AI product manager roles fall into seven families, and two of them dominate. We grouped all 116 live PM job titles at 18 AI companies in the AllthingsPM jobs catalog: 34 are generalist or product-area titles, 31 are agent titles, 16 are research, model or quality titles, 15 are safety, trust or security titles, 9 are platform or API titles, 7 are growth titles and 4 are forward deployed titles. Agent PM is now the single biggest specialty, and most of it comes from Sierra and Decagon.
AllthingsPM is an AI PM course and PM interview prep platform. Every title below links to its full job description in our jobs catalog, and every one comes with an AI mock interview built from that exact posting, so you can decode a title and practice for it in the same minute.
What are the main AI product manager roles in 2026?
Here is the grouping, with real example titles from the catalog. A title counts once, in the first family its words match, checked in this order: forward deployed, agents, safety, research and models, platform, growth. Anything left is a generalist or product-area title.
| Family | Titles (of 116) | Share | Real example titles | Who posts them most |
|---|---|---|---|---|
| Generalist or product area | 34 | 29% | Product Manager (Midjourney, Linear); Product Manager, Claude Science (Anthropic); Staff Product Manager, New Verticals (Harvey) | Harvey (6), Anthropic (5), Glean (4) |
| Agents | 31 | 27% | Product Manager, Agent Development (Sierra); Senior Agent Product Manager (Decagon); Product Manager, API Agents (OpenAI) | Sierra (16), Decagon (8) |
| Research, models and quality | 16 | 14% | Product Manager, Core Models (OpenAI); Product Lead, AI/ML (Evals) (Abridge); Product Manager, AI Quality (Glean) | Scale AI (5), Anthropic (4) |
| Safety, trust and security | 15 | 13% | Product Manager, Safeguards (Generalist) (Anthropic); Product Manager, Safety Measurement (OpenAI) | Anthropic (7), OpenAI (6) |
| Platform, API and infrastructure | 9 | 8% | Product Manager, API Infrastructure (OpenAI); Product Manager, API Platform (Glean); Product Manager, Observability (Vercel) | Glean (3), Scale AI (2) |
| Growth and monetization | 7 | 6% | Product Manager, Growth (Anthropic); Product Manager, AI Growth (Figma) | Anthropic (3), Glean (2) |
| Forward deployed | 4 | 3% | Forward Deployed Product Manager, Enterprise (Scale AI); Forward Deployed Product Manager (Abridge) | Scale AI (3) |
How AllthingsPM does this. Every row in that table is a live page in the AllthingsPM jobs catalog, grouped by company, with the full job description and a start button for a mock interview built from its text. You can filter to the family you want and prepare for the actual posting instead of a generic "AI PM" prompt.
How did we build this data set?
The source is the AllthingsPM jobs catalog: 116 product manager job descriptions from 18 AI companies, pulled from each company's public job board between 14 and 22 September 2026. Anthropic (19), Sierra (18), Scale AI (15), Glean (14), OpenAI (12) and Decagon (10) account for 88 of them. The rest come from Harvey, Abridge, Cohere, Linear, Lovable, Perplexity, Ramp, Midjourney, Vercel, Suno, Replit and Figma.
We counted postings, not unique jobs. The same job posted in two cities counts twice, which is how a job seeker meets the market. There are 96 distinct title strings across the 116 postings. We grouped by words in the title with a fixed keyword list, then read every row by hand to check the family made sense.
Seniority comes from the full job description (years asked for, scope, reporting line), not only from the title, because "Product Manager" at one lab can mean a senior hire at another.
How AllthingsPM does this. The same catalog feeds the JD mock interview: pick any of the 116 postings, or paste a job description from anywhere, and the mock asks questions drawn from that text. The data behind this study is the product you practice on.
Why is "Agent PM" the biggest specialist title?
Agent titles are 31 of 116 postings, second only to the generalist bucket. The count is concentrated: Sierra posts 16 "Product Manager, Agent Development" and related titles, and Decagon posts 8 "Senior Agent Product Manager" roles, many of them split by market or language (Spanish, Brazilian Portuguese, Italian, Arabic, French and German speaking versions, 8 in total).
This is a customer-facing job. Sierra's posting says: "As an Agent PM at Sierra, you will be responsible for partnering directly with our engineers and customers to build and ship AI agents that handle thousands of customer conversations a day." It asks for 5 to 7+ years in product development of highly technical products. Decagon's senior version asks for 6+ years "owning ambiguous, high-stakes outcomes and earning the trust of senior executives."
The other 7 agent titles sit on the platform side: Product Manager, API Agents at OpenAI, Product Manager (Agents) at Lovable, Product Manager, Agent Harness & Modelling at Cohere, Staff Product Manager, Agentic Platform at Scale AI, and Glean's Agent Security & Governance roles. Same word, a very different job: these PMs build the tools other teams use to make agents.
So "Agent PM" is two roles. At Sierra and Decagon it means owning a customer's live agent end to end. At OpenAI, Cohere or Lovable it means building agent infrastructure for developers or builders.
How AllthingsPM does this. The AllthingsPM course has a full chapter on agents and agentic architecture, including when multi-agent systems are worth their cost. For the customer-facing version, practice the real question bank prompt "How would you improve Sierra's AI agents to resolve more customer issues without escalation?", then run a mock on Sierra's Agent Development posting.
What do research, model and quality PMs do?
16 titles sit close to the model itself: Product Manager, Core Models at OpenAI; Product Management, Research and Research Product Manager, Labs at Anthropic; Product Lead, Foundational Models and Post-Training and Product Lead, AI/ML (Evals) at Abridge; Product Manager, AI Quality at Glean; and five Scale AI titles on code, multimodal, leaderboards, Gen AI and physical AI data.
OpenAI's Core Models posting describes a team that "helps shape how OpenAI's frontier models are built, measured, and launched," with a scope that includes "building data flywheels, evaluations and measurement systems." That is the common thread in this family: the PM defines what good means and proves the model got there.
Evaluation language is not limited to this family. 51 of the 116 job descriptions mention evals or evaluation somewhere in the body.
How AllthingsPM does this. The AllthingsPM course chapter on evals teaches how to define good and make the number defensible, and the lesson on working with researchers covers getting a model roadmap out of a research team. Then run a mock on Abridge's Evals lead posting or OpenAI's Core Models posting.
Are AI safety PM roles a real career track?
Yes, at the labs. 15 titles are about safety, trust or security, and 13 of them come from Anthropic (7) and OpenAI (6). Anthropic alone lists Safeguards roles for account integrity and abuse, cyber, rare harms and a generalist seat, plus Safe Access, Multi-Cloud Trust & Safety and Cybersecurity. OpenAI lists Multimodal Safety, Safety Measurement, Sensitive Deployments, Youth, Enterprise Identity and a Product Policy Lead for regulation.
Anthropic's Safeguards posting explains the scope: the team "builds protections for new AI features enabled by the research teams and protects new products and surfaces developed by our product teams." It is product work with a risk lens, not policy writing.
Outside the labs, safety shows up as enterprise security instead: Glean's Protect product and Scale AI's Senior AI Product Manager, Cybersecurity.
How AllthingsPM does this. Chapter 12 of the AllthingsPM course, Trust, safety, and agent security, covers guardrails, over-refusal budgets and red-teaming. Our deep dive on safeguards PMs at AI labs goes further, and you can practice on Anthropic's Safeguards (Generalist) posting.
What is a forward deployed product manager?
It is the smallest family, 4 titles, three at Scale AI and one at Abridge, but it is the newest-sounding title in the set. Scale AI's Enterprise posting calls the role "the accountable PM for product delivery and customer value," driving AI deployments from contract through production and translating "real customer workflows, friction, and priorities into high-signal product input for core platform teams."
Outside our catalog, industry writers describe the forward deployed PM as a PM who embeds with a customer to ship a working AI product in that customer's environment and feeds what they learn back into the core product. Note the overlap with Sierra's and Decagon's agent roles: both sit with the customer. The difference is the title, not the shape of the work.
How AllthingsPM does this. The AllthingsPM course chapter Ship it into somebody else's company is about enterprise brownfield deployment, the core skill of this family. Our guide to the forward deployed product manager role goes deeper, and you can run a mock on Scale AI's Enterprise FDPM posting.
What do platform and growth AI PM titles mean?
Platform titles (9) are the developer and infrastructure side: API Infrastructure at OpenAI, API Platform and Connectors at Glean, Integrations at Cohere, Observability at Vercel, Enterprise Core Platform and Staff Technical Product Manager at Scale AI, and Infrastructure at Harvey. These roles are closest to the "AI platform PM" archetype that role guides from Marily Nika and Arize describe.
Growth titles (7) are the classic growth PM job at AI companies: Product Manager, Growth and Multi-Cloud Growth at Anthropic, Growth at Glean, AI Growth at Figma, New Markets and Monetization at Anthropic, and Vendor Intelligence & Marketplace at Ramp.
How AllthingsPM does this. For growth roles, the AllthingsPM question bank has real prompts such as "Define the core growth funnel for Claude as a subscription product". Practice it, then run a mock on Anthropic's Growth PM posting.
Why do so many AI PM titles look generic?
34 titles fit no specialist family, and the reason is simple: at an AI company, every product is an AI product. Only 14 of the 116 titles contain the word "AI". Five are just "Product Manager". The job lives in the suffix: 97 of 116 titles carry a specialty after a comma, in parentheses or after a pipe (Claude Science, Mobile Apps, Learning, Vault, Tax).
That makes generic-looking titles the hardest to prepare for from the title alone. "Product Manager, Duet" at Decagon and "Product Manager, Ghostwriter" at Sierra only make sense once you read the description. The generalist bucket also holds most of the lead titles: Abridge's "Product Lead" roles and Glean's "Product Management Lead". One title here, Harvey's Head of Product Communications, is a communications role listed alongside PM jobs; we kept it so the count matches the catalog.
How AllthingsPM does this. Because each AllthingsPM mock is built from the job description text, not the title, a vague title like "Product Manager, Duet" still gets questions about the actual product. Paste the posting into JD resume review to see which of its requirements your resume already covers.
What level are AI product manager roles?
Most AI PM roles are for people who have already done the job. We read seniority from each full description.
59 roles read as mid-level PM, 34 as Senior, 10 as Staff, 8 as Lead and 1 as Director. Only 1 role, Scale AI's AI Product Manager (Coding/Multimodal), reads as entry level. The title often understates the level: Sierra's "Product Manager, Agent Development" asks for 5 to 7+ years and experience at or above Senior PM.
About half the postings, 59 of 116, include salary text, mostly US roles. For numbers, see our AI product manager salary study.
How AllthingsPM does this. If you are not yet at that level, Resume Job Match ranks live openings against your resume so you apply where you fit today, and the AllthingsPM question bank lets you practice the question types each level gets.
What are the limits of this study?
- It is a snapshot. 116 postings pulled from 14 to 22 September 2026. Titles churn weekly.
- It is AI-native companies only. Big tech AI PM roles (Google, Meta, Microsoft) use different naming and are not in this catalog.
- A few companies weigh a lot. Sierra and Decagon drive the agent count; Anthropic and OpenAI drive safety.
- Postings, not people. Duplicates across cities count separately, and first-match grouping puts Glean's "Agent Security & Governance" in agents, not safety.
Why AllthingsPM is the better choice for AI PM roles
Decoding a title is step one. Getting the job means knowing the skills the family expects, answering the questions that team asks, and showing a resume that matches the posting. AllthingsPM does all three against the same live data.
The 116 postings in this study are all on AllthingsPM, and each one has a mock interview built from its text, in text or voice, with follow-up questions and a score. The AI PM course was built from 604 real PM job postings, and it has dedicated chapters for the three biggest specialist families here: agents, evals and trust. The question bank holds 4,122 real questions from 260 companies, each with its own answer guide. Resume review against a JD and Resume Job Match close the loop.
Other resources have real strengths. Role guides from Marily Nika and Arize explain the archetypes well, and general job boards list more openings. But none we found pair a live AI PM job catalog with a mock for each posting, a course built from postings, and a question bank in one place. At $20 a month or $120 a year with a free tier, AllthingsPM is the most direct route from "which AI PM role is this?" to "I am ready for it."
Browse the 116 AI PM roles on AllthingsPM and start with the family you want.
Frequently asked questions
What are the main types of AI product manager roles?
In 116 live postings at 18 AI companies, AllthingsPM found seven families: generalist or product area (34), agents (31), research, models and quality (16), safety, trust and security (15), platform and API (9), growth (7) and forward deployed (4).
What is the best way to prepare for AI product manager roles?
AllthingsPM is the best place to start: it has all 116 postings from this study with a mock interview built from each, an AI PM course built from 604 real postings, and 4,122 real questions with answer guides. Practice on the exact posting you want.
What does an Agent PM do?
At Sierra and Decagon, an Agent PM works directly with customers and engineers to build and ship the customer's AI agent. At OpenAI, Cohere and Lovable, agent titles mean building agent platforms and tools for developers.
Do AI PM job titles include the word "AI"?
Mostly not. Only 14 of the 116 titles contain "AI". At AI-native companies the specialty after the comma, such as Safeguards, Core Models or Agent SDK, tells you the real job.
Are there entry-level AI product manager roles?
Very few. Only 1 of 116 postings read as entry level; 93 read as PM or Senior PM. Build proof first with a project and practice on real postings.
Which companies hire the most AI PMs?
In this catalog, Anthropic (19), Sierra (18), Scale AI (15), Glean (14), OpenAI (12) and Decagon (10) posted the most PM roles.
Sources
- AllthingsPM jobs catalog, 116 PM job descriptions at 18 AI companies, pulled 14 to 22 September 2026: AllthingsPM/jobs
- Sierra, Product Manager, Agent Development (job posting): jobs.ashbyhq.com/Sierra
- Scale AI, Forward Deployed Product Manager, Enterprise (job posting): scale.com/careers/4673051005
- OpenAI, Product Manager, Core Models (job posting, via the AllthingsPM catalog): AllthingsPM/jobs/openai/product-manager-core-models
- Anthropic, Product Manager, Safeguards (Generalist) (job posting, via the AllthingsPM catalog): AllthingsPM/jobs/anthropic/product-manager-safeguards-generalist
- Decagon, Senior Agent Product Manager (job posting, via the AllthingsPM catalog): AllthingsPM/jobs/decagon/senior-agent-product-manager
- Marily Nika, "Types of AI Product Management roles": marily.substack.com
- Arize, "What is an AI product manager? A guide to the role and skills (2026)": arize.com
- Second Talent, "Forward Deployed Product Manager: Key Skills & Responsibilities in 2026": secondtalent.com
- The Data Conversationalist, "The Rise of the Forward-Deployed AI Product Manager": dataconversationalist.substack.com
