The hire-first pattern.
The conversation often starts like this: the CEO has read enough about AI to be sure the operation needs to "do AI." They post a Head of AI role at $185K-$280K loaded. They hire a strong engineer in three to six months. The engineer joins, ramps for two months, ships the first thing in months four through six.
Then it gets interesting. The engineer finishes the first thing and looks for the next thing. There's no scoped second project; the operation thought the first was the project. The engineer maintains the first build for three months, builds a small second thing, maintains both, and at month fifteen starts looking for a job somewhere with more surface area.
This pattern shows up at firms that hired AI talent in 2023-2024 thinking they were ahead of the curve. They were ahead. They also burned through six to nine engineering hires across the segment trying to find the right fit, often because the issue wasn't the candidate but the queue.
The commission-first pattern.
The alternative sequence: commission the first build. The boutique scopes against the operation's biggest constraint, ships in 4-7 weeks, hands off code the operation owns. The build runs in production. The senior staff who shepherded it become AI-fluent through the experience.
By month four, the AI-fluent senior is naturally identifying the second workflow that would benefit. By month seven, the second build has shipped (commissioned again, or built internally if the operation has the bench). By month nine, a real queue exists: three workflows shipped, three more visible, the cultural muscle memory in place.
That queue is what justifies the Head of AI hire. The hire arrives at month nine to a firm with three production systems, a clear roadmap of three more, and a senior operator (the AI-fluent one) who can be the hire's day-one collaborator. The hire ships their first deliverable in month two of joining instead of month six, because the infrastructure is already there.
Why the sequence matters.
The hire-first sequence creates a chicken-and-egg problem. The Head of AI is brought in to figure out what to build; the operation hasn't built anything yet so doesn't know what to ask for; the engineer's first six months are exploration that often surfaces the wrong things to build because the operation hasn't yet learned to think about its workflows the right way.
The commission-first sequence inverts this. The first commissioned build is itself a learning vehicle. The operation learns what AI does well, what guardrails feel right, what change-management patterns work in their culture, what data shape they actually have. By the time the Head of AI joins, all of this is settled. The hire's first project lands in production fast because the operation has internalized the right questions.
The economics, candidly.
One year of a Head of AI hire's loaded cost is $185K-$280K. That's roughly two to three commissioned builds. Two years is $370K-$560K, four to six commissioned builds.
If your business has four to six distinct AI builds queued for the next two years, the math favors the hire. If you have one or two, the math favors commissioning. Most $8M-$50M firms have one or two clearly-named workflows in their first 12 months of AI work, with three to four more becoming visible after the first build ships.
This is why the sequence matters. Commission-first reveals the queue. Hire-first hopes the queue exists.
What we'd do.
If your business has one named workflow with a dollar figure attached: commission. The playbook library describes 19 specific architectures we'd ship across five verticals.
If your business has 4-6 named workflows already scoped and a culture that has absorbed AI fluency in its senior staff: hire. The Head of AI joins to a real, justified queue.
If your business has neither yet: don't hire and don't commission. Take the AI-Ready Course first, run the diagnostic, find the first workflow. Then decide.
The honest read.
This memo is self-interested. Of course we'd recommend commissioning before hiring; we are the boutique. The honest defense: across the firms we've audited, the sequence we describe holds even when we're not the boutique commissioned. Firms that commission first (with us, with another boutique, with their existing dev shop) generally end up with stronger AI capability at month 18 than firms that hire first. The sequence is the leverage; the boutique-vs-not-boutique is downstream of that.