Choosing an AI consultancy is hard because every firm now says the same things: strategy, implementation, responsible AI, proven methodology. The way through is to ask questions whose answers cannot be marketing. Here are the ten we would ask, with our own answers, so that comparing is easy.

1. Can we see the method, not just its name? Every consultancy names a framework. Ask to see the artefacts: the scorecard, the templates, a redacted output. Ours is published in full under our approach pages: the Converge Methodology, the Opportunity Pyramid, the four levels of AI, the adoption pathways and the convergence scorecard.

2. Who will actually be in the room, and who writes the code? Rates are quoted at senior level and delivered at junior level more often than anyone admits. Ours: Dr Paul Gibson leads the consultancy and the sessions; Ian Lawton leads engineering; the engineers are the group’s own, in York. Nobody is subcontracted.

3. What do we own at the end? For consultancy: every idea, every score, every use case, in editable form, usable without you. For a build: the code, the evaluation set and the documentation, with no licence and no per-seat fee. That is our answer, and it should be any consultancy’s.

4. What happens after the prototype? The gap between a working pilot and a running system is where most AI spend disappears. Ask for the production pilot to be in the quote, and for the handover or managed-service cost to be stated before the build starts. Ours is.

5. How will we know it works? If the answer is “you’ll see”, walk away. Ours: an evaluation set of real cases with agreed outcomes, built with your experts before build, run on every change, with the results reported. For consultancy, a roadmap with a first quarter you can start on Monday.

6. What do you do when the exciting idea is not the ready one? A consultancy that tells you every idea is a go is selling days. Ours sequences by readiness, assigns a data position to every use case, and tells you in writing which opportunities should be bought rather than built and which should wait for data.

7. Which vendors are you tied to? Reseller margin and partner quotas bias recommendations. We are platform-agnostic, build on Claude, OpenAI and open-weight models, and take no margin on any licence.

8. Can you name clients and show results? Logo walls and anonymised cases are not evidence. Ours: Nidec Drives, Immediate Media and Agency51, with case studies that say what was done and what came of it, and a clear line between the sectors where we have delivered AI and those where we bring the group’s wider experience.

9. What are your credentials and your delivery history? AI expertise without delivery experience produces slides; delivery experience without AI expertise produces the wrong system. Ours: a PhD in Artificial Intelligence leading the consultancy, and a group that has built and integrated enterprise platforms for Siemens, Nestlé, Hearst, Krispy Kreme and West Ham United since 2006.

10. When is the price fixed? Ask for every engagement to be scoped and the price fixed before it starts, and for the production pilot and the handover or managed service to be in the quote rather than discovered later. Ours are: scoped first, then fixed, with nothing priced during delivery.

Ask the same ten of anyone else, and compare the answers, not the brochures. Ours are on the services and about pages.