When fractional makes sense
You have 30–200 people. AI decisions are landing on your desk weekly and you're guessing more than you'd like. A full-time head of AI would be six months of hiring, six figures of comp, and overkill for the current surface area. Fractional gives you the judgment layer without the org commitment.
When it doesn't
You need engineering hands, not judgment. You want someone to "handle AI" while you look away. You expect weekly deliverables priced like a full-time role. In any of those cases, either hire or don't — fractional in the middle is where trust decays fast.
What a good engagement looks like
Monthly operating review with the leadership team. Quarterly re-scoring of the opportunity backlog. One owned initiative per quarter — something that ships, not a slide. Clear boundary on what stays in-house. Written notes after every session so continuity survives the fractional cadence.
Signs it's not working
- Every session ends with "we'll circle back" and no owner.
- The operator keeps recommending tools you'd have to buy through them.
- Your team doesn't know what the operator worked on last month.
- The roadmap changes shape every session with no cause named.
Any two of those and it's time to renegotiate scope or wind down. Fractional works on trust and cadence — when either breaks, end the engagement early, not late.