The question behind the question
“Contractor or consultancy?” is usually asked as a question about cost. It is really a question about responsibility: who directs the work, who carries the risk if it does not deliver, and who owns the result afterwards. Get those right and the commercial model follows. Get them wrong and no rate is cheap enough.
The six questions in the tool test the things that decide it:
- Direction. An individual contractor is effective when someone in your team directs the work. Without that, the supplier needs to own the plan as well as the doing.
- Definition. A clear outcome with acceptance criteria can be bought as scoped delivery. An outcome that is still forming cannot, yet.
- Breadth. One senior person can cover a lot of AI engineering. Several specialists working at once need coordination, which is what firms sell.
- Duration. Short needs suit a contract or a bounded project; long or ongoing needs suggest planning for ownership to move in-house.
- Ownership afterwards. Your existing team, a team still being hired, nobody yet, or the supplier on an ongoing basis.
- Budget model. Some organisations can only fund fixed prices; others prefer day rates.
How the recommendation is made
Each answer adds points to one or more models. The highest total is recommended, with the answers that supported it, and any that counted against it. When the top two are within a point, the tool says it is a close call. It always shows all four models with a fair note on each, because the right choice often depends on something six questions cannot capture.
Illustrative example
Illustrative example, not a client engagement. A product company’s CTO has an AI feature that a departed engineer prototyped. Her answers: a tech lead can direct the work partly; the outcome is roughly defined; one senior person could cover it; three to six months; the team that will own it is still being hired; either budget model works.
The tool recommends hybrid: contractor now, scoped delivery later, two points ahead of the independent contractor. The reasons: direction is available but thin, the outcome needs sharpening before it can be fixed, and the owning team is still arriving. In practice that means a few weeks of embedded work to measure where the prototype fails, followed by a scoped delivery with acceptance criteria written from what was learned. The note on hybrid warns about the main risk: setting a date for the switch, so it does not drift into an open-ended contract.
Assumptions and limitations
This is a rule of thumb, not procurement advice. It does not know your procurement rules, risk appetite, internal politics or the specific suppliers available to you. It does not compare firms or quotes. The scoring favours simpler commitments when options tie, on the basis that an engagement that is easy to end is easier to correct.
What to do next
For an individual contractor, define the role with the contract brief builder and compare quotes with the contract cost model; the roles I take on are under contract AI engineering. For scoped delivery, write the scope with the AI project scope template; outcomes I deliver are listed under services. How each model is priced is on engagement models and pricing.
If the real choice is between training a team and changing how it works, that is a different decision: see AI training versus enablement.