Your Intelligence Heist: When Help Isn't Actually Help

The inside man is already in your building. Last week, a service provider emailed me offering to help find Forward Deployment Engineers. I almost forwarded it to my head of talent. Then I looked up what the role actually was. Every major technology wave creates a new job title. Cloud computing gave us cloud architects. Continuous deployment gave us DevOps engineers. Mobile developers didn't exist as a profession in 2005 and were everywhere by 2012. Security complexity has built an entire category of security architects and consultants. Each of those roles followed the same pattern. A new technology emerged. It was hard to adopt. Vendors sent in specialists. Businesses got a faster time-to-value. Vendors got stickier customers. Everyone went home. I hired a lot of those people over the years — vendor-employed, independent consultants, eventual full-time hires. Genuinely useful, most of them. And every single one worked for an employer who wanted the same thing: your spend, not your market. AWS wanted your compute bill. HashiCorp wanted your infrastructure tooling spend. Mobile agencies wanted retainers. Security vendors wanted your seats. None of them was in the business of building the product that competed with you. Their success was your adoption. Not your decline. Forward Deployment Engineers are different. What looks like help is a heist. Palantir invented the FDE model over a decade ago — engineers embedded inside client organizations to accelerate adoption. It worked. Now every major AI provider is scaling the same model, with more money behind it than Palantir ever had. Amazon just committed $1 billion to a new internal organization of AI-focused engineers embedded inside enterprise clients. OpenAI raised over $4 billion for “The Deployment Company” — a joint venture with TPG, Bain Capital, and Brookfield. Anthropic built a $1.5 billion joint venture with Blackstone and Goldman Sachs. Microsoft announced a $2.5 billion “Frontier Company” to embed engineers inside customers. Google committed $750 million. Over $10 billion. Five providers. Same bet. That’s not a service offering. That’s a heist budget. When I added up that number, I stopped thinking about law firms for a moment and thought about my own company. As a software CEO, I carry two existential threats at all times: my customers might decide to build it themselves, or my AI model provider might decide to own my category. Most days, I worry about the first one. The FDE is how the second one walks through your front door. There’s a version of the API conversation worth having here. When you use an AI provider’s API — every query you run through it — you’re choosing to send data out. Opted in. Something leaves the building because you decided it should. An FDE is different. You’re inviting someone in. And then walking them through the vault. Embedded inside your firm, they’re observing which workflows create value, which tasks consume the most time, where billing concentrates, what clients actually need versus what your firm delivers. That gap — the blueprint. And it goes back to an employer whose business model is built on knowing exactly where gaps large enough to automate actually are. On April 17, 2026, Anthropic launched Claude Design — a tool that generates working prototypes from natural language. The same day, Figma’s stock dropped 7%. One announcement. One company that had never competed with Figma. And the market immediately understood. Harvey AI — built for legal — is now valued at $11 billion. $190 million in annual recurring revenue. Over 100,000 lawyers across 1,300 organizations, 25,000 custom agents running legal workflows — M&A, due diligence, contract drafting, document review. Thomson Reuters reports that $143 billion in U.S. legal and accounting revenue is now under active reconsideration based on AI delivery. Legal software built over decades is being rebuilt in three years. Not by law firms. By AI providers who learned enough about legal workflows to build the product themselves. Here’s where the McKinsey objection usually lands. McKinsey has sent consultants inside firms for sixty years. They’ve learned your processes, sold those insights across clients, and generated enormous returns. Law firms are still standing. Accounting firms are still standing. So why is this different? Because McKinsey doesn’t build software. McKinsey’s output is advice. When an engagement ends, you’re still the service provider. Work still requires humans. Category survives. An AI provider’s output is a product — and the product scales without you. Microsoft didn’t become an accounting firm after building Excel. Google didn’t become a law firm after indexing legal documents. Those were tools for the industry. What’s being built now is the next layer of the industry itself. That’s what makes this different from every tech consultant who came before. I’m still working out exactly where my own line is. But one question clarifies it quickly. Who does this person work for, and what does their employer build? Cloud provider charging for compute — proceed. Business intelligence doesn’t register in their model. AI provider actively building software in your category — stop. You’re not hiring a consultant. You’re showing a potential competitor around. Hire your own AI engineers. People whose career depends on your success, not on mapping your workflows to a product roadmap somewhere else. Your intelligence. Your workflows. Your client relationships. That’s not data to be mapped. That’s the company. Their job is to understand all of that. Ask yourself who benefits when they do. Every tech consultant you’ve ever hired worked for a company that wanted your money. FDEs work for companies that want your market.