Build vs. Buy vs. Embed

It depends on your timeline and expertise. Building in-house gives maximum control but takes 6-12 months to hire and ramp an AI team, and you're building institutional knowledge from scratch. Outsourcing gets you started faster but creates vendor dependency. The embed model is the third option: an external team operates inside yours, builds production systems, and transfers knowledge until you can operate independently.

An embed partner operates as part of your team, not as an external vendor. They use your tools, your repos, your comms channels. They build production systems alongside your people, transferring knowledge continuously. When the engagement ends, your team has the skills and the systems to operate independently. It's not outsourcing; it's temporary capability injection.

A minimal AI engineering team (2-3 people) costs $400K-$700K annually in salary alone, plus 6-12 months of ramp-up time before they're productive on your specific domain. An embed engagement delivers production systems in 4 weeks at a fraction of the annual team cost, and you can hire the team afterward with the knowledge already captured in Business-as-Code.

Two things: lock-in and generalization. Platforms lock you into their architecture, pricing, and roadmap. And they generalize, they build for every industry, not your industry. Your business-specific processes, rules, and knowledge can't be captured by a generic platform. For production AI that operates on your business, you need custom implementation, whether in-house or embedded.

That's exactly the model. The embed engagement has a designed end (Escape Velocity) where your team can operate independently. Business-as-Code artifacts serve as the institutional knowledge base. Many clients hire AI engineers after the engagement, who ramp faster because the context is already captured and operational.