The Structural Gap
In 2026, AI labs (OpenAI, Anthropic, Palantir, Microsoft) employ Forward Deployed Engineers (FDEs). These are senior engineers who embed in customer accounts, build the integration, and stay until production-ready.
The economics: FDEs cost $350K-$550K total comp. Lab pricing makes them viable for $40M+ contracts. They concentrate roughly 76% in regulated whale accounts (banks, government, healthcare, insurance).
The result: AI labs created visible demand for "embedded AI engineering expertise" but cannot serve 95% of the market they generated demand for. SMBs and mid-market companies need the same capability the labs sell to Fortune 500. They cannot afford the lab version.
The gap exists. Someone has to fill it.
The Forward Deployed Studio
A Forward Deployed Studio is a small studio (2-10 people) that operates like a Forward Deployed Engineer team but at SMB and mid-market price points.
Three structural differences from traditional agencies:
1. Production code as primary output. No staging environments controlled by the studio. Code lives in the buyer's repo from day one.
2. Integrated brand layer. The AI feature ships with the launch surface around it. One team, one engagement.
3. Vendor neutrality by design. A studio that does not sell tokens is free to pick the right model per task.
This matches the actual structural difference between FDEs at labs and FDEs at studios: labs sell tokens (so model selection has a thumb on the scale), studios sell delivery (so model selection is task-driven).
The Economics
A Kastling Embedded Pod is typically:
- 2-4 person team
- 3-6 month engagement
- Rolling priorities with weekly business reviews
A full Embedded Pod engagement lands in the same range as one mid-band AI-lab FDE's annual compensation. For that, the buyer gets a multidisciplinary team (engineer + designer + PM + brand) instead of a single specialist.
For an SMB doing a serious AI rollout, this is closer to a 12-month FTE equivalent at fully-loaded cost. For a mid-market company, it's a fraction of what a dedicated AI hire would cost in their first year.
When To Use It
The Embedded Pod is the right shape when:
- The work needs a team in the room (literally or via dedicated async time)
- The scope crosses multiple workstreams (build + brand + launch + operate)
- The engagement is 3-6 months minimum
- Integration with existing systems is complex
- The buyer wants single-throat-to-choke accountability
It's wrong when:
- The scope is one specific deliverable (use Project instead)
- The buyer needs ongoing capacity with multiple rotating priorities (use Partnership)
- The budget is below $200K total
What This Doesn't Mean
The Embedded Pod is not consulting. The team writes code, ships features, owns the delivery. The brand work isn't a deck. It's a launched website, a delivered video, a printed pitch deck.
The Embedded Pod is not staff augmentation. The team operates as a unit with internal coordination, not as individual contractors plugged into the buyer's existing team.
The Embedded Pod is not a longer Project. The structure is different: rolling priorities, weekly business reviews, a defined exit ramp, full IP transfer at engagement end.
How To Evaluate Embedded Pod Vendors
The same disqualifying questions from the vendor evaluation checklist apply, with three additions specific to embedded engagements:
- What does a typical week look like? Vendors should be able to describe a concrete weekly cadence (standups, business review, sprint planning, etc.).
- Who is the single point of accountability? Embedded engagements need one named lead, not a committee.
- What is the exit ramp? When does the engagement end, what does handoff look like, who owns the operational responsibility after?
If a vendor can't answer these, the pod won't be embedded. It will be billed-by-the-hour staff aug.
The Honest Read
The Forward Deployed Studio model exists because the AI lab FDE model created a structural gap. The labs are not going to fill it. Traditional agencies aren't built for it. Offshore outsourcing doesn't have the integrated brand layer.
Studios that combine real engineering, real design, real strategy, and real launch capability in one engagement are the answer. We built Kastling deliberately as one. The Embedded Pod is how we deliver it.