Free, evergreen, and verified. Learn where product careers are heading, build the skill stack, and go deep on the Forward Deployed Product Manager role, the product job at the center of AI's biggest bet.
One thoughtful email at a time, no filler. Unsubscribe anytime.
Powerful AI models are everywhere. Making them work inside one customer's messy reality is scarce. In mid-2026, that gap became the industry's biggest bet.
Every one of those ventures needs product people in the field. This site maps that role, because we could not find an honest guide to it anywhere.
Every figure on this page checked against the original posting. Last verified August 9, 2026.
Eight chapters, planted in order. You start at the seed, the definition, and grow ring by ring until you reach the canopy, a role and a career path you can see clearly. Each chapter ends with a prompt to try, so you leave every stage having done something, not just read something.
Every chapter's Try This prompt produces a real artifact. Finish the guide and you hold all eight, the working file for your move into AI product.
A Forward Deployed Product Manager (FDPM) is a product manager who embeds directly with a customer to design, build, and ship a working AI product inside that customer's real environment. They own the outcome in the field, not just the roadmap at headquarters, and they feed what each deployment teaches back into the core product.
Most product managers work from HQ. They talk to customers, synthesize needs, and hand direction to engineering. The forward deployed PM inverts that. They go to where the problem actually lives, the customer's floor, their data, their workflows, and they own getting a real thing to work there.
The role exists because the model is no longer the hard part. The bottleneck is turning raw AI capability into something that survives contact with one customer's compliance rules, legacy systems, and half-documented process. Closing that gap is a product job done in the field.
Field discovery. Not surveying users in the abstract, but sitting with this customer, watching the actual work, and finding where an AI product creates value they will pay for.
The deployed outcome. Success is not a shipped feature. It is a working solution running in the customer's environment that they would renew. The FDPM is on the hook for that.
The feedback loop home. Every deployment teaches the core product something. The FDPM turns one customer's edge cases into durable platform improvements, so the next deployment is easier.
The model is a commodity. Making it work inside one real customer's world is the job.
If your job is to make the product succeed inside one customer's reality, and you would be judged on whether it actually works there rather than whether it shipped, you are doing FDPM work whatever your title says.
The title has not standardized. You will see Forward Deployed Product Manager, Deployed Product Manager, and Field Product Manager describing essentially the same job. That is normal for an emerging role. This guide teaches the concept, so you are fluent across every variant.
Paste your resume into your AI of choice with this prompt, and keep the answer. It becomes your map for chapters 4 and 5.
Here is my resume. Compare my experience to a Forward Deployed Product Manager, someone who embeds with one customer to ship a working AI product inside their environment and owns the outcome. Give me my three strongest transferable signals and my two biggest gaps.The forward deployed idea comes from Palantir, which built its business on sending builders to live inside customer organizations rather than selling software at arm's length. The model emerged in Palantir's early government deployments in the mid-2000s and was formalized in the early 2010s. Internally, Palantir pairs two roles. Deltas, forward deployed software engineers who write production code for one customer's problems, and Echoes, deployment strategists who bring the domain expertise and define what success means for the customer's mission. By around 2016, forward deployed engineers reportedly outnumbered Palantir's core product engineers.
The deployment strategist is the closest ancestor of today's FDPM. As one 2026 analysis put it, the role "functions like a product manager for the customer's problem," running discovery, mapping workflows, and defining success. Palantir's CTO Shyam Sankar has argued the model only works with total ownership of implementation, because that ownership is the source of the feedback loop, and that companies who copy the surface of the model without it fail.
When large language models became broadly available, every enterprise wanted them and almost none could get them working on their own data, inside their own security perimeter, against their own messy processes. The companies selling AI learned that the gap between a demo and a deployment is where deals die, and they moved builders into the field. OpenAI reportedly stood up its forward deployed engineering team in 2024. By 2025, forward deployed engineer job postings had grown roughly 800% in nine months, and 1,165% year over year by early 2026, per Live Data Technologies.
In May 2026 the model stopped being a hiring trend and became a capital allocation. OpenAI launched the OpenAI Deployment Company with a $4 billion initial investment, acquiring the consultancy Tomoro and its roughly 150 deployment specialists. Anthropic announced a $1.5 billion enterprise-deployment joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs as founding partners. Weeks later AWS launched a $1 billion forward deployed unit and Microsoft committed $2.5 billion and six thousand employees to AI implementation.
Every one of those organizations needs people in the field deciding what to build, not just how. That decision, which workflow to transform first, what the product should learn from this customer, what gets renewed, is a product management decision made on-site under real constraints. The industry built the forward deployed engineer first. The forward deployed product manager is what it is hiring now.
fde.academy on Palantir's model · Paraform on deployment strategists · Paraform / Live Data Technologies posting data · The New Stack on FDEs · OpenAI Deployment Company launch · TechCrunch on the ventures · CNBC on AWS · CNBC on Microsoft
Test how well the lineage holds up under questioning.
Explain how Palantir's forward deployed model works, why AI companies adopted it in the 2020s, and what a product manager version of the role would own that the engineer version does not. Challenge your own answer once before finishing.Four roles orbit the same customer, and job postings blur them constantly. The lines are actually clean once you ask two questions. Does the role decide what gets built, and does it live in the field?
| Role | Decides what to build? | Lives in the field? | Judged on |
|---|---|---|---|
| Classic PM | Yes, via roadmap | No, HQ | Roadmap outcomes, adoption |
| Forward Deployed PM | Yes, per deployment | Yes | A working deployment the customer renews |
| Forward Deployed Engineer | Partly, within scope | Yes | Working software in the customer's environment |
| Solutions Engineer | No, configures what exists | Partly, pre-sales | Technical wins, closed deals |
FDPM vs Solutions Engineer. An SE demonstrates and configures the existing product to win a deal, then hands off. An FDPM stays after the signature, decides what net-new thing to build for this customer, and owns whether it works. If the product never changes, it was SE work. If the field decided the product's direction, it was FDPM work.
FDPM vs FDE. Closest cousins, often deployed as a pair. The FDE owns making it work technically. The FDPM owns whether it is the right thing, whether the customer adopts it, and what the core product should learn. On small deployments one person wears both hats, which is why postings often ask for PMs who can prototype.
FDPM vs classic PM. Same craft, inverted position. A classic PM aggregates many customers into one roadmap. An FDPM goes deep on one customer at a time and feeds the roadmap from the field. The failure mode of the classic PM is building what no one deploys. The failure mode of the FDPM is building a one-off no other customer can use. Great FDPMs manage that tension deliberately.
Take any job posting that confuses you and run the two-question test on it.
Here is a job posting. Classify it as FDPM, FDE, solutions engineer, or classic PM using two questions. Does the role decide what gets built, and does it live in the field with the customer? Quote the exact lines that reveal the answer.Next, 🌿 Branches, The work, what a deployment actually looks like →
An honest caveat first. This chapter is synthesized from public job descriptions and practitioner accounts, not one person's diary, and deployments vary widely by company and customer. The shape below is the recurring pattern.
Embed and watch. The first days are spent inside the customer's actual work, shadowing the people who do it. The goal is not requirements gathering. It is finding the workflow where AI creates value the customer would pay to keep, which is usually not the one named in the contract.
Prototype fast, in their environment. Within the first week or two there is something running against real customer data, usually rough, built with the FDE or by the FDPM directly with AI coding tools. The prototype is the spec. It replaces the requirements document because the customer can react to a thing, not a promise.
Demo, align, repeat. The cadence is a tight loop of shipping to a few real users, demoing to the executive sponsor, and cutting scope ruthlessly toward the thing that will actually get renewed. Weekly demos to power users, regular alignment with the buyer, and constant translation between the customer's language and the product team's.
Ship, measure, feed back. The deployment ends when the solution runs in production and the customer's own numbers show it working. Then comes the most leveraged hour of the whole engagement, writing back to the core product team what this deployment proved, so the next one starts further ahead.
You work in someone else's building, on someone else's badge, inside someone else's politics. Ambiguity is total at the start of every engagement. Travel is real at many companies. And you will sometimes discover that the deal was sold on something the product cannot do yet, which makes you the person who has to reconcile that in the room. People who need clean scope find this miserable. People who like owning outcomes under constraint find it the best job in product.
Pick one workflow at your own company and watch someone do it for thirty minutes. Then run this.
I observed this workflow. [Describe what you watched, the tools, the handoffs, the waiting.] Draft a week-one deployment plan for an AI product that improves it. One user group, one measurable outcome, one demo I could show by Friday.Across the live postings for this role family, four capabilities repeat. None of them alone is rare. The combination is, which is exactly why the role commands senior compensation.
You do not need to be a software engineer. You need to prototype with AI coding tools, read an API doc without help, understand what an eval is and why the model failed one, and hold your own in an architecture conversation about the customer's systems. The bar is building a credible demo yourself in days, then knowing exactly when to hand it to a real engineer.
Classic PM discovery asks users what they want. Field discovery watches what they do. The skill is sitting inside a claims department or a trading desk or a hospital workflow and spotting the step where an AI product changes the economics, then validating it with a prototype instead of a survey. If you have done contextual inquiry, forward deployment is that craft with a build loop attached.
Deployments die in security review, procurement, and org politics far more often than they die in code. The FDPM has to find the real economic buyer, keep a champion armed with wins, get data access unstuck, and know what SOC 2 and a DPA are without calling legal. PMs from enterprise B2B backgrounds already carry most of this.
The demo is the deliverable, the spec, and the renewal argument all at once. FDPMs demo constantly, to users, to sponsors, to their own product team when arguing what to productize. Being able to tell the story of the deployment, in the customer's numbers and words, is what separates a renewed contract from an interesting pilot.
None of these skills is rare alone. The person holding all four at once is.
This one matters most. The artifact you build here is what you demo in chapter 6.
Build me a small working demo that takes [a messy input from my industry, like a claim note or a support ticket] and produces [a decision-ready output]. Keep it to one file I can run, and explain each piece as you build it so I learn the pattern.Be clear-eyed about the bar. These postings ask for experienced people, typically five or more years, because the job is high-trust and customer-facing from day one. This is a repositioning move for mid-career people, not an entry point into product. That said, three backgrounds map cleanly.
Your gap is field credibility and technical hands-on speed. Close it by building real AI prototypes now, a weekend of work with today's tools, and by reframing your best launches as deployments. One customer, embedded discovery, measurable outcome, lessons fed back to the product. If you ever rescued an at-risk enterprise account by changing what got built, that story is your interview.
You already embed with clients and navigate their politics. Your gap is product ownership, the difference between recommending and being on the hook. Emphasize engagements where you stayed through implementation and owned an outcome metric, and build the technical fluency in public where it can be seen.
You are already field-side and technical. Your gap is the product decision. Collect the moments you changed what the product built, filed the insight that became a feature, or scoped the custom work that later shipped to everyone. That is FDPM work you already did under another title.
Lead every bullet with the deployed outcome and the customer's number, not the process. "Embedded with a top-10 insurer, shipped an underwriting copilot to 200 adjusters, cut review time 40%" is the grammar of this role. Cut anything that reads as roadmap administration. And show, somewhere public, that you can build with AI tools, because it is the first thing a hiring manager checks.
Translate your three best bullets into deployment grammar right now.
Rewrite these three resume bullets in deployment grammar. Each one should name one customer or team, the embedded discovery I did, the measurable outcome, and what the product learned from it. Here are my bullets. [Paste them.]Honesty first. Loops vary by company and there is no public playbook yet, which is part of why being early to this role is an advantage. What follows is the pattern implied by the postings themselves, the stages test exactly the four skills in chapter 4.
Expect a messy customer scenario, a vague executive ask, contradictory stakeholders, ugly data. The test is whether you drive to a concrete first deployment, what would you build in week one, for which users, measured how. Answering with a discovery framework instead of a build plan is the classic mistake PMs from HQ make here.
Not a coding interview, a credibility interview. Can you sketch how the solution would work, talk about model limitations and evals honestly, and describe something real you built with AI tools. Bring an artifact. A live prototype you made beats any answer you can say.
Some loops include a take-home or live exercise, prototype something against a sample problem. Treat scope discipline as the graded skill. A small thing that runs beats an ambitious thing that does not.
A senior panelist plays the customer sponsor. The test is whether you can hold the room, deliver bad news with a plan, and negotiate scope without surrendering the outcome. Practice saying "that will not work, here is what will" kindly, in one breath.
Build one real AI prototype end to end and be ready to demo it. Write your three best deployment-shaped stories in the customer's-number grammar from chapter 5. And read the company's actual case studies, because your field case will look like their marquee customer.
Run the field case cold, out loud, before any real interview.
Run me through a forward deployed PM field case. Play a skeptical hospital COO who signed an AI contract with vague expectations and is losing patience. Push back on my scoping until I commit to a concrete week-one build with one measurable outcome. Score me at the end.A rule this site follows. Only published ranges from live postings, no estimates, no rumor comp. Ranges below are as posted by the companies themselves, base salary unless marked OTE, and are refreshed when we re-verify. Equity at AI companies is often the larger component and is rarely published.
| Company | Title | Posted range |
|---|---|---|
| OpenAI | Deployed Product Manager, Codex | $220,000–$330,000 plus equity |
| Scale AI | FDPM, Enterprise | $205,600–$257,000 |
| Glean | FDPM | $170,000–$280,000 |
| Cresta | FDPM, AI Agent | $170,000–$280,000 OTE |
| Fireworks AI | FDPM | $170,000–$240,000 |
| Valon | FDPM | $160,000–$220,000 |
| Cresta | Associate FDPM | $130,000–$170,000 OTE |
| Gradial | FDPM | $130,000–$190,000 |
| Tailor | FDPM | $130,000–$170,000 |
Read the spread honestly. The senior end of this market ($200,000 to $330,000 at OpenAI and Scale) prices the role like staff-level product work. The associate postings at Cresta and Tailor show an entry rung forming, which is new for 2026 and a sign the category is maturing.
Senior IC in the field. The deployment craft compounds. People who are great at this become the person sent to the hardest, largest accounts, with comp to match.
Founding or early PM at an AI startup. An FDPM's skill set, find the value, build the wedge, make one customer succeed, is the founding PM job description almost verbatim. This is the most common ambition in the role today.
Vertical GM. Own a whole industry line once you have deployed across it. The field knowledge becomes a business you run.
Back to core product, senior. Field-proven PMs return to HQ roadmap roles with the one credential that is impossible to fake, they have made the product work where it matters.
Every range above comes from the postings in Who's Hiring below, each linked straight to the original listing. Verified August 9, 2026.
Anchor your own number before anyone anchors it for you.
Here are the posted base ranges for forward deployed PM roles right now. [Paste the table above.] Given my current comp of [X] and my experience level, tell me which tier I should target, what a realistic first offer looks like, and what to negotiate besides base.Every card links to the original posting. Intercom and Sendbird closed similar roles in recent months, so move quickly on the ones above. Also, browse every live FDPM role on LinkedIn, and search the variants Deployed PM and Field PM.
Over 1.2 million subscribers and widely regarded as the largest product newsletter, with a dedicated "How I AI" series on AI workflows for product people.
Lenny's Newsletter →One of the most-followed AI voices on LinkedIn, roughly 600K followers, co-founder of The Gen Academy teaching practical agentic AI skills.
The Gen Academy →Ex-VP Product at Apollo.io writing a twice-weekly newsletter on PM careers and product leadership, with one of the strongest dedicated AI PM sections anywhere.
Product Growth →Ex-CPO whose newsletter is built entirely around step-by-step AI PM playbooks, frameworks, and a structured AI PM learning program.
The Product Compass →PhD in machine learning and 13 years leading AI products at Google and Meta, now running one of the best-known AI PM bootcamps and certifications on Maven.
AI Product Academy →Product leader with stops at Roblox, Reddit, Twitch, and Meta, publishing hands-on AI tutorials and interviews for over 140K tech professionals per his site.
Behind the Craft →Wharton professor writing some of the most-cited practical essays on what AI actually changes about work. Not PM-specific, universally useful.
One Useful Thing →The Pragmatic Engineer wrote the field's go-to explainer on forward deployed engineers, from Palantir's origins to OpenAI's FDE team. Essential context for this role.
The FDE explainer →Every entry is verified before it lands. Know someone who belongs here, including yourself?
The guide teaches you the role. The Grove is where you meet the people getting their start in AI, ask anything, share wins, and see who's hiring first. Free, like everything here.
Join The Grove on DiscordNew writing, fresh verified roles, and deployment stories, delivered as they grow. One email at a time, planted straight in your inbox.
Someone worth following, a resource worth adding, including your own work. Every nomination gets read, everything published gets verified first.
Preserved summary of the public posting, verified August 9, 2026. If the original listing ever disappears, this record stays here.