The forward deployed engineer (FDE), the role OpenAI and Anthropic turned into one of the most sought-after jobs in AI, is headed for the finance department.

Ramp said Wednesday it is launching Applied AI Solutions, an offering that sends Ramp engineers inside enterprise finance teams to build and deploy AI agents across work like accounts payable, procurement, and the monthly close. The launch lands days after Ramp raised $750 million at a $44 billion valuation, and it points the FDE model at the part of the enterprise that has been slowest to automate.

Applied AI Solutions is not a self-serve product. It is a high-touch services offering. Ramp engineers work alongside a customer’s finance team to find high-value workflows, pull the context an agent needs out of ERPs, contracts, and approval chains, then build and deploy those agents inside the systems the team already uses.

This is new territory for Ramp. The company built its brand on letting finance teams sign up and start using software quickly. This is a different motion; if anything, it’s more human. Ramp is positioning Applied AI Solutions for select enterprise customers, which makes it closer to embedded engineering or consulting than traditional SaaS.

I wrote last month that the forward deployed engineer had become AI’s hottest job, as OpenAI and Anthropic race to staff enterprises with engineers who can turn AI products into something that works in production. Ramp is taking the same idea and selling it to finance teams.

The reason is familiar: There is a large gap between AI ambition and AI results. Ramp says AI token spend across its customer base has jumped 13x since the start of 2025, drawn from its own card and bill-pay data, even as returns lag. In a Deloitte survey Ramp cites, 87% of finance chiefs called AI very or extremely important, while only 21% of active users said it had delivered clear, measurable value.

That gap is where Ramp sees this role making an impact. Finance work depends on context that is often buried across systems: vendor history, contract terms, policy exceptions, approval chains, and prior decisions. Off-the-shelf finance AI tools often struggle because they lack that context, or because they break once they touch legacy production workflows.

“In finance, every decision depends on buried layers of context: the policy, the vendor, the contract, the approval chain, and the exception history,” said Ori Daniel, head of AI solutions at Ramp, said in a statement. “Applied AI Solutions helps enterprises capture that context and turn it into agents that can complete work safely, with the controls finance teams need.”

Those controls are the load-bearing part. Ramp says the offering keeps humans in the loop through approval paths, audit trails, and human review for high-risk decisions. That is essential in finance, where an agent that acts too freely is not a productivity tool. It is a governance problem.

The launch fits a pattern. Ramp spent the spring shipping agents across procurement and accounting and even built a credit card for AI agents to use. It also runs on the model it is now selling. Ramp says its own finance team operates with a fraction of the headcount you would expect, with agents handling capital planning, variance analysis, board reporting, and the close.

“Just about every CFO has been promised AI will change everything and most still have nothing to show for it.” said Daniels to the The New Stack. “We didn’t dream our AI Solutions up in a lab to sell; we run our own books on these agents with a fraction of the office of the CFO headcount, and now we send our engineers in to do the same thing inside your finance team.” 

It also fits a broader shift. The forward deployed engineer has moved from a Palantir-era curiosity to one of the main ways the biggest AI labs get their technology adopted by large customers. OpenAI and Anthropic have built dedicated teams around the model. Ramp is now betting the same role can help enterprises deploy AI inside legacy finance systems faster than software alone can.

It is becoming clear that the next generation of enterprise AI will not be won by models alone. It will be won by teams that can make models work inside the messy systems where companies actually run.

YOUTUBE.COM/THENEWSTACK

Tech moves fast, don't miss an episode. Subscribe to our YouTube channel to stream all our podcasts, interviews, demos, and more.

Created with Sketch.