Bring the engineer
Part four: forward deployed engineers and the next firms to get bought
Part one showed the platform-led deal is gone. Part two showed the platforms buying their AI from startups. In part three, Salesforce itself said the screens are optional. So where is consulting growing?
Next to the model companies. The firms that work hand in hand with OpenAI, Anthropic and Palantir are hiring as fast as they can, and they deliver differently.
Forward deployed
The model is the forward deployed engineer, a term Palantir made famous. An engineer sits with the client, builds against the client’s real problem and ships. Next to that engineer sits someone who understands the business process cold.
A friend who runs a large practice told me they cannot hire those engineers fast enough. Every firm I talk to says the same thing.
Remember Amodei’s number from part three: companies capture maybe five or ten percent of what the models can already do. These engineers are how the rest gets captured.
The deals are getting bigger while the teams get smaller.
Small, targeted teams, with some of the fee at risk on the outcome. That is a very different firm from the one I grew up in.
Who else pays
The money around the deal has moved too.
The hyperscalers will fund part of an implementation if the client commits to more compute. The payments companies do the same to win the flow. A startup that sells through a client’s existing cloud marketplace contract skips months of procurement and retires part of the client’s committed spend in the same move.
That same friend has been coaching a few young founders through this, for free. Every one of them was surprised to hear it.
Tokens are the rough edge. Every model is priced differently and the prices keep moving, so who carries the compute is the hardest line in the contract. That will settle. A year ago you could do very little for a lot of tokens; now you can do a lot for very few.
Depth, spread wider
The scarce input is not the engineer. It is the person who has spent thirty years inside how a distributor actually takes an order.
One friend tells junior partners that their kids will still be implementing digital systems for B2B companies, because there is that much left to do. Industrial companies and distributors are still far behind.
A person with that kind of depth used to advise one team at a time. Paired with a handful of forward deployed engineers, the same person can reach many teams at once.
Thirty years of knowing where it hurts is worth more now than it has ever been.
The firms that get bought next
Crown was tech-enabled by other people’s platforms, not by our own technology. The firms I am looking at now, $10 million to $20 million in revenue, can be tech-enabled for real: their own layer, their own agents, small teams doing very good work.
They will not compete with Deloitte on a $50 million transformation. They will own the mid-market, and the mid-market is enormous.
What they usually lack is the part I learned the hard way. The founders are very smart and have never sold to an enterprise. They do not know how long it takes or who has to say yes. Growing from 100 people to 500 is harder than picking the platform to bet on, and that is the work I have spent my career on.
At Bambu Capital this is the Services-Tech thesis we are building now: B2B firms that sit next to one of the model companies, deliver with forward deployed engineers, and get bought by the firms that cannot hire those engineers fast enough.
So I owe the friend from part one twice now. Once for Crown, and once for the best hour on this market I have had all year.
The platform used to bring the deal.
The client brings the problem.
The firms that win bring the engineer.
A four part series on where B2B consulting goes next.
One. The lead used to come from the platform. Crown, Razorfish, and why the platform-led deal is gone.
Two. The platforms are buying their AI. Software companies are buying small AI firms instead of building, and somebody still has to make it work.
Three. Salesforce says the UI is going away. The Dreamforce keynote gave up the screens, kept the data and partnered for the intelligence. What that does to the thesis.
Four. Bring the engineer. You are here. The consultancies growing fastest sit next to OpenAI, Anthropic and Palantir, and send forward deployed engineers.