The Engineer in the Building: Why MSPs Are Built to Win the Forward Deployed Era

The AI labs proved the embedded-engineer model works. The version that reaches small businesses belongs to the MSP.

By RocketHub Content Studio · 2026-10-26

The AI labs just spent a fortune proving that the most valuable engineer is the one inside the client's building. MSPs have been that engineer for twenty years. Now they need to claim the AI work before someone else walks through the door.

Kyle Stockwell — CGO, RocketHub.ai

The labs just validated the MSP model

For twenty years, MSPs have sold one thing above everything else: a trusted technical partner who knows the client's business from the inside.

This year, the biggest names in AI reached the same conclusion. OpenAI and Anthropic are building teams of forward deployed engineers. These are vendor engineers who sit inside the customer and build against its real systems until the work runs. Their answer to making AI work was simple: put an engineer in the building.

MSPs have been that engineer since before most AI labs existed.

What we see from inside other people's companies

At RocketHub.ai, Atlas runs AI agents inside other people's businesses, in their real inboxes, calendars, files and systems. Here is what those agents find.

None of those came from a clever model on its own. Each came from an agent inside the business, with access to the right systems and a clear sense of who to tell. Agents do their best work where someone already understands the client's systems, permissions and people. That is a description of an MSP.

The money is moving fast

Anthropic has committed $100M to train 10,000 engineers for deployment work through Claude Frontier Academy. Business Insider reported in May that job postings for forward deployed engineers grew more than 700% in a single year.

700%+ growth in forward deployed engineer job postings in one year (Business Insider, May 2026)

The lab version costs more than $200,000 a quarter. Your clients will never see it.

Here is how the AI labs sell it. The vendor places its own engineer inside a large customer to build AI for one specific job, and keeps that engineer there until the work runs. Gartner's research, as reported by Channel Dive, puts that at more than $200,000 a quarter for each job.

Even large enterprises are walking away from it. Gartner predicts that 70% of enterprises will abandon agentic AI built by vendor forward deployed engineering teams by 2028, citing cost and dependency.

70% of enterprises will abandon agentic AI built by vendor forward deployed engineering teams by 2028 (Gartner, as reported by Channel Dive)

So the idea works: an engineer inside the business, building on its real systems. What breaks it is the price and the dependency on the vendor. A 15-person accounting firm will never get that engineer. A 40-person manufacturer could never justify one.

But those businesses already have an engineer in the building who knows their systems and isn't leaving for the next account: their MSP.

What the MSP already has

What many MSPs are still building is AI build capacity. The Futurum Group's research, reported by Channel Dive, found that 71% of partners already sell AI software. Yet only 48% of AI-confident partners report robust AI revenue. That is a capacity gap, not a trust gap, and capacity is the easier one to close.

The questions worth asking

I don't think any one company has the answer for every MSP, and I'm curious how the best ones are approaching this.

If you've answered some of these, I'd love to hear how.

A pivot like this comes along rarely

MSPs got one chance to redefine what they sell when businesses moved to the cloud. They got another when security became a board-level worry. Both times, the providers who leaned in early became something bigger.

This one is larger. The labs have proven the model. The enterprise version is expensive, and Gartner expects most of it to be abandoned. The small-business version doesn't exist at scale yet. The MSP is already standing where it needs to be built: inside the building, trusted with the keys, known by name. The engineer in the building should be the MSP.

That is why we built Atlas at RocketHub.ai. In the hands of an MSP, Atlas becomes the forward deployed team the client already trusts. Take the 15-minute interview and get a preview of the integrated workforce Atlas would build for your clients.

Take the interview and preview your integrated workforce

Sources

All posts