"AI consulting" describes an enormous range of things in 2026 — everything from a single person repackaging ChatGPT prompts into a retainer, to firms that genuinely build and ship production automation. Both use nearly identical marketing language. The questions below are the ones that actually separate them, and we think any firm — including us — should be able to answer all eight without hedging.
1. What have you actually shipped, not just advised on?
"AI strategy" and "AI implementation" are different businesses. A firm that only produces recommendations has a fundamentally different (and lower) bar to clear than one whose work has to actually run in production and keep working.
2. Can I see real work, not just a case study writeup?
A polished case study PDF is marketing copy. A live product you can sign into, or an open-source repo you can read, is evidence. If a firm's strongest proof is prose, ask why there's nothing you can independently verify yourself.
3. Who sees our code and data, and can we control that?
Any serious AI engagement touching your codebase should have a clear answer here — not "we take security seriously," but specifics: what model provider is involved, whether you can bring your own API key so your code goes straight from your infrastructure to your chosen provider, and what's stored versus processed transiently.
4. Does anything auto-merge, or is there always a review gate?
This is the single question that tells you the most about how a firm thinks about risk. Automation that opens a PR for your team to review is a fundamentally different risk profile than automation with unsupervised write access to your repository.
5. Is this a wrapper around a single AI API, or a real integration into our workflow?
"Real integration" means it hooks into the tools you already use — your issue tracker, your CI, your version control — and takes action there. A chat window bolted onto your docs is not that, no matter how good the model behind it is.
6. What happens to what's been built if the engagement doesn't work out?
You should own the code, prompts, and workflows produced during the engagement, full stop. If continuing to use what was built requires staying locked into the firm's proprietary tooling, that's a lock-in risk worth pricing in before you start, not after.
7. How is pricing structured, and when do you find out the number?
Hourly, retainer, and fixed-scope engagements all have reasonable use cases — but you should know which one you're getting into, and roughly when pricing gets discussed, before you've invested hours in a sales process.
8. Who is this a good fit for — and who isn't it for?
A firm that can articulate who they're wrong for is telling you something real about their positioning. A firm that claims to be right for everyone usually isn't especially right for anyone.
We built our own consulting FAQ around these same eight questions, because we'd rather a prospective client read our honest answers before a call than discover the gaps on one.