Build the team or borrow it
The most expensive AI hiring mistake is treating org design as a hiring problem instead of a function problem. Three functions have to happen — direct, redesign, build — and the question for each is whether you build it internally or borrow it. Score each factor below, then read the band.
The three functions
Before you score anything, confirm which function you're actually trying to fill. Conflating them into a single hire is the error.
| Function | What it means | The wrong version of it |
|---|---|---|
| Direction (vCAIO) | Decide which bet goes first, what the firm won't do, and what gets killed when it isn't working | A $300K Head of AI who writes strategy decks but has no authority to change how a workflow runs |
| Redesign (operations leader) | Take a broken workflow apart and rebuild it around what AI can now do — with real authority to change the work | A technologist without an operational mandate, or a strategy deck nobody implements |
| Build (AI Engineer) | Wire the model to your systems, build the tools it calls, set up evals, handle exception paths | Hiring an ML Engineer (builds/trains models from data) when you need an AI Engineer (builds with existing models via APIs) |
Which function are you evaluating right now? ☐ Direction (vCAIO) ☐ Redesign ☐ Build (AI Engineer)
Scored criteria — rate each factor 1 (low) to 3 (high)
| Factor | Guidance | Score (1–3) |
|---|---|---|
| Volume of AI decisions | 1 = a handful of bounded projects; 3 = continuous backlog, agents proliferating | |
| Stakes of a wrong call | 1 = low-cost to correct; 3 = a bad decision costs a client, a lawsuit, or significant revenue | |
| Ability to keep the role busy | 1 = a few days a month of real work; 3 = full-time work with clear growth path | |
| Bench depth (not a single point of failure) | 1 = hiring one person into isolation; 3 = hiring into a team of peers | |
| Speed to production matters | 1 = can wait for the right hire; 3 = need to be in production this quarter | |
| Long-term internal capability goal | 1 = fine to rent this forever; 3 = must own this function as a strategic asset | |
| Authority to redesign work (for the redesign function only) | 1 = no internal candidate has the operational mandate; 3 = clear internal owner with real authority over the workflow |
Total score: _______ / 21
Recommendation bands
| Score | Indication | Typical right move |
|---|---|---|
| 15–21 | High volume, high stakes, real bench — building in-house is justified | Hire, but hire into a function (not into isolation). Define the role precisely before writing a job description. |
| 9–14 | Mixed signals — some factors favor building, others favor borrowing | Borrow to start; set a 6-month milestone to re-score. Use the partner period to learn what you actually need before hiring. |
| 7–8 | Not enough volume, stakes, or bench — full-time hire is premature | Borrow the function. For Direction: fractional vCAIO. For Build: implementation partner. For Redesign: own the authority internally, borrow facilitation if needed. |
Hard rules from the chapter
- ☐ Redesign is always built internally. You cannot outsource the authority to change how your own workflows run. Borrow the facilitation; own the function.
- ☐ Direction is usually borrowed at first. A 200-person firm does not have enough AI decisions to keep a $300K executive busy. The fractional vCAIO exists for this reason.
- ☐ Build goes either way — hiring is slower and riskier in this talent market; a partner gets you to production faster but you're renting the capability. Neither is wrong; defaulting to a full-time hire because it feels more serious is.
- ☐ Hire AI Engineer, not ML Engineer, for mid-market work. You are building with existing models via APIs, not training your own. The titles get confused constantly; the wrong one is more expensive and less useful.
- ☐ Whoever builds it must still be around when it breaks. A production agent with no owner is a liability waiting for a bad Tuesday.
Want a second set of eyes on this in your firm? The no-sell promise applies — if it isn't a fit, I'll tell you in the first ten minutes.
Book a 30-Minute Call →