We are building an AI implementation business. This guide explains our proposed selection criteria; it is not an independent ranking.
Ask for evidence that resembles your problem.
A polished demo is useful, but ask whether it uses sample data or a customer workflow. Request the baseline, measurement period, sample size, review effort and limitations behind any performance claim.
Find out who does the work.
Understand who scopes the project, builds integrations, tests outputs, trains staff and provides support. Clarify any partner or subcontractor responsibilities.
Inspect the difficult cases.
What happens with incomplete documents, contradictory records, unexpected customer requests or a failed integration? Ask to see the review queue and recovery process.
Make ownership and costs explicit.
Confirm code and data ownership, infrastructure access, third-party licences, ongoing fees, documentation and the ability to move to another provider.
Agree on success before starting.
The proposal should define the workflow, exclusions, acceptance criteria, dependencies, timing assumptions and the decision after the pilot.
A practical shortlist checklist
- Relevant, verifiable project evidence.
- A named delivery and support owner.
- A measurable pilot with clear boundaries.
- A review and fallback process.
- Transparent build and running costs.
- Documented data handling and exit arrangements.