Forward Deployed Engineer vs Project Manager: Who Actually Builds It?

A project manager coordinates the people who build. A forward deployed engineer is one of the people who build. If engineers are already on your problem and what keeps slipping is sequencing, dependencies and dates across three teams, hire the project manager. If nothing is actually being built, or a pilot has been sitting at eighty percent since the spring because everyone who could finish it has a day job, hire the forward deployed engineer, because a coordinator with nobody to coordinate produces a very accurate weekly report about a system that does not exist.

One of these jobs has a census category. The other has eleven mentions in history.

Before comparing them, look at how differently the two titles sit in the world, because that asymmetry explains most of the confusion I hear in boardrooms.

Project management is one of the largest professions in the American economy. It has its own federal occupation code, 13-1082, project management specialists, and the Bureau of Labor Statistics counted 1,066,670 people doing it in the United States as of May 2025, at a median wage of $102,320, a mean of $110,740 and $167,970 at the 90th percentile. In the New York, Newark and Jersey City metropolitan area alone there are 68,910 of them, at a median of $126,660. For comparison, the same programme counts 1,687,890 software developers nationally at a median of $135,980.

The forward deployed engineer has none of that infrastructure. Search all 1,104 occupations in that federal wage file for the words forward deployed and you get nothing. Not a low count. Zero. Nor has the title reached the audited page in any quantity: search every annual report ever filed with the Securities and Exchange Commission and the phrase project managers appears in 4,303 Form 10-K filings, while forward deployed engineers appears in 11.

Eleven, against four thousand three hundred. So when someone frames this as a straight choice between two comparable hires, remember what they are actually comparing. One is a mature profession with a measured population, a known price and a hundred thousand people you could interview this month. The other is a two year old answer to a problem the mature profession does not solve, growing at around 800 percent in job listings between January and September 2025 and over 1,000 percent year on year into 2026.

The federal definition settles most of the argument

The government wrote the job description for you, and it is worth reading slowly. Occupation 13-1082 is defined as follows: analyze and coordinate the schedule, timeline, procurement, staffing, and budget of a product or service on a per project basis. Lead and guide the work of technical staff. May serve as a point of contact for the client or customer.

Analyze. Coordinate. Lead and guide the work of technical staff. Serve as a point of contact. Every one of those verbs takes an object that has to already exist. A schedule presumes work to schedule. Technical staff presumes technical staff.

Now read what the other job is asked to do, in the employer's own words. Anthropic's Forward Deployed Engineer posting, covering New York City, San Francisco and Seattle, asks the person to work within customer systems to build production applications with Claude models, and to deliver technical artifacts for customers like MCP servers, sub agents and agent skills that will be used in production workflows. Databricks asks its forward deployed engineers to own the architecture, lead design decisions, and implement end to end systems spanning data engineering, AI, and application development. Palantir's Forward Deployed AI Engineer posting in New York says the responsibilities look similar to those of a hands-on AI startup chief technology officer.

Build. Deliver. Implement. Own. One vocabulary organises the work and the other performs it. That is not a difference of seniority or a difference of branding. It is a difference of what leaves the building at the end.

At the companies doing this at scale, the two sit next to each other

Here is the part that most comparison articles get wrong, including some written by people selling forward deployment. These are not rival answers to one question. At the firms that have run this model longest, they are two chairs at the same table.

Databricks was advertising 874 open roles this week, 91 of them with forward deployed in the title. Roles titled project manager: zero. But read its Senior Forward Deployed Engineer posting for New York City and one responsibility says the engineer will work with the Databricks technical team, Project Manager, Architect and Customer team to ensure the technical components of the engagement are delivered to meet the customer's needs. Another says to work with engagement managers to scope technical delivery work with input from the customer. The coordination has not disappeared. It has been given to somebody else so that the engineer can stay in the code.

Palantir, which invented this job, had 310 open postings this week, 77 with forward deployed in the title and 13 program managers, none of them called project manager. What it has instead is 35 Deployment Strategists, and their posting says plainly that they identify relevant datasets through deep engagement with customer problems and workflows, and work with Forward Deployed Engineers to integrate the data into a stable and extensible pipeline. Anthropic runs 594 open roles with six forward deployed and 20 technical program managers, and again not one project manager. Datadog: 451 roles, two forward deployed, zero project managers.

Four companies, 2,229 open roles between them, and the phrase project manager appears in none of the titles while the coordination job exists at every one of them under a name that carries an outcome instead of an artefact. Deployment strategist. Engagement manager. Technical program manager. I read that as fashion in the naming and substance in the structure. The structure says somebody holds the plan and somebody else holds the tools, and the second person is never the one taking the minutes.

The money does not run the way people assume

The lazy version of this comparison says the forward deployed engineer is the expensive senior one and the project manager is the cheaper coordinator. Held against real bands, that falls apart.

At Databricks, in the same compensation zone that covers New York, the Senior Forward Deployed Engineer band is $182,000 to $250,208. The Lead Engagement Manager for forward deployed engineering, hiring in Dallas, New Jersey and New York, is $178,800 to $245,850. Those are effectively the same money. At Anthropic the order reverses: the Forward Deployed Engineer band is $280,000 to $320,000, and three of the four technical program manager postings I sampled this week list $365,000 to $435,000, with the fourth at $290,000 to $365,000. The coordinator out-earns the builder by around eighty thousand dollars at the top of the band.

I should be fair about that comparison rather than use it as a gotcha. A technical program manager at a frontier lab is coordinating an organisation, not a client project, and that is a genuinely harder and rarer job than running a delivery plan for one customer. The honest conclusion is the useful one anyway: this is not a ladder. Neither title is the senior version of the other. They are priced by what they are accountable for, and you should buy the accountability you are missing.

Read it as a chief financial officer and the line gets sharper

Twenty years running finance inside high growth companies, through three exits, taught me to ignore the org chart on questions like this and ask which budget the person comes out of and what number they are measured on.

Databricks writes it into the postings without meaning to. Its forward deployed engineers, in its own phrase, are billable and know how to complete projects according to specification. Its lead engagement manager contributes to team level booking targets and consumption goals and leads the sales of complex professional services offerings. One seat converts money into a working system. The other converts a relationship into signed work and keeps the plan honest. Both are legitimate. Only one of them adds a system to your company this quarter.

The second CFO number is coverage. A project manager can carry several workstreams at once, which is why a million of them are employed and why the median sits near a hundred thousand dollars. A forward deployed engineer covers one problem at a time and is priced accordingly. On a cost per project basis the forward deployed engineer is not marginally more expensive. They are several times more expensive, and worth it in exactly one circumstance: when the alternative is a plan that is beautifully tracked and never gets built.

The third is hireability, and it cuts the other way. There are over a million project management specialists in the United States with a published wage distribution, standard certifications and a market you can hire from in three weeks. Forward deployed engineering has no occupation code, no wage series and a title that means different things at different companies, which means every candidate has to be investigated individually. Budget for that. The scarcity is real and it is part of the price.

Where the project manager wins

Plainly, and more often than the current enthusiasm for forward deployment admits.

If the work is already staffed and the failure is between the teams rather than inside them, the project manager is the correct hire and the forward deployed engineer is an expensive person with nothing to build. Three engineering teams, a vendor, a compliance review and a hard date is a coordination problem. Nobody solves that by writing better code. They solve it by holding the dependencies, forcing the sequencing decisions early and making the tradeoffs visible to the person who can settle them, which is exactly what Anthropic asks its technical program managers to do: run the planning and operating cadence with clear owners, sequenced dependencies and a single source of truth on coverage and status.

Regulated programmes are the same story with more paperwork. Procurement gates, security reviews, audit evidence and a board committee asking for a date are not engineering work, and asking an engineer to absorb them is how you turn an expensive builder into a mediocre administrator.

And the project manager scales in a way the forward deployed engineer never will. One good one covers a portfolio. The market is deep, the price is public and the skill transfers between industries. If your company is at the stage where the constraint is that nobody knows what is happening across eleven concurrent initiatives, hire the coordinator and stop reading.

Where the forward deployed engineer wins

When the thing does not exist yet, and nobody on your payroll is going to make it exist.

This is the failure I get called into, and from the inside it never announces itself as a building problem. It looks like a project that is managed impeccably. There is a plan, a weekly call, a status colour and a risk register. What there is not, six months in, is a system doing real work. Every week the plan is updated to reflect the fact that the two engineers who could have built it were pulled onto the platform migration again. The coordination is working perfectly. It is coordinating an absence.

The companies with the best engineering in the world reached the same conclusion with their own balance sheets. In May 2026 OpenAI launched a four billion dollar deployment subsidiary. In July 2026 Anthropic launched Ode with Anthropic, built on its acquisition of Fractional AI and backed by Blackstone and Hellman and Friedman. Google Cloud is hiring 59 forward deployed engineers this year. None of those organisations is short of project managers. They were short of people who would sit inside somebody else's company and build the thing.

Project manager, product manager, program manager

These three get collapsed in conversation and they generate three different arguments, so it is worth separating them before you write a job description.

The project manager owns delivery of a defined scope to a date. The argument against a forward deployed engineer there is the one above: coordination versus construction. The product manager owns what should be built and why, and that argument is about authority over the roadmap rather than about who writes the code, which I have written about separately at Beyond Elevation. The program manager owns the connective tissue between many projects, which at frontier labs is the seat that pays most, as the Anthropic bands show. A forward deployed engineer replaces none of the three. On a small enough team they can absorb the first one for a single workstream, and that is the honest limit of it.

The mistake I see most

Hiring the coordinator to fix the building problem, because it is the hire the company already knows how to make. The requisition template exists, the market is deep, the salary is defensible and the first month feels like progress: a plan appears, a spreadsheet appears, the meetings get shorter. Then the second quarter arrives and the AI pilot is still at eighty percent, now with excellent documentation of why.

There is a mirror image of it that costs just as much. A company reads about forward deployment, hires a very good engineer into a programme that spans six teams and two regulators, and gives them no coordination support at all. Within a month that person is in status meetings all day and their hands are nowhere near a keyboard. You paid a premium for build capacity and then spent it on calendar management. If the work genuinely spans teams, buy both, and if you can only afford one, buy the one that matches the thing that is missing.

When I read any job description, mine or anyone else's, I ignore the title and look for two sentences. Does this person get write access to a system they do not own? And are they still accountable for it after the invoice is paid? Two yeses is forward deployment whatever it says at the top of the page. Anything less is project management, and it should be priced, measured and staffed as project management.

One question that settles it in a meeting

Name the person who will write the code. Out loud, with a first name, this week.

If you can name them and their calendar is real, your problem is sequencing and you want the project manager. If the answer is a team that is already committed elsewhere, or a vendor who has been quoting for two months, or a shrug, then no schedule will save you and the forward deployed engineer is the hire. Most founders know the answer before they finish the sentence, which is why the question is worth asking in front of everybody rather than alone.

Where I come in

I am Hayat Amin. At Beyond Elevation I take the second answer. I move into the company, work inside your systems rather than alongside them, and put my name against something that has to run in production, which on this particular comparison is the whole offer: you are not buying a plan for the AI work, you are buying the AI work. The promise on meethayat.com/services/fde is deliberately a delivery promise. Week one is spent in your meetings and your systems mapping where the work actually happens, the first system is live inside eight weeks watched by a named owner, and you keep the code, the agents, the accounts and the documentation. On the Beyond Elevation side, fractional runs one to two days a week from $5,800 per month, whole projects start at $30,000 and run eight to fourteen weeks, and a company that is not sure yet can start with a two week AI audit at a fixed $3,000. No salary, no equity, no notice period.

For founders in New York and across the United States sitting on this exact decision, the useful half hour is not a debate about two job descriptions. It is naming the person who will write the code and watching what happens in the room. Book that at meethayat.com/services/fde. If the real constraint is money rather than delivery, the CFO seat is at meethayat.com/cfo. The nearest neighbours to this question are forward deployed engineer vs solutions architect, which turns on who does the building, and forward deployed engineer vs sales engineer, which turns on which side of the signature the work sits. If you suspect the whole category is a rebrand, I answered that head on in is a forward deployed engineer just a consultant. Otherwise start at meethayat.com.

Questions I get asked about this

Is a forward deployed engineer a project manager?

No, and at the companies running forward deployment at scale the two sit beside each other rather than replacing each other. Databricks lists 874 open roles and 91 with forward deployed in the title, none called project manager, yet its Senior Forward Deployed Engineer posting in New York City says the engineer will work with the Databricks technical team, Project Manager, Architect and Customer team to ensure the technical components of the engagement are delivered. Palantir does the same under a different name: 310 open postings, 77 forward deployed, and 35 Deployment Strategists asked to work with Forward Deployed Engineers to integrate data into a stable and extensible pipeline.

Forward deployed engineer vs project manager: which pays more in the United States?

Against the national average, the forward deployed engineer, by a wide margin. Against a senior program manager at a frontier AI company, often not. Project management specialists have a median wage of $102,320 nationally and $126,660 in the New York, Newark and Jersey City metro, rising to $167,970 at the 90th percentile. Forward deployed bands read this week were $280,000 to $320,000 at Anthropic and $182,000 to $250,208 for a senior forward deployed engineer at Databricks in New York City. But Anthropic's technical program manager postings list $365,000 to $435,000, above its own forward deployed engineers. Buy the accountability you are missing, not the title you assume is senior.

Do forward deployed engineers code?

Yes, and that is the entire point of the job. Anthropic asks for production applications built inside customer systems and technical artifacts delivered into production workflows. Databricks asks for six or more years in data engineering or software engineering, code in Python, Scala or TypeScript, and end to end systems implemented rather than specified. Palantir compares the role to a hands-on AI startup chief technology officer. The federal definition of project management, by contrast, contains no verb about building anything, which is not an insult to project managers. It is the job.

Do forward deployed engineers travel?

Usually, and less than the reputation suggests. Anthropic's forward deployed engineer posting estimates 25 percent travel to customer sites and the senior role at Databricks in New York City says 20 percent. The coordination seats travel more: Palantir asks its deployment strategists for 25 to 75 percent depending on location and team, and Databricks asks its lead engagement manager for up to 25 to 50 percent. Being in the room is useful. It is not the thing that produces a working system. Write access is.

Do I need a project manager or a forward deployed engineer for my AI project?

Ask whether anybody is already building the thing. If engineers are on it and the failure is sequencing, dependencies, dates and handoffs between teams, hire the project manager and hire a good one. If nothing is being built, or a pilot has been stuck at eighty percent since the spring because the people who could finish it have day jobs, a project manager will give you an accurate weekly account of a system that still does not exist. Most companies under a few hundred people who ask me this have a building problem wearing the costume of a coordination problem.

Can a project manager become a forward deployed engineer?

Only with the engineering, and there is no way around that sentence. The transferable half is real and undervalued: navigating a company that did not ask for you, finding out what is actually true, holding a scope against pressure. The part that does not transfer is that the artefact changes. A project manager is measured on an accurate picture of the work, and a forward deployed engineer is measured on the work. Six years of engineering experience is the floor in the Databricks posting and four years in a technical customer facing role is the floor at Anthropic. Take the coordination skill with you. It will make you better than the engineers who never learned it.

Where do these numbers come from?

Every figure above was read on 11 September 2026 from the source itself rather than from a summary. The definition of occupation 13-1082, the employment counts, the median, mean and 90th percentile wages and the New York, Newark and Jersey City metro figures come from the Bureau of Labor Statistics Occupational Employment and Wage Statistics programme, national and metropolitan estimates for May 2025, pulled from its public data API, with the occupation and area definitions taken from the programme's own reference files. The filing counts come from the Securities and Exchange Commission full text search across Form 10-K filings. The open role counts, responsibilities and every salary band come from the live careers feeds at Palantir, Anthropic, Databricks and Datadog, read the same day. The growth figures for forward deployed engineer hiring, and the notes on OpenAI's deployment subsidiary, Ode with Anthropic and Google Cloud, come from reporting by Pragmatic Engineer, Perspective AI and Paraform. Job boards move. If you are reading this months later the bands will have shifted, and the ratios are the part worth keeping.