You won't hire your way to scale — the market is thin, and even a full roster wouldn't carry the domain knowledge that makes the work useful. The stronger move is often the person already sitting inside the business.
There is a reflex on every new AI program to solve the people problem by hiring: post the roles, compete for scarce specialists, staff up. It rarely works on its own. The market for AI talent is expensive and thin, and even if you could hire everyone you need, you would not get the thing that makes the work useful — deep knowledge of how your business runs.
Hire the core that is genuinely hard to build internally — the platform and ML engineering that has to be strong from the start. Upskill the people you already have, which is usually cheaper, stickier, and keeps the domain knowledge you cannot buy. Partner for surges of demand you cannot staff permanently. Most programs over-index on the first and neglect the second, which is exactly backwards for durability.
Here is the move that quietly matters most. It is usually easier to teach an experienced insider the data skills than to teach a data specialist your business.
The insider already carries the scarce ingredient. Every company has them, often hidden — the analyst buried in finance who quietly runs half the company's reporting, the engineer who does in Python what everyone else does by hand. Find these people early; they are your first recruits and your first allies. Give someone who already knows how the operation works the technical skills, and you have something rare and hard to replicate. Hire a brilliant outsider and you still have to teach them the business — which takes longer and never fully takes.
A specialist who doesn't know your business builds clever, useless work. An insider you've taught builds the thing people adopt.
Individual upskilling doesn't scale on its own, so build the supporting frame. A data and AI academy to train people in cohorts rather than one at a time. A safe path for capable non-specialists to do their own analysis without creating a mess. And literacy for the leaders who fund and challenge the work — because if executives cannot read a model's output or push on it sensibly, the culture will not shift no matter how good the technology is.
This is also why the embedded domain expert is non-negotiable on every delivery team. The person who knows how the business actually works is the difference between a model that is technically correct and one that is useful. Upskilling is how you grow more of those people instead of forever renting them.
Step back and the point of all this is the exit. A program staffed entirely by hired outsiders stalls the moment those people leave. A program that has taught its own insiders keeps running, because the capability now lives in people who were always going to stay.
So treat talent the way you treat the rest of the office: build it to outlast you. Hire the core, teach the insiders the rest, partner for the surges — and measure success not by how many specialists you recruited, but by whether the team can run the next use case, and the fortieth, without you.
Bring us where you're stuck — a mandate, a stalled pilot, or the whole build. We'll tell you where we'd start.