The AI consulting market has more firms than it has firms worth hiring. Many are rebranded strategy shops, agencies that added “AI” to the homepage, or teams that have built one chatbot and now sell transformation. The good ones exist, but you have to ask the right questions to find them.
Here are the twelve that matter, and what a good answer sounds like.
1. What have you actually built?
Ask for specific systems the firm has designed, built, and deployed, with the business problem each one solved. A strong answer names the industry, the use case, the technology, and the measurable result. A weak answer talks about frameworks, methodologies, and thought leadership.
Follow up: can we talk to that client?
2. Who will be on our engagement?
Consulting firms sell with senior people and deliver with junior ones. Ask for the names and backgrounds of the people who will do the work, not the people who pitched it. If the firm will not commit to specific individuals, that tells you something.
3. Where is your team based?
Location affects communication, time zone overlap, data handling, and compliance. If your data is sensitive or regulated, you need to know where it will be processed and by whom. A US-based team is not always required, but you should know what you are getting.
Our guide to US-based AI consultants and developers covers the firms that keep engineering onshore.
4. How do you decide whether AI is the right solution?
A good firm will tell you when AI is the wrong answer. Many business problems are better solved with conventional software, process changes, or better data hygiene. If a consultant has never recommended against AI, they are selling a product, not solving a problem.
5. What is your approach to data?
AI systems are only as good as the data behind them. Ask how the firm assesses data readiness, what happens if your data is not ready, and how they handle data quality issues during the engagement. Firms that skip this step build models that fail in production.
6. Build, buy, or integrate?
There are three ways to get an AI capability: build a custom model, buy a commercial product, or integrate an existing model like those from OpenAI, Anthropic, or Google into your workflow. A good firm will explain the trade-offs for your specific case rather than defaulting to the option that generates the most billable hours.
7. How do you handle model risk and governance?
Ask about bias testing, explainability, monitoring for drift, and what happens when the model produces a wrong answer. If the firm has no answer beyond “we test it,” they are not ready for regulated or customer-facing use cases.
8. What does the first 30 days look like?
A concrete answer describes discovery activities, deliverables, decision points, and who from your team needs to be involved. A vague answer means the firm has not done this enough times to have a repeatable approach.
9. How do you price, and what changes the price?
Fixed fee, time and materials, or retainer? What assumptions is the quote based on? What causes it to go up? The answer should be specific enough that you could predict the invoice.
10. What happens after deployment?
AI systems require monitoring, retraining, and maintenance. Ask who owns that, what it costs, and how knowledge is transferred to your team. If the firm’s model is to build and disappear, you will be back in the market within a year.
11. What are you not good at?
Every firm has a sweet spot. A confident firm will tell you where they are strongest and where they would refer you elsewhere. A firm that claims to do everything usually does nothing particularly well.
12. Can we start small?
A firm confident in its work will agree to a scoped pilot before a large commitment. If the only option is a six-figure engagement up front, walk away.
Red Flags
- Case studies with no named clients and no numbers
- A sales process that skips discovery and goes straight to proposal
- Pricing that requires a long-term commitment before any work is done
- No engineers in the room during sales conversations
- Heavy emphasis on a single vendor’s platform without discussion of alternatives
- Promises about accuracy or ROI before they have seen your data
Green Flags
- They ask more questions than they answer in the first meeting
- They can describe a project that did not go as planned and what they learned
- They propose a phased approach with decision points
- They have opinions about when not to use AI
- Their engineers can explain technical choices in business terms
Making the Decision
Score each firm on the twelve questions. Weight the ones that matter most for your situation. If data sensitivity is critical, weight location and governance heavily. If speed matters, weight the 30-day plan and the ability to start small.
Then talk to references. Not the ones the firm hands you, but the ones you find yourself.
If you want a starting list, our guide to the top US-based AI consultants and developers covers the firms that pass these questions. Xcelacore is on it, and we are happy to answer all twelve.
