely for a repost.
Portfolio match: The CPA looking at AI-assembled audit evidence and deciding whether it meets the standard for sign-off is an inference-aware design problem: how much of the AI's reasoning, confidence, and source-document chain does the auditor need to see before they'll sign? The forward-looking artifact for inference-aware design — how cost, latency, reasoning depth, and uncertainty shape product experiences — maps directly to Fieldguide's core surface, where the auditor's trust calibration depends on visible reasoning. Thermo Fisher mySupply is the production proof: six pharma partners had to trust a new system enough to route $20M+ in margin through it, and adoption had to be 100% because partial adoption in a supply chain breaks the chain. Fieldguide's CPAs face the same binary — partial trust in the audit evidence means the audit doesn't close. Lead with inference-aware design thinking, support with adoption in regulated professional context.
Verdict + action: Ignore-until-trigger. Trigger: New Staff+ requisition ID, repost with materially changed scope, or Director+ design search.

3. Hightouch. Customer data activation platform with AI-powered audience building, journey orchestration, and creative tools for marketing teams.

Stage / Size / Capital: Series D, $150M at $2.75B (April 2026), ~300 employees.
Offices: San Francisco. US-remote for the Staff role.
Value chain: Midstream activation layer between the data warehouse and marketing execution surfaces. Design shapes how marketers build audiences, configure journeys, and evaluate AI-generated creative.
Best at: Composable architecture that treats the data warehouse as the source of truth for marketing, avoiding yet another data silo.
Signals: Staff Product Designer, AI Creative Tools posted June 26, updated in August without resetting the clock. Series D funding is outside the 22-day window.
Tier: Growth-stage platform.
Rubric scores: Company 10/12 — AI centrality 2, Stage/equity 3, Design influence 2, Trajectory 3. Role 12/15 (scored from decayed posting) — Total comp 2, Scope 3, Craft depth 3, AI exposure 2, Portfolio value 2. Like Fieldguide, this role cleared the Act threshold before it decayed.
Portfolio match: The AI Creative Tools surface is where marketers evaluate whether AI-generated content represents their brand accurately enough to deploy. Agentic Labs Brand Pulse is the match — both products ask the same question from different directions: does the marketer trust the AI's read on their brand enough to act? Brand Pulse monitors sentiment; Hightouch generates creative. The gap between "AI generated this" and "I'd put my brand's name on this" is the design problem in both. Add Allē's dual-surface redesign (30M members, consumer + provider) for multi-stakeholder proof.
Verdict + action: Ignore-until-trigger. Trigger: Fresh Staff+ requisition ID, scope-changing repost, or Director+ product-design role.

4. Vanta. Continuous security compliance platform automating evidence collection, monitoring, and audit readiness across SOC 2, ISO 27001, HIPAA, and other frameworks.

Stage / Size / Capital: Series C ($150M, 2023), ~500 employees, Sequoia-backed.
Offices: San Francisco. US-remote for design roles.
Value chain: Midstream compliance infrastructure between a company's cloud environment, its security controls, and the auditor's assessment.
Best at: Making compliance continuous rather than periodic — the product watches infrastructure in real time and surfaces gaps before the auditor arrives.
Signals: Head of Design posted July 6, now day 47. Staff Design Systems posted June 26, also stale. Senior Systems Designer, EPD appeared August 21 but doesn't qualify. Prior coverage noted this search as a backfill. Specific posting URLs for the decayed roles are no longer directly linkable; the public board reflects current inventory.
Tier: Growth-stage platform.
Rubric scores: Company 9/12 — AI centrality 2, Stage/equity 2, Design influence 3 (Head of Design search signals intent), Trajectory 2. Role 11/15 (scored from decayed Head of Design posting) — Total comp 2, Scope 3, Craft depth 2, AI exposure 2, Portfolio value 2.
Portfolio match: Vanta's compliance monitoring is a supervision problem — hundreds of controls watched continuously, and the security team needs to distinguish which alerts require human investigation from which resolved automatically. Carrier IQ is the proof: an InsurTech platform where agents process quotes and the human reviewer needs provenance — why did the system flag this, what data drove the decision, is the audit trail complete? Vanta's compliance surface asks the same questions about security controls. Lead with the provenance layer for automated decisions in a regulated context.
Verdict + action: Ignore-until-trigger. Trigger: New Head/VP/Director requisition ID, repost explicitly identified as a renewed search, or new Staff product-design mandate.

5. Amplitude. Product analytics platform with Amplitude Wave, an AI agent that analyzes customer feedback and product data to surface insights.

Stage / Size / Capital: Public (NYSE: AMPL), ~700 employees.
Offices: San Francisco. North America remote for the Head role.
Value chain: Downstream analytics layer where product teams make decisions based on behavioral data. Wave adds an agentic layer that interprets data and recommends actions.
Best at: Behavioral cohort analysis at scale — the product defined event-based product analytics as a category.
Signals: Head of Product Design posted July 7, now day 46. Wave launched June 10. The combination — new AI product surface plus design leadership search — would have been a strong Act signal six weeks ago. Specific posting URL for the decayed role is no longer directly linkable; the public board reflects current inventory.
Tier: Enterprise platform.
Rubric scores: Company 8/12 — AI centrality 2, Stage/equity 1 (public, limited equity upside), Design influence 3 (Head of Product Design search), Trajectory 2. Role 10/15 (scored from decayed posting) — Total comp 2, Scope 2, Craft depth 2, AI exposure 2, Portfolio value 2.
Portfolio match: Wave's specific problem: a product manager looks at an AI-generated insight about user behavior and decides whether to change their roadmap based on it. The stakes are organizational — a bad insight acted on wastes a quarter. Brand Pulse is the match: real-time AI-surfaced intelligence where the user decides whether the signal is strong enough to act on. Both surfaces require calibrating how much context the AI needs to show for the human to trust the recommendation. Lead with the confidence-calibration interaction model from production.
Verdict + action: Ignore-until-trigger. Trigger: Fresh Head/VP/Director requisition ID, or Staff+ posting tied to Wave or the AI product organization.

Funded, No Design Seat

Five companies that raised significant capital recently, build AI-native products, and have no qualifying design role posted. The money is in, the product is shipping, but the design organization hasn't scaled to match. The prior dormant inventory distinguished definition windows from application windows. Every company here is pre-definition — when the design leadership search appears, the mandate will still be shapeable. Ordered by how soon the trigger is likely to fire, based on funding recency and hiring velocity.

6. Rillet. AI-native ERP automating journal entries, close, reconciliation, reporting, and multi-entity accounting.

Stage / Size / Capital: Series C, $100M at $1B (August 17, 2026), 51–200 employees, total >$200M. Sequoia, a16z, ICONIQ.
Offices: San Francisco, New York, Barcelona.
Value chain: Midstream financial-data platform between source systems, the general ledger, accountants, and financial reporting. Accountants review, approve, and close on the product's primary surface.
Best at: Treating accounting as an AI-native problem from the ground up rather than bolting AI onto legacy ERP.
Signals: Product Designer ($150K–$230K + equity) and Design Engineer posted in SF. Neither is Staff+. VP of Product and Head of Product are publicly identified; no Head/VP/Director of Design found. Series C closed five days ago — design leadership search is likely 2–4 months out.
Tier: AI-native.
Rubric scores: Company 11/12 — AI centrality 3, Stage/equity 3, Design influence 2, Trajectory 3. Role: pending.
Portfolio match: Rillet's close process is sequential and deadline-driven: journal entries flow to reconciliation, reconciliation to review, review to sign-off, all under a hard month-end deadline across multiple entities. The AI's accounting pipeline literally becomes the user's workflow, and the accountant needs to see where each entity stands in that pipeline, which entries the AI generated vs. which require manual creation, and where the process is blocked. The forward-looking artifact for agent infrastructure as a user-experience problem maps directly to this close surface. Thermo Fisher mySupply is the production proof: a sequential, deadline-driven supply chain where six pharma partners moved $20M+ in margin through a new system, and the design had to make each handoff trustworthy enough that the next participant would accept the output without re-verifying from scratch. Both are chain-of-custody problems where trust at each step determines whether the next participant accepts the output or starts over.
Verdict + action: Ignore-until-trigger. Trigger: Staff+ product-design posting, or a publicly identified design leader followed by a qualifying mandate.

7. HappyRobot. AI agents for logistics and industrial operations — voice, chat, and back-office workflows serving 150+ enterprises.

Stage / Size / Capital: Series C, $150M at $1.2B (August 4, 2026), fivefold growth since prior round. a16z, YC.
Offices: San Francisco.
Value chain: Midstream operations layer between shippers, carriers, dispatchers, and back-office systems. Operations staff supervise AI agents handling real-time logistics communication on the product's primary surfaces.
Best at: Deploying AI agents into the specific communication patterns of freight logistics — load tendering, tracking updates, carrier negotiations. Domain-specific agents, not generic chatbots.
Signals: Two Product Design Engineer openings in SF, plus a Brand Designer. No Staff/Senior Staff Product Designer or Director+ role. The design-engineer routes are at day 43–45, outside the IC-cluster window. Series C closed 18 days ago — inside the funding signal window but decaying.
Tier: AI-native.
Rubric scores: Company 11/12 — AI centrality 3, Stage/equity 3, Design influence 2, Trajectory 3. Role: pending.
Portfolio match: The dispatcher supervising AI agents handling live freight operations needs to know which calls the agent is handling, which are going wrong, and when to intervene — all in real time across 150+ enterprise deployments. Carrier IQ is the direct domain match (InsurTech agent automation with auditability), and the Trust essay's intervention handoff addresses the specific question: when does the human take over from the agent mid-task? Cummins/ZED Connect adds logistics-domain credibility — dual-surface fleet management (driver-facing + fleet-operations), IoT data, predictive maintenance. The combination of agent-supervision design framework and logistics/fleet-operations domain experience barely exists in the candidate pool.
Verdict + action: Ignore-until-trigger. Trigger: Staff/Senior Staff Product Designer or Director+ product-design posting, or a new two-role product-design cluster posted within two weeks.

8. Corgi. Full-stack commercial insurer that underwrites, issues policies, handles claims, and operates embedded-insurance products across trucking, small business, and sports.

Stage / Size / Capital: Series B1, $106M at $2.6B (May 2026, three weeks after a $160M Series B at $1.3B). ~250 employees. YC-backed.
Offices: San Francisco.
Value chain: Vertically integrated across distribution, underwriting, policy administration, and claims. Design has access to both customer-facing quotation surfaces and internal regulated decision surfaces — unusually broad scope.
Best at: Speed of vertical integration. Doubling valuation in three weeks suggests the underwriting model works, and expansion into trucking, small business, and sports means the product surface is multiplying fast.
Signals: 59 open roles including marketing-oriented design positions, but no product-design role at qualifying level. Aggressive hiring velocity suggests a design leadership search is a matter of timing.
Tier: Healthcare/Regulated. The insurance regulatory context — underwriting compliance, claims adjudication rules, state-by-state filing requirements — shapes how this company will evaluate a design leader more than the AI-native product architecture does. The AI layer is real, but the evaluation will center on whether you understand regulated decision surfaces.
Rubric scores: Company 10/12 — AI centrality 2, Stage/equity 3, Design influence 2, Trajectory 3. Role: pending.
Portfolio match: Corgi's vertical integration means the designer owns both the policyholder-facing quotation surface and the internal underwriting decision surface — two audiences, two trust calibrations, one system. Allē is the match: dual-surface redesign (30M consumer members + provider-facing tools) where the design system served both sides of a transaction with different trust requirements and different success metrics (3.2× redemption on consumer side, $42 CAC on provider side). Corgi's policyholder and underwriter are the same structural pair. Lead with dual-surface regulated-transaction proof.
Verdict + action: Ignore-until-trigger. Trigger: Staff+ or Director+ product-design role covering underwriting, claims, policy administration, or a shared design system.

9. Assured. Agentic platform for P&C insurance claims — intake, first contact, fraud detection, adjudication, and claimant communication across tens of millions of claims.

Stage / Size / Capital: Equity round at reported ~$1B valuation (March 2025; ICONIQ disputed the reported terms without correcting them). 201–500 employees. ICONIQ, Kleiner Perkins.
Offices: Palo Alto. Described as fully remote.
Value chain: Midstream claims-orchestration layer between insurers, adjusters, claimants, and payment decisions. Every surface where a human reviews, overrides, or approves an AI-driven claims decision falls within the design scope.
Best at: End-to-end claims automation with deterministic controls, explainability, and human-in-the-loop intervention across the full claims lifecycle.
Signals: Zero open positions at cutoff. The company describes real-time monitoring and human-in-the-loop controls across its product, which signals design maturity even without a posted role.
Tier: AI-native.
Rubric scores: Company 10/12 — AI centrality 3, Stage/equity 2 (valuation disputed, round age >12 months), Design influence 2, Trajectory 3. Role: pending.
Portfolio match: Assured's claims adjudication surface runs the Trust essay's full five-handoff framework in production: the adjuster watches the agent process claims, reviews flagged decisions, approves automated resolution paths, catches fraud signals, and audits completed claims. Carrier IQ is the InsurTech domain match, but the specific proof is the five-handoff framework itself — Juno decomposed "human-in-the-loop" into five distinct design problems, each with different interaction requirements. Most candidates can talk about human-in-the-loop as a concept. Juno has the published framework that breaks it into component design surfaces.
Verdict + action: Ignore-until-trigger. Trigger: Any Bay/US-remote Staff+ product-design or Director+ design role tied to claims adjudication, claimant communication, or human review.

10. Infinitus Systems. Healthcare communication agents for benefit verification, prior authorization, and prescription follow-up — AI making phone calls to payers on behalf of providers.

Stage / Size / Capital: Series C, $51.5M (October 2024), total $102.9M. 51–200 employees. a16z, Coatue, Kleiner Perkins.
Offices: San Francisco (hybrid, 3 days/week).
Value chain: Midstream coordination layer between providers, payers, PBMs, and patients. Healthcare staff use the product to configure, monitor, and intervene in AI agent phone calls.
Best at: Deploying voice agents into the procedural requirements of payer-provider communication — hold times, IVR navigation, verification scripts — where the agent must follow healthcare-specific protocols.
Signals: No open positions at cutoff. Series C press release emphasized "AI guardrails" as a differentiator, signaling that safety and oversight design is central to the product thesis.
Tier: Healthcare/Regulated.
Rubric scores: Company 10/12 — AI centrality 3, Stage/equity 2, Design influence 2, Trajectory 3. Role: pending.
Portfolio match: The design problem is the intervention moment during a live agent phone call — the healthcare worker monitoring a benefit verification call needs to know when the agent is stuck, when the payer is asking something the agent can't handle, and when to take over. Red Cross is the match: high-stakes coordination where caseworkers made disbursement decisions during disaster response and the system had to surface the right information at the right moment for a time-pressured human decision ($847K disbursed, 6 systems consolidated into 1, national deployment). Both are real-time, high-consequence coordination problems where the human's decision window is narrow. Lead with the real-time decision surface for high-stakes human intervention.
Verdict + action: Ignore-until-trigger. Trigger: Staff+ or Director+ product-design role for agent-building, patient communications, safety guardrails, or human escalation.

Domain-Fit Sleepers

Four companies whose product problems map to your portfolio with unusual precision, but where the current hiring state doesn't qualify — the posted role is below level, geographically mismatched, or absent entirely. These are the longest-horizon entries on this list. The value is the strength of the match when the trigger fires.

11. LangChain. Open-source agent frameworks plus LangSmith, a platform for observing, evaluating, debugging, and deploying production AI agents.

Stage / Size / Capital: Series B, $125M at $1.25B (October 2025). 201–500 employees. IVP, Sequoia, Benchmark.
Offices: San Francisco, New York, Cambridge, Amsterdam, London.
Value chain: Midstream agent-engineering and observability layer between foundation models and the teams building production applications. Design shapes how developers and operators understand what their agents are doing and why.
Best at: Owning the developer workflow for agent construction — from framework to observability to deployment — which gives the design team influence over how the ecosystem thinks about agent debugging and evaluation.
Signals: Product Designer posted in SF/NYC, 3+ years experience, no Staff level established.
Tier: AI-native.
Rubric scores: Company 10/12 — AI centrality 3, Stage/equity 3, Design influence 2, Trajectory 2. Role: current posting doesn't qualify (below level).
Portfolio match: LangSmith's core surface is the agent trace — a developer or operator looking at a sequence of agent decisions and evaluating whether the agent behaved correctly. The agent's plumbing — its decision chain, tool calls, retrieval steps, failure points — becomes the product's primary interface. The forward-looking artifact for agent infrastructure as a user-experience problem maps directly to LangSmith's trace and evaluation surfaces. Carrier IQ's provenance-tracking is the production proof: tracing each agent decision back to its inputs, understanding why the agent chose this path, deciding whether to change the agent's behavior. TinyFish production experience amplifies the match — Juno worked daily with agent traces, auditability, and attribution in enterprise deployments, and has the practitioner's understanding of which information in a trace actually changes the operator's decision.
Verdict + action: Ignore-until-trigger. Trigger: Staff+ product-design role with authority over agent evaluation, observability, or deployment, or a Director+ mandate for LangSmith.

12. Pylon. Agentic B2B support platform spanning Slack Connect, Teams, email, in-app chat, and customer-success workflows — the system of record for B2B customer communication.

Stage / Size / Capital: Series B, $31M (August 2025), total $51M. ~90–120 employees. a16z, General Catalyst, YC.
Offices: San Francisco.
Value chain: Downstream system of record where support teams and AI agents investigate and resolve account-level customer issues across multiple channels simultaneously.
Best at: Treating B2B support as a multi-channel, account-level problem rather than a ticket queue — the platform unifies Slack, Teams, email, and chat into a single customer record.
Signals: Visual Product Designer posted in SF with no stated experience threshold. The role emphasizes design-system construction, broad user flows, and direct founder collaboration without product managers. Series B is 12 months old.
Tier: Growth-stage platform.
Rubric scores: Company 10/12 — AI centrality 2, Stage/equity 2, Design influence 3 (founder-direct, no PM layer), Trajectory 3. Role: current posting doesn't qualify (no level established).
Portfolio match: Pylon's "massive surface area" across channels is an information-architecture problem before it's a visual-design problem — how does a support agent see the full history of an account's communication across Slack, email, and chat, and how does the AI agent's work interleave with the human's across those channels? Alibaba's platform redesign is the match: $50B+ GMV, cross-functional sprints across homepage, search, and PDP, where the design challenge was making a massive multi-surface platform coherent for users with fundamentally different workflows. The specific proof is the +20% transaction lift from structural redesign. Lead with multi-surface coherence at enterprise scale and the measured business impact.
Verdict + action: Ignore-until-trigger. Trigger: Staff/Senior Staff leveling of the current mandate, a new Staff+ requisition, or a Director/Head role.

13. SmarterDx. Clinical AI for revenue integrity — reviews medical records to surface findings affecting reimbursement, denials, documentation quality, and care measures across 85+ health systems.

Stage / Size / Capital: Series B ($50M, March 2025) plus strategic investment from New Mountain Capital (April 2025, terms undisclosed). 501–1,000 employees.
Offices: New York (HQ). US-remote roles predominant. Boundary note: NYC-headquartered, but the company's operating model is US-remote across its open roles, which clears the geographic filter for Bay Area candidates. Included here because the domain fit justifies monitoring.
Value chain: Midstream clinical-to-financial decision layer between patient records, CDI/coding teams, claims, payers, and hospital reimbursement. Clinical reviewers evaluate AI-surfaced findings on the product's primary surface — findings that directly affect revenue.
Best at: Connecting clinical findings to specific reimbursement and denial outcomes — the AI links clinical accuracy to revenue impact, not just documentation gaps.
Signals: 17 open roles, no product-design posting at any level.
Tier: Healthcare/Regulated.
Rubric scores: Company 9/12 — AI centrality 3, Stage/equity 2, Design influence 1 (no visible design org), Trajectory 3. Role: pending.
Portfolio match: The clinical reviewer evaluating an AI-surfaced finding that will change a hospital's reimbursement needs to see the clinical evidence, understand the AI's reasoning, assess the revenue impact, and decide whether to accept, modify, or reject the finding. The stakes are clinical and financial simultaneously. Red Cross is the match — $847K disbursed through a system where caseworkers made high-consequence decisions under time pressure, and the design had to surface the right evidence at the decision moment without overwhelming the reviewer. Both are high-consequence review surfaces where the human's trust in the system's recommendation directly affects outcomes. The design influence score of 1 is a real concern — monitor for a design leader hire as the first signal of organizational intent.
Verdict + action: Ignore-until-trigger. Trigger: US-remote Staff+ or Director+ product-design role for clinical review, documentation integrity, denials, or human-in-the-loop revenue decisions.

14. Numeric. AI accounting platform unifying close management, reconciliation, reporting, technical-accounting research, and cash management.

Stage / Size / Capital: Series B, $51M (November 2025), total $89M. ~115 employees. IVP, Founders Fund.
Offices: San Francisco (HQ). Geographic mismatch: Current Product Designer, Experience role is New York City, in-person five days/week. No Bay or remote route despite SF headquarters.
Value chain: Midstream financial-data and workflow platform between source systems, the general ledger, accountants, and reporting. Overlaps with Rillet's space but emphasizes cash management and variance analysis rather than journal-entry automation. The design problem is investigative — the accountant traces a discrepancy back to its source — which differs from Rillet's sequential close workflow.
Best at: The investigative layer: cash matching, variance explanation, and technical-accounting research where the accountant follows a discrepancy to its origin.
Signals: Product Designer, Experience posted in NYC. Design Engineer also listed. No Bay-eligible route. No design leadership signal.
Tier: AI-native.
Rubric scores: Company 9/12 — AI centrality 3, Stage/equity 2, Design influence 2, Trajectory 2. Role: geographically disqualified.
Portfolio match: Numeric's variance-explanation surface is an inference-aware design problem: the accountant sees a discrepancy, the AI traces it back through source systems, and the accountant decides whether the explanation is sufficient to close the item. How much of the AI's reasoning chain — which source documents it consulted, what matching rules it applied, where its confidence drops — needs to be visible for the accountant to trust the resolution? The forward-looking artifact for inference-aware design addresses this directly. Carrier IQ's provenance-tracking is the production proof: tracing a decision back to its source documents and making the chain of evidence visible enough for the reviewer to trust it. TinyFish production experience adds the attribution and auditability work with agent traces that had to be auditable for enterprise customers. The combination of AI agent source-tracing and enterprise platform design (Alibaba) is what Numeric's variance-explanation product requires.
Verdict + action: Ignore-until-trigger. Trigger: Bay-area or US-remote Staff+ product-design role, or Director+ mandate spanning AI accounting and the financial-data platform.

Things to follow up on...

Rillet's missing design leader: The research identified VP of Product and Head of Product on Rillet's public LinkedIn page, but no Head/VP/Director of Design — worth watching whether the Series C triggers a design leadership hire before a Staff+ design posting appears.
Freshworks design reporting line: The CPTO appointment named product strategy, engineering, and AI in scope but didn't mention design, leaving it unresolved whether the US design org will eventually post qualifying Bay-eligible roles under the new structure.
Brex's post-acquisition design org: The Capital One closing filing acknowledged integration work while Brex publicly says it operates independently, but no public evidence establishes whether Brex designers have been absorbed into Capital One's design organization or remain a distinct team.
Assured's disputed valuation: Bloomberg reported an approximately $1B valuation for the March 2025 round, but ICONIQ called the reported information inaccurate without supplying corrected terms — a future funding announcement would resolve both the valuation question and the stage/equity rubric score.