On September 8, Accenture announced a new business group with Google. Part of the plan is a workforce of 1,000 forward deployed engineers.
Today you can also open the website of an offshore development firm and hire a forward deployed AI engineer for $22 to $46 an hour. The page helpfully adds that this is 60 to 70 percent below US rates.
Same three words. One of the largest consultancies in the world at one end, a rate card at the other.
When a label fits both ends of a market, it has stopped telling the buyer anything.
That is a shame, because the idea behind it is a good one. A forward deployed engineer, in the original sense, works for the vendor but sits inside the client’s organization, often for months. They work next to the people who will use the system, learn how the business actually runs, ship working software into production, and answer for whether it does the job. It is the opposite of a contractor waiting for the next ticket.
Maybe the label has already reached your leadership meeting. Someone proposes “forward deployed teams” for the homepage. It sounds modern. The AI labs use it. Your sales lead thinks it will open doors, and your delivery lead points out, fairly, that your people have always worked inside client teams anyway.
Before you sign off on it, ask what the new headline actually changes.
The Repositioning That Skips the Work
Here’s the thing about a trendy label. It lets a company change how it sounds without deciding anything.
Real repositioning is slow and uncomfortable. You pick the clients you serve best and accept that you will turn some others away. You find out why your best clients actually bought, which is rarely the reason on your website. You talk to buyers who chose someone else and listen to why. You test the new story with the market before you print it. Then you rewrite the proposals, the case studies, and the first sales conversation so they all say the same thing.
That takes months, and it ends with choices someone can be blamed for.
A new label takes an afternoon and offends nobody.
In my years working with software and technology companies, this is the work that gets postponed first. It feels soft, slow, and hard to measure, so it keeps losing to the next delivery sprint. The cost shows up in the data. Bain surveyed more than 1,000 B2B companies this January. Only 4% think they have a clear, consistently understood value proposition. The ones that do grew about 1.6 times faster than average.
So I understand the appeal of forward deployed. It arrives with borrowed prestige from Palantir, OpenAI, and Anthropic. It sounds like a strategy. It feels like progress. And it spares everyone the conversation about who you are for and why anyone should care.
But look at what it describes. It describes you: how your people work, where they sit, how close they get. Your client never started a purchase by thinking about any of that.
Gartner describes the first job in a B2B buying process as problem identification, which in plain words means “we need to do something.” The thought that starts a purchase sounds like “this part of the business cannot keep running like this.” It never sounds like “we need forward deployed engineers.”
That moment, when the client decides something has to change, is where you want to be. Ideally before they have finished the thought, while they are still working out what the problem even is. Nobody gets there by describing their delivery model more fashionably. You get there by being known for the problem.
I’ll concede part of this. Some buyers do search for the format. Job postings for forward deployed engineers grew about 800% between January and September 2025, according to Indeed data reported by the Financial Times.
That number measures who is hiring engineers. It says nothing about who is buying them. I looked for tenders and requests for proposals that ask vendors for forward deployed engineers and found none. That does not prove they don’t exist. It does suggest that a buyer who arrives asking for a format has already turned your service into a commodity, and will compare you on the only thing left to compare. The rate.
What Your Clients Already Calculated
The rate is exactly where your clients are pushing right now.
They have read the same AI headlines you have, and many of them have already done the math on your behalf. In a survey HFS Research ran with EY among Global 2000 executives, three in four expect AI-led services to cost at least 10% less. The average expected reduction is 19%. One procurement VP at an industrial manufacturer put it plainly: “AI should be more efficient, but pricing rarely shows where the efficiency goes.”
The providers hear it too. In May, Endava’s CEO told analysts that clients “are harvesting a little bit more of that benefit than we would prefer.” In July, Cognizant’s CEO said some clients are “starting to ask us to deliver an AI-infused rate card.”
Let’s be clear about what is happening, because the scary version is wrong.
Rates are not collapsing. Endava’s CFO described price per work day as “pretty stable.” Globant’s CEO said on his August call that the deflationary scenario is “something that we’re not seeing right now.” What shrinks is the number of days. The client keeps paying the same for a day of your team and quietly expects to buy fewer of them.
So the pressure lands on the unit. Every hour your team’s AI saves is an hour the client expects not to pay for.
If you run a SaaS company, you know the same conversation in a different costume. The customer’s CFO asks why they still pay per seat when the agent does the work. Bain looked at the published pricing of about 200 B2B SaaS vendors and found that about four in five of those adding AI pricing chose capacity models. You still sell a quantity. The buyer still gets to ask why the quantity costs what it does.
Now go back to the homepage proposal. A forward deployed team is a team. You will almost certainly bill it per person, per month. You would be answering pressure on the effort unit by giving the effort unit a better name.
Where the Label Came From
Palantir popularized the title. By the account a16z published this January, it gave its solutions and integration engineers a new name. Their work was “critical to customer success and business reputation, but traditionally treated as replaceable help,” and the new title gave it the standing it deserved. The label started life as a rebrand, and a very good one.
What made it work was the business behind it. Palantir sells software: platforms called Gotham, Foundry, and AIP, licensed to governments and companies on multi-year contracts. Its forward deployed engineers set up and extend Palantir’s own platforms inside the client’s operation. As a16z’s Marc Andrusko put it, at Palantir “services are a means to drive product adoption, not the primary revenue stream.”
And for years, those services lost money on purpose. Palantir’s 2020 IPO filing says it ran early pilots “at a loss.” In 2019, customers in that first phase brought in $0.6 million in revenue and a $65.4 million contribution loss. Customers in the second phase, where the engineers were building out deployments, ran at a contribution margin of minus 43%. Only accounts that reached scale made money, at 55%. By 2025, Palantir’s gross margin was 82%.
Read that again as a CEO. The engineers were an investment. Software revenue from the accounts that grew paid it back, years later.
The venture firm that did the most to make the model fashionable says the same thing in its own words. Its 2025 essay on services-led growth describes the FDE as “a professional services employee, sometimes rebranded as a forward deployed engineer,” and tells founders to “design incentive structures that allow services to be sold at cost.” Sell the engineers at cost, accept the margin hit, and let the product earn it back.
And then the inventor quietly stepped away. In its 2023 quarterly filings, Palantir replaced “our forward-deployed engineers” with “our employees” in the same sentence. Its latest annual report talks up bootcamps as “shorter, more cost-effective programs.” OpenAI and Anthropic, who revived the title in 2025, have since set up separate, investor-funded services companies to do this work. Microsoft launched its own deployment unit and says it “goes beyond what has been labeled as Forward-Deployed Engineering.”
The company that made the title famous stopped using it with investors about two years before your competitors started using it on their homepages.
The Part That Does Not Travel
A services firm cannot copy the one thing that made the model pay. Palantir had software revenue waiting at the end of each deployment. You don’t.
When you build software for a client, the client usually owns it. There is no license fee that grows as the account grows.
Your engagement is your revenue.
You cannot sell your people at cost, because your people are the product. Take away the software that pays for them later, and a forward deployed engineer becomes an engineer with a monthly rate who works on the client’s site.
Andrusko warns software startups about exactly this. If you copy only the embedded engineers, he writes, “you aren’t ‘Palantir for X.’ You’re ‘Accenture for X’ with a nicer front-end.” For a services firm, that is the starting point.
The published price lists agree.
I went through 32 services firms that sell forward deployed engineering, from global integrators to nearshore shops, agencies, and consultancies. The big firms publish no prices, so I can’t tell you how Accenture, Deloitte, or EY charge for it. Among the firms that do publish their terms, most charge for time or headcount: hourly rates, a monthly price per engineer, a monthly retainer. None prices it mainly on outcomes. Where outcome pricing appears at all, it is an optional bonus on top of the monthly fee.
One nearshore firm lists “Single Senior FDE: $9,500–14,000” per month, with a three to four month minimum. The page sits in the staff augmentation section of its website. At least that firm is honest about it.
Sure, some firms argue the difference well. HatchWorks, a nearshore firm, says on its page: “Staff augmentation adds capacity. A HatchWorks AI Forward Deployed Engineer adds AI execution capability.” Capgemini renamed its people “Outcome Deployed Engineers,” and Cognizant launched “Frontier Certified Engineers,” which it says differ from forward deployed engineers by “permanence, accountability and something that cannot be trained overnight.” A good embedded team that owns a result is a genuinely different thing from a contractor who takes tickets.
But the name does not decide which one you are selling. The invoice does.
Look at the Invoice
Four questions tell you what you actually sell, whatever the homepage says.
- Do you bill per person, per period?
- Who decides what your people work on this week, you or the client?
- Who is accountable if the business result does not arrive?
- Does the engagement have an end date, or does it run until someone cancels it?
If the answers are “per person,” “the client,” “nobody,” and “it runs,” you are selling staff augmentation. It has a better name now. It has the same unit, and your client’s procurement team is pricing that unit down at every renewal.
No CEO decides to sell effort. It happens by default. Delivery describes the work as the team that does it. Sales writes proposals as roles and rates. Marketing publishes case studies that count engineers and months. Nobody owns the one sentence that says what the client gets, so every asset in the company describes the part the client is now trying to buy less of. A new label changes the top of the homepage. It leaves the rest of that structure exactly where it was.
The label has also left engineering. Stripe has a role called Forward Deployed AI Accelerator inside its marketing team. Agencies and fractional operators now sell “forward deployed marketing.” Guilty as charged, in a way: fractional leadership is embedded work by design. I sit inside the company, work with its people, and stay until the thing runs. I still wouldn’t put the word “embedded” in a headline. Nobody hires a marketing leader because they want one embedded. They hire one because pipeline has stalled.
Forward deployed is a fine answer to the question “how do you work?” It is a poor answer to “why should we buy from you?”
Keep it for the first question. Earn the answer to the second: the problem you end, the situation your best clients were in when they called you, and the result they can see afterward. Where you can, price against that result instead of against the calendar.
The Test to Run Before Your Next Renewal
Put two things side by side: your homepage headline and the pricing line from your last proposal.
If both describe people and time, your client’s procurement lead will price you as people and time. This year, that lead also has a survey saying their peers expect AI-led services to cost about a fifth less.
Now try to write one sentence, in your client’s words, about what they would lose if your team stopped showing up next month.
If you can write it, that is your headline, and the delivery model can go on page two.
If you can’t, what exactly would the new name be for?
