The AI labs are already hiring for your industry

Government and life sciences lead the vertical hiring
Open roles tied to finance, life sciences & health, legal or government, by industry

Among the five labs, government and defense is the largest bucket: 51 roles across OpenAI (20), Anthropic (23) and Mistral AI (8). Life sciences and health follow with 35, finance with 27, legal with 5. The chart adds two comparisons: Cohere, and Google Cloud's industry roles.

The mix differs by lab. OpenAI spreads across all four, including a founding engineer for "AI-powered products and experiences for the legal industry". Anthropic has no legal-industry roles; its finance sales roles cover banking, insurance, asset management and hedge funds. Mistral AI is a public-sector story: 8 of its 9 vertical roles sell to or deploy in government and defense, among them "Critical and Sovereign Institutions" strategists in Paris and the UK. xAI's 4 finance roles all name X Money. DeepMind's four are "Gemini Data" expert leads.

Google Cloud does the selling for Alphabet. Searches of Google's careers site found at least 55 Google Cloud roles titled for these industries: 38 in government (among them "Forward Deployed Engineer, National Security, Google Public Sector"), 13 in finance, 4 in healthcare and life sciences. 28 sell, 18 deploy and 9 build or run government cloud, 4 of those for sovereign cloud. That is at least as many as Anthropic's 55.

Cohere bets on governments. All 13 of Cohere's vertical roles serve public sector and defense, from forward-deployed engineers for the UK public sector to defense delivery managers in Canada, the UK, France and Germany.

Who sells into the verticals, and who builds there
Open roles tied to finance, life sciences & health, legal or government, by what the role does

A count says a lab cares about an industry; the role type says how. We sorted every vertical role into sell, deploy, build or M&A (definitions in the Method section).

Anthropic sells. 31 of its 55 vertical roles are sales roles, 13 deploy and 10 build. Seven of the 10 build roles sit in one place, a Life Sciences research team, whose job text says it is "focused on making fundamental biological discoveries" and combines "cutting-edge AI with hands-on biological research". The one M&A role, Corporate Development Lead, Life Sciences, covers "acquihires, tuck-in acquisitions" in drug discovery, clinical development, lab automation and healthcare data.

OpenAI deploys and builds. 19 of its 45 vertical roles deploy and 12 build; 14 sell. The build roles read like a product company entering an industry. Backend engineers for "Codex for Finance" join "OpenAI for Financial Services", which the ad places in "OpenAI's Verticals organization". A product manager for financial services reports "to the GM of Financial Services". The Life Sciences team builds Rosalind Workbench around "GPT-Rosalind, our dedicated life sciences model", and its engineers connect scientists to "laboratory services".

DeepMind, xAI and Mistral put domain knowledge into training data. DeepMind's finance, legal and health expert leads sit on "Gemini Data"; the legal posting asks for experience "optimizing [LLMs] for legal use cases". xAI hires an "AI Tutor - Legal & Compliance". Mistral hires a "Finance Data Annotation Quality Analyst", and its AI for Science team builds training data for engineering simulation and chip design. That is vertical entry through the model rather than through a product or a sales team.

For scale, a vertical-native company. Harvey, which builds AI for law firms and legal departments, has 358 open roles, 106 of them in its go-to-market department alone. The five labs have 5 legal-industry roles between them. In legal, at least, the labs are still suppliers to the specialists.

Both large labs bought a biotech company this year, and Anthropic has an open role whose job is to buy more. Alphabet funds the same bet through Isomorphic Labs, whose AI-designed drugs WIRED reported in April are headed to human trials.

Finance and health launches went back and forth between the two labs all year; the life-sciences push grew through the year and came with acquisitions.

Careers boards show only roles open today, so they cannot show month-by-month growth: roles posted in spring and already filled are gone, and the Internet Archive refused our automated reads of its old copies. They do show when each open role was posted (on Ashby boards the date may reflect a re-publication). Greenhouse (Anthropic, xAI) gives a first-published date; Ashby (OpenAI, Mistral AI, Cohere) gives a publication date; we use neither board's "updated" date.

By that measure OpenAI's vertical roles are newer than its average role: 35 of 45 (78%) were posted in August–October 2026, against 66% of all its open roles. Anthropic's are as new as the rest (65% against 66%). 54 of Anthropic's 55 vertical roles, and 43 of OpenAI's 45, were posted in 2026.

One reply in the discussion of Mollick's point asked: "in short, embodiment?" The hiring data says yes for OpenAI, and yes in a different sense for xAI.

• OpenAI has 15 open roles on its Robotics team (motors, actuators, control systems) and 25 in Consumer Devices (firmware, on-device inference): 40 of 796, more than its finance and legal roles combined.
• xAI, now part of SpaceX, is mostly hiring for the physical plant: 170 of its 298 open roles name data centers, power generation, construction or its Memphis-area sites, from medium-voltage electricians to power-plant operators.
• Google DeepMind has 7 robotics roles, Mistral AI 3. Anthropic has none; its 30 physical roles are in data-center build-out and operations.

Read through the hiring data, Mollick's forecast already describes two labs, and they are not following one strategy.

Anthropic's openings point to distribution, with building mostly in biology, where it is also buying companies. OpenAI's point to products: a verticals organization, a life-sciences model, legal and finance product teams, and acquisitions that feed them. Alphabet splits the work between DeepMind's training data and Google Cloud's sales force; Mistral and Cohere sell sovereignty to governments.

For a company that builds on top of these models in finance, life sciences, legal or government, the useful signal is the "build" column. Where a lab hires product engineers and buys startups, it is not only a supplier any more.