If you're buying enterprise AI, you're increasingly likely to be offered an engineer who comes to work inside your business, a role called the Forward Deployed Engineer. Palantir popularised the FDE and the major AI labs have been following suit: 9,000+ FDE roles have been created in recent months,1 with Anthropic committing $100 million to train 10,000 partner staff2 by the end of 2027.
The role places a specific kind of engineer inside customers, one who can understand your business problem, apply that to the product you've been sold, and produce heaps of code to integrate your data and systems with the vendor's platform to maximise your value (and of course stickiness and retention). For the vendors selling their products on the value of needing all of the data, this makes sense, and the AI labs can afford to make this work however it adds up. They have lots of cash, and an engineer embedded inside a customer is highly valuable and hard to replace.
Now that the concept is spreading to the enterprise, customers need to determine if FDE is right for them, and to understand how they will pay for the role.
Is it a loss leader?
Some say so openly. C3.ai told investors in September that "selective investments in a forward-deployed engineering organization" would take its adjusted gross margin from 50% to the mid-40s.3 Its chief executive described it as overinvesting in existing customers "in the short run", with a coding product expected to reduce the need for the engineers later.
An a16z essay from 2025 makes the case for doing this on purpose: sell the services at cost and win on retention.4 It points to ServiceNow, whose gross margin was 63.2% at IPO and 79% in 2024, and Workday, at 54.1% and then 75%. The margin recovered because the software did the earning once the early work was done.
It's (probably) not just consulting
If your assigned FDE will be genuinely enhancing the vendor product whilst delivering you incremental value, this could be the right approach. Matthew Mayo's test in KDnuggets gives a useful framing: an FDE "is defined by whether what they learn at the customer changes what the company builds next".5 His warning signs include customer-specific features that never reach the product, and engagements that "get extended rather than concluded".
Considerations for the customer
The customer-specific features warning will be familiar to any mature product management organisation that builds the product for the market, not for customers, but customers don't always see it that way and may have one of two concerns:
- "Why am I subsidising the product roadmap development?", or
- "When will this custom work be properly productised?"
I didn't find any independent data on how often the work of an FDE reaches the product.
The extension warning deserves attention too – an engagement that never concludes is a dependency with a day rate. Consider what happens once the engineers leave, or if they don't.
If you're being offered an FDE, three questions are worth asking:
- Can they show you something built last quarter that is now in the product?
- Who owns and maintains the FDE outputs?
- Is the service priced to earn money, or subsidised to win the account (and what happens when the subsidy ends)?
Be sure you're not paying premium day rates for a custom build in lieu of product maturity and documentation.
References
- https://www.thediff.co/archive/forward-deployed/ | "Forward Deployed", Nikhil Davar and Byrne Hobart, The Diff, 21 September 2026 (free registration required)
- https://www.anthropic.com/news/claude-frontier-academy | Claude Frontier Academy, Anthropic
- https://www.fool.com/earnings/call-transcripts/2026/09/09/c3ai-ai-q1-2027-earnings-call-transcript/ | C3.ai Q1 FY2027 earnings call transcript, 9 September 2026
- https://a16z.com/services-led-growth/ | "Trading Margin for Moat: Why the Forward Deployed Engineer Is the Hottest Job in Startups", Joe Schmidt, a16z, June 2025
- https://www.kdnuggets.com/forward-deployed-engineer-ais-hottest-new-career-or-consulting-with-a-better-title | "Forward Deployed Engineer: AI's Hottest New Career, or Consulting With a Better Title?", Matthew Mayo, KDnuggets, 2 October 2026