Old Wine, New Bottle

How AI companies spent billions rediscovering something TCS / Infosys / Accenture / Thought works  figured out decades back — and why the pattern was always going to repeat.

The AI industry loves a good origin myth.

The latest one is the "Forward Deployed Engineer" — a term that takes images of elite technologists parachuting into Fortune 500 war rooms, building bespoke AI systems, bending the arc of enterprise history. 

Palantir made it famous. Now every AI company with a sales problem has one.

There's just one issue. This job is 40 years old. And in last few weeks, two of the most valuable AI companies in the world made it official. 

OpenAI launched the "OpenAI Deployment Company" — $4 billion, backed by TPG, Bain Capital, and Brookfield — explicitly to send forward-deployed engineers into enterprises.

Anthropic also launched its own $1.5 billion equivalent. 

Both dressed up in the language of transformation. 

Job Description Hasn't Changed Much

Every Generation of Complex Software Has Spawned This Layer

human layer required to bridge "works in demo" to "works in production" never goes away.

What challenge LLM Companies Face ? 

Real Reason for FDE: 6X money on the table

As per sequoia capital for ever 1$ , $6 is spent on services 

Source:https://sequoiacap.com/

OpenAI and Anthropic looked at that ratio and made a simple decision: we are currently capturing the $1. We want the $6. That's the actual strategic next step. 

Not innovation. Not a new deployment model. Pure margin expansion into an adjacent market that their own product creates demand for.

The remarkable thing is that OpenAI's DeployCo includes McKinsey, Bain & Company, and Capgemini as investors. Three of the world's most powerful consulting firms wrote checks to fund their own potential disintermediation. 

The same calculation Accenture made when it partnered with Salesforce in 2005 instead of fighting it.

Value Tier Structure Looks Familiar Too

Only structural difference: Consulting company billed the client directly for those hours. FDE costs are bundled into the SaaS ACV and called "customer success." 

The labor is identical. The accounting is different.

Why Not Just Let the AI Do It?

This is hard question for AI industry and they don't want to answer directly. If the models are as capable as advertised, why do you need humans to implement them?

Claude Code and similar tools are genuinely compressing the coding portion of implementation work. A junior FDE spending three days writing glue code is being replaced by an agent in three hours. That's real gain and every one understand this.

But the hardest part of implementation was never the code. It's knowing that the CRM data is unreliable because the sales team doesn't update it. It's navigating the political disagreement between the CFO and the CTO about who owns the AI governance question. It's being the accountable human in the room when something breaks in production.

Enterprises aren't ready to give AI agents the access, the trust, or the accountability that implementation requires. That's not a model capability problem. It's an institutional and legal problem. And institutional problems take longer to solve than technical ones.

Who is winning ?

FDE: You're doing genuinely valuable work. But you're not a new species. You're the latest instance of a role that has existed since enterprise software existed. The role matters and getting new packaging.

If you're McKinsey or Bain: You just funded the company that is supposed to make you obsolete. History will decide whether that was a savvy hedge or a very expensive mistake.