The short answer
Partly, yes: outside experts embedded in client organizations is as old as consulting itself, and plenty of ordinary consulting now wears the forward-deployed label. What is genuinely new is the deliverable. A consulting engagement ends with recommendations. A forward-deployed engagement ends with a production system the client owns. Judge any provider by which of those they leave behind.
Spend time in engineering forums and you will meet the sharpest version of the charge: that calling consulting revenue "forward-deployed engineering" is the greatest rebrand in enterprise software. We sell the model, and we think the skeptics deserve a straight answer rather than a brochure.
The charge
The argument goes like this. Companies have embedded outside experts since the invention of the invoice. Palantir's innovation was operational, not conceptual, and what the AI industry did in 2025 and 2026 was notice that "consulting" prices like the past while "forward-deployed engineering" prices like the future. Same staffing model, new sticker, billions of dollars of it.
What the skeptics get right
- –The label is doing heavy lifting. When a term gets hot, everything migrates toward it. Staff augmentation, bodyshopping, and slideware consulting are all being sold as forward-deployed engineering right now. The costume is real and common.
- –Embedding is not new. On-site experts, secondments, integrated delivery teams: consulting firms have run every variant for decades. Presence alone differentiates nothing.
- –The incentive to blur is structural. Demand for the role is exploding faster than the supply of people who can actually do it, which guarantees that the gap gets filled with relabeled generalists.
When a label gets valuable faster than the skill it names, most of what wears the label is costume.
What is actually different
Strip the branding from both models and compare what is contractually true at the end of an engagement.
Consulting's unit of delivery is judgment: analysis, options, recommendations. That is not an insult. Good judgment is scarce and valuable. But the engagement is complete when the judgment is delivered, which is why the deployment gap survives every consulting engagement that was pointed at it. Somebody still has to build the thing, run it against real data, and carry the pager.
Forward-deployed engineering's unit of delivery is a running system. The engagement is not complete at the recommendation, the demo, or even the launch. It is complete when the system is in production and ownership of it has moved: code, prompts, infrastructure, and evals in the client's accounts, documented well enough that the vendor becomes optional. The definitional details are in what is forward-deployed AI engineering, but the distinction fits in one line: consulting transfers advice, forward-deployed engineering transfers a working asset.
The test that settles it
Which means the rebrand question has an empirical answer, engagement by engagement. Ask one question of any provider wearing the label: what exists in my accounts the day you leave?
- –A strategy deck and a roadmap: consulting. Fine, if that is what you wanted to buy.
- –A system only the vendor can operate, hosted in the vendor's accounts: consulting with hosting revenue. The worst of both.
- –A production system with real users, transferred completely, with the evals to prove it works: the discipline, whatever anyone calls it.
We hold ourselves to the third answer in public. Our own funnel runs on an agent we built, and when operating it went wrong, we published the incident rather than the brochure version. That is what the label is supposed to mean. Buyers who make vendors prove it, using questions like these, are the mechanism that keeps the term honest.