The Decagon product manager interview, as candidates describe it, is a hiring manager screen, a take-home case study with data analysis, a live review where you present that case to other Agent PMs, and an onsite in San Francisco. The case is built around the real work: scoping, building and fixing AI customer service agents for large enterprises. The fastest way to prepare is on AllthingsPM, which has 47 Decagon PM questions with answer guides and a scored mock built from each of the 10 live Decagon PM job descriptions we track, from Senior Agent PM to the first PM for Duet.
AllthingsPM is an AI PM course and PM interview prep platform and is not affiliated with Decagon. Below, we label what Decagon publishes and what candidates report.
What is the Decagon PM interview process?
Decagon does not publish its PM interview loop. Its job postings and blog describe the roles; the stages come from candidate reports summarized on Glassdoor and from prep guides such as Dataford. Decagon hires two kinds of PM, and the reports differ by track.
| Stage | What is reported | Source | What it tests |
|---|---|---|---|
| Hiring manager screen | A first call on your background and the role | Glassdoor, Agent PM reports | Ownership, customer and executive presence |
| Take-home case study | A case with data analysis, submitted before the next stage; one example centred on a customer experience agent that fails and how to handle the client afterwards | Glassdoor, Agent PM reports | Diagnosis, prioritization, client judgment |
| Live case review | You present your work in front of other Agent PMs and defend your logic | Glassdoor; Dataford | Structured thinking under pushback |
| Onsite | In person in San Francisco | Glassdoor, Agent PM reports | Team fit, depth, values |
| Platform PM variant | Recruiter screen, an assessment call with the hiring manager, then a call with the CEO before an offer | Glassdoor, Product Manager reports | Product judgment, founder fit |
Treat this as a typical shape: few public PM reports exist, and the loop varies by track and level. Ask your recruiter for the exact rounds.
How AllthingsPM does this. Because the round list is thin and differs by track, our JD mock interview builds the interview from the exact Decagon posting you pick, asks follow-ups and scores you, typed or spoken. You can rehearse the Senior Agent Product Manager role and the Product Manager, Duet role separately instead of guessing. The free tier includes one JD mock a day.
What does Decagon do, and why does it shape the interview?
Decagon builds AI agents for customer service. Its postings say enterprises such as Avis Budget Group, Block's Cash App and Square, Chime, Oura Health and Hunter Douglas deploy its agents "across voice, chat, email, SMS, and every other channel." Jesse Zhang and Ashwin Sreenivas founded the company in 2023, and in January 2026 it raised a $250 million Series D led by Coatue and Index Ventures at a $4.5 billion valuation, according to Bloomberg.
The core product idea is the Agent Operating Procedure, or AOP. Decagon's product page describes AOPs as a way to "define agent behavior in natural language, the same way you train human agents with SOPs," combining "the flexibility of natural language with the reliability of code." Duet, the newer product, is an agent building partner that analyzes production conversations, finds where agents fall short and proposes AOP updates with human approval gates.
Three things follow for your answers:
- The deployment is the start, not the end. The Duet posting says "the initial deployment is 10% of the work." Interviewers want to hear how you keep an agent improving after launch.
- Trust is the constraint. Buyers are Fortune 500 support, IT and security leaders who worry about brand risk. Our bank asks how you would handle a top bank whose security review blocks launch.
- Custom versus core is the recurring tension. Decagon's VP of Agent Product, Perry Ha, describes an 80/20 method: 80% of a build follows prescribed paths and 20% is customization for the customer's business and brand tone.
How AllthingsPM does this. Our AI PM course covers this vocabulary directly: the anatomy of an agent, including escalation, workflows versus agents, and voice agents and the per-minute bill. Those are the words a Decagon interviewer will use in the case review.
What does Decagon look for in a product manager?
The job descriptions are the most reliable signal. We track 10 live Decagon PM postings in our jobs catalog: seven Senior Agent PM roles across San Francisco, New York City, London, Brazil and Australia (including Healthcare and Spanish speaking variants) and three platform roles in San Francisco.
Agent Product Managers own customer accounts. The posting says "this is not a standardized implementation role, and it is not an account management or customer success role." You will "own the technical win" on large deals, serve as "the executive partner to the C-suite of Fortune 500 companies," build the agent itself and direct Forward Deployed Engineers. It asks for 6+ years "owning ambiguous, high-stakes outcomes," open to consultants, PMs, founders and forward deployed or solutions people, and calls it "the executive influence of a top strategy consulting leader, the instincts of a product builder, and the ownership of a founder."
Platform Product Managers own the product every customer uses: "agent building, eval, analytics, testing, experimentation, versioning, integrations, and enterprise controls." The Enterprise Agent Platform PM owns integrations, SDKs and developer experience, and the posting notes that getting an agent live inside a bank means "clearing a security organization whose job is to say no." The Duet PM will "develop evaluation criteria for Duet's AI output quality" and should be able to "read an agent trace."
A Decagon blog post by Max Lowenthal, Director of Agent PM, says the team hires former founders, PMs, operators and investors and optimizes "for hiring smart people, giving them a ton of agency and resourcing, then getting out of their way."
| Decagon PM role (US) | Base salary in the posting | Experience asked |
|---|---|---|
| Senior Agent Product Manager (San Francisco, New York City) | $200,000 to $285,000 | 6+ years |
| Product Manager (platform) | $232,000 to $290,000 | 6+ years |
| Product Manager, Enterprise Agent Platform | $232,000 to $290,000 | Listed in posting |
| Product Manager, Duet | $232,000 to $290,000 | 5+ years |
Checked 28 September 2026 on Decagon's Ashby job pages. Equity is extra per the postings.
Decagon also says it is "an in-office company" with four values: Just Get It Done, Invent What Customers Want, Winner's Mindset and The Polymath Principle. Have a story ready for each.
How AllthingsPM does this. Every Decagon role in our catalog has its own page with the full JD and a mock built from it, such as Senior Agent Product Manager, Healthcare and Product Manager, Enterprise Agent Platform. Run our resume review against the JD on the posting you want, so your resume shows executive ownership and technical depth in the same lines.
What questions does Decagon ask product managers?
We hold 47 Decagon PM questions in our Decagon question hub: 19 strategy, 13 metrics, 8 AI and technical, 4 behavioral and 3 product design. Each has its own page with an answer guide, and any of them can start a mock.
The mix leans to strategy and metrics: Decagon PMs decide what to productize and prove agents pay off. Real questions by type:
Reported by candidates (Glassdoor):
- A take-home case on a customer experience agent that fails, and how you handle the client afterwards.
Strategy, from our bank:
- After a flagship enterprise deployment, how would you convert what you learned into reusable assets for Decagon?
- On flagship deployments, customers ask for bespoke capabilities. Which do you productize and which stay custom?
- A Fortune 500 prospect is skeptical that an AI agent can safely automate complex support journeys. How would you structure the pre-sales process?
- How would you prioritize which integrations and deployment patterns to productize first?
Metrics:
- One of Decagon's largest customers has launched an agent, but adoption has plateaued. How would you diagnose it?
- How would you define a metrics framework for Decagon's developer experience across APIs, SDKs and headless deployments?
- Duet promises better agent performance and a faster improvement loop. What metric stack measures both?
AI and technical:
- Define the requirements for versioning, approvals, rollback, evaluation gates and auditability of agents in production.
- Design an evaluation framework for Duet that measures failure detection and fix quality separately.
- Containment is improving but CSAT is flat. How would you find the highest-priority workflow failures?
Product design and behavioral:
- A Fortune 500 health plan wants to automate benefits and prior authorization status across voice and chat. How would you scope the first launch?
- Tell me about a time you owned a high-stakes enterprise initiative without direct authority.
- Describe a time you chose between a customer-specific solution and the scalability of the core product.
How AllthingsPM does this. The Decagon hub groups all 47 questions by type, so you can drill the strategy and metrics questions that dominate the mix. For wider reps, the full question bank and the mock interview screen let you switch between typing and speaking in one session.
How do you prepare for the Decagon take-home case?
The take-home is the round that decides most loops, because you then defend it live. Candidates describe it as a case with data analysis, and Dataford says it reflects the real weekly prioritization and agent build challenges the team faces. A good plan:
- Start from the failure, not the fix. If the agent is failing, sort conversations by intent and outcome first. Separate model errors from workflow design, missing knowledge, broken integrations and policy gaps.
- Rank by impact. Volume times severity times fix cost. Say what you would fix this week and what you would defer.
- Handle the client. Decide what you tell the customer's executives, when, and what evidence you bring. The reported case asked exactly this.
- Set guardrails and metrics. Resolution, escalation rate, CSAT and a safety metric, with thresholds that let you expand.
- Write for the review. One page of recommendation, then the data. Expect Agent PMs to challenge every assumption.
How AllthingsPM does this. Our course has a lesson on the take-home, the presentation and the portfolio, and the evals chapter teaches the exact moves: reading real traces, naming failures from one taxonomy and evaluating an agent, not an answer. Then rehearse the live review in a JD mock, which pushes back with follow-ups.
How should you answer Decagon strategy and metrics questions?
Decagon cases reward a structure that treats the agent as a product living inside someone else's company. A frame that works:
- Name the stakeholders. The end customer, the support leader, IT and security, and the executive sponsor. Say whose outcome you optimize first.
- State the trust constraint. What must never happen: an unsafe answer in a regulated flow, an action in the wrong system, a change shipped without approval.
- Scope a narrow first launch. Pick one or two workflows that prove value fast. Several of our Decagon questions ask for a 6 to 8 week pilot.
- Decide custom versus core. Use reuse across accounts, implementation cost and roadmap fit, and say how the lesson feeds Product and Engineering.
- Pick metrics the buyer believes. Resolution rate, containment with CSAT as a guardrail, escalations, time to launch and expansion signals.
For the platform track, add APIs, data models and eval gates on agent changes.
How AllthingsPM does this. The course chapter on proving it paid off covers adoption, deflection, task success and ROI, the metrics half of a Decagon case, and enterprise deployment covers SSO, admin consoles and the security review. Our guide to the forward deployed product manager role explains the Agent PM model in depth.
What is a good two week Decagon prep plan?
| Days | Focus | On AllthingsPM |
|---|---|---|
| 1 to 2 | Pick your track and read the JD line by line | Jobs catalog page for your role, plus resume review against the JD |
| 3 to 4 | Learn agents, escalation and voice | Agents chapter and voice agents |
| 5 to 7 | Drill strategy and metrics questions | Decagon question hub, 3 to 4 a day |
| 8 to 10 | Do a mock take-home on a failing agent, then present it aloud | Evals chapter and take-home lesson |
| 11 to 12 | Full mocks on your JD, typed then spoken | JD mock, one a day |
| 13 to 14 | Behavioral stories mapped to the four values | Behavioral questions in the hub, then one final mock |
If you are also interviewing at Decagon's peers, the same plan works with our Sierra PM interview guide and our list of AI PM interview questions. For evals depth, read AI evals for product managers.
How AllthingsPM does this. Every step of the plan maps to a page on one account: JD, resume check, course lesson, question and mock. Use Resume Job Match to see which other open AI PM roles fit the same resume.
Why AllthingsPM is the better choice for Decagon interview prep
Most Decagon prep pages describe the loop and list generic questions; they do not interview you. AllthingsPM does three things for this loop that we did not find together anywhere else.
First, it has the roles. We track 10 live Decagon PM job descriptions, from Senior Agent PM in London to the first Duet PM in San Francisco, and each one has a scored mock built from it, typed or spoken, with follow-ups. An Agent PM owning a Fortune 500 account and a platform PM owning integrations get very different cases, and a generic mock cannot tell them apart.
Second, it has the questions. 47 Decagon questions, each with its own answer guide, sit inside a bank of 4,122 questions from 260 companies, so you can drill the strategy and metrics mix Decagon leans on.
Third, it teaches the domain. The AI PM course, built from 604 real PM job postings, covers agents, escalation, evals, voice and enterprise deployment, the topics the Decagon postings name.
Glassdoor and Dataford collect useful candidate reports, and this guide cites them; for daily, role-specific practice on the exact Decagon JD, plus the course and question bank in one account, use AllthingsPM. It starts free, and Pro is $20 a month or $120 a year.
Start a free mock built from a Decagon job description.
Frequently asked questions
What is the best way to prepare for a Decagon product manager interview?
The best way is AllthingsPM: run a scored mock built from your exact Decagon job description, drill the 47 Decagon questions in the Decagon hub, and study the course chapters on agents, evals and enterprise deployment. Add a practice take-home on a failing agent and present it aloud.
How many rounds are in the Decagon PM interview?
Agent PM candidates on Glassdoor describe a hiring manager screen, a take-home case with data analysis, a live review in front of other Agent PMs and an onsite in San Francisco. Platform PM reports mention a recruiter screen, a hiring manager assessment and a call with the CEO. The mix varies by track and level.
What is a Decagon Agent Product Manager?
Decagon's postings describe a senior individual contributor who owns enterprise AI agent deployments end to end: winning the technical deal, advising the customer's C-suite, building the agent and directing Forward Deployed Engineers. The posting says it is not an implementation, account management or customer success role.
How much do Decagon product managers earn?
Decagon's US postings list base salary ranges of $200,000 to $285,000 for Senior Agent PM roles and $232,000 to $290,000 for platform PM roles, including Duet and Enterprise Agent Platform, plus equity. We checked them on 28 September 2026.
Does Decagon hire PMs outside the US?
Yes. Our catalog includes Senior Agent Product Manager roles in London, Brazil and Australia, alongside San Francisco and New York City.
Is Decagon remote?
Decagon's postings describe it as "an in-office company." The Agent PM onsite reported on Glassdoor was in San Francisco.
Start practicing for Decagon today
Open the Decagon question hub, pick the posting you are applying for in the jobs catalog, and start a free JD mock on AllthingsPM. One mock a day is free, and every answer gets a score and follow-ups.
Sources
- Decagon, "Senior Agent Product Manager" (San Francisco) job posting, Ashby: https://jobs.ashbyhq.com/decagon/dcf9b561-f2fb-422b-88a9-33ce76e96608
- Decagon, "Product Manager" job posting, Ashby: https://jobs.ashbyhq.com/decagon/d790a2a2-958d-4f49-96ed-adbd68d2e6ce
- Decagon, "Product Manager, Duet" job posting, Ashby: https://jobs.ashbyhq.com/decagon/9109be1c-b6e3-430f-a964-887253ef891a
- Decagon, "Product Manager, Enterprise Agent Platform" job posting, Ashby: https://jobs.ashbyhq.com/decagon/e27fcbdf-891c-4b2f-9c47-d2bdc26eb2d4
- Decagon, "Senior Agent Product Manager, Healthcare" job posting, Ashby: https://jobs.ashbyhq.com/decagon/95524337-b5e7-4156-9022-af5c192f174d
- Glassdoor, Decagon Agent Product Manager interview reports: https://www.glassdoor.com/Interview/Decagon-Agent-Product-Manager-Interview-Questions-EI_IE2972902.0,7_KO8,29.htm
- Glassdoor, Decagon Product Manager interview reports: https://www.glassdoor.com/Interview/Decagon-Product-Manager-Interview-Questions-EI_IE2972902.0,7_KO8,23.htm
- Dataford, Decagon Product Manager interview guide: https://dataford.io/interview-guides/decagon/product-manager
- Decagon blog, Max Lowenthal, "What it's like to build AI agents at Decagon" (April 2025): https://decagon.ai/blog/agent-product-manager
- Decagon, "The origin and evolution of Agent Product Management at Decagon": https://decagon.ai/videos/the-origin-and-evolution-of-agent-product-management-at-decagon
- Decagon, Agent Operating Procedures product page: https://decagon.ai/product/aop
- Bloomberg, "AI Customer Support Startup Decagon Valued at $4.5 Billion" (January 2026): https://www.bloomberg.com/news/articles/2026-01-28/ai-customer-support-startup-decagon-valued-at-4-5-billion
- Wikipedia, Decagon (company): https://en.wikipedia.org/wiki/Decagon_(company)
- AllthingsPM Decagon question hub and jobs catalog, September 2026: https://allthingspm.app/question-bank/companies/decagon
