Hiring an AI consultant for a franchise restaurant is a buying decision most owners only make once or twice — which means almost no operator has the pattern recognition to spot the difference between a $5,000 engagement that delivers a working tool and a $25,000 engagement that delivers a slide deck, and the seven questions below are the ones I wish I'd been asked before I became one of those consultants myself.
Why this matters
AI consulting is having a moment. Lots of people calling themselves "AI consultants" who were doing something else 18 months ago. The price range for an "AI implementation" at a single-location franchise is currently $3,000 to $50,000 for what's often the same scope of work. The questions below help you tell the difference. Before you even start evaluating consultants, decide if you actually need one — start with the 5 AI categories that work for franchise operators. Off-the-shelf tools cover most of the ROI; consultants are for the gaps.
The 7 questions (with what good answers look like)
1. "Have you personally operated a restaurant or franchise location?"
Why ask: Restaurant operations have specific patterns (Sysco shortage processes, scratch-vs-prep ratios, peak-hour staffing, walk-in space constraints) that don't show up in tech-consultant playbooks. A consultant who hasn't worked a closing shift will design a count workflow that takes 90 minutes instead of 20.
Good answerSpecific operational stories — "I ran X location for Y years" or "I currently operate Z, here are 3 issues I deal with weekly."
Red flag"I've consulted FOR many restaurant chains" — not the same as having been responsible for one. The operator-vs-consultant difference matters for reasons I unpack here.
2. "Show me a working AI tool you've shipped to production for a customer my size."
Why ask: AI consulting has a high deliverable-to-output ratio of slide decks, audits, and PDF reports. The actual rate of consultants who have shipped working software to a small-business customer is much lower than the marketing implies.
Good answerA live URL, a customer reference, a specific feature the consultant can demo end-to-end on the call.
Red flag"I've worked on enterprise deployments at [chain X]" — ask whether they could ship to YOUR scale (single location, $99/mo operator tools, no IT department). Most enterprise AI consultants can't price-down without losing money.
3. "What's your specific failure mode? When have you been wrong, and what did it cost the customer?"
Why ask: Every consultant has shipped something that didn't work. The ones who can articulate it specifically have learned from it. The ones who deflect haven't — or they're too new to have hit a real failure yet.
Good answerA specific story — "I underestimated X, the customer ended up in Y situation, here's what I learned." Bonus: they refunded or fixed it on their dime.
Red flag"I haven't really had any failures" — this is impossible in real consulting and signals either a very thin track record or a willingness to lie.
4. "What's the smallest engagement you've done? What did the customer get?"
Why ask: Most franchise operators don't need a 6-month $50K engagement. They need a focused 2-week assessment or a specific tool built. A consultant who can't price down is a consultant who's only configured for big projects — which means yours will be padded to fit their template.
Good answerA clear small-engagement offering with a fixed scope and price ($2,500-$5,000 range is realistic for a real audit). See what an actual operator-level audit looks like for comparison.
Red flag"We don't really do projects under $X" where $X is more than your monthly tech budget.
5. "Who actually does the work — you or someone else?"
Why ask: Many consulting practices that sell to small-business customers are actually selling sales-call-senior plus subcontracted-junior delivery. The senior shows up to pitch; a less-experienced person does the work; the customer doesn't realize this until the deliverable feels generic.
Good answer"I do the work personally" or "Here's the specific person who'll be on your account, here's their background." Bonus points if they offer to introduce you to the actual delivery person before contract signing.
Red flagVague "our team handles this" without naming the actual human.
6. "What happens if I want to fire you?"
Why ask: Some consulting contracts have early-termination penalties, deliverable lock-up clauses, or "all custom IP belongs to us" clauses that make ending the engagement expensive. Read the contract before signing — these often hide in 8-point font on page 6.
Good answerMonth-to-month, you own all deliverables, no penalty for ending early.
Red flagMulti-month minimums, kill fees, vague IP language. The good consultants don't need to lock you in because they deliver value early enough that you don't want to leave.
7. "What's the first thing you'd actually do in week one?"
Why ask: This filters consultants who have a cookie-cutter process ("we always start with stakeholder interviews and a discovery phase") from those who can engage specifically with your situation on the call itself.
Good answer"Based on what you just told me about your inventory process, week one I'd shadow a closing shift and watch how counts actually happen" — operator-specific, not generic.
Red flagA 5-step methodology slide that looks identical regardless of the customer's actual situation.
What questions NOT to ask (and why)
- "Are you certified in AI?" — Real-world AI engineering moves faster than certifications. The good practitioners often don't bother with them.
- "What models do you use?" — Operator's question is "what does the tool DO" not "which LLM API powers it." If the consultant leads with model names, they're optimizing for a different audience.
- "What's your AI background?" — Most useful AI work in 2026 is built on LLM API plumbing that any thoughtful generalist can pick up. "Background" matters less than judgment.
- "Can you guarantee X% ROI?" — Anyone who guarantees a number on a sales call is making it up. The good ones give ranges with conditions.
A 5-minute self-check before any consulting call
Run this before booking the discovery call
- Do you have a specific named problem you want solved? (If not, do the audit-level engagement first, not a custom build.)
- Do you have an upper-limit budget you'd actually spend? (Consultants will price to your stated max if you let them; have a real number.)
- Do you have an internal owner who'll work with the consultant? (Engagements without an internal owner consistently fail to deliver.)
- Are you ready to make a decision in 30 days? (Some operators "evaluate consultants" forever; that's not really shopping, that's avoidance.)
If you want to see what an actual AI Operations Audit looks like — the deliverable, the timeline, the price — that's the first tier of services on this site. Two weeks, $2,500, fixed scope, written priority list at the end. Most operators leave with the audit and don't need the build phase. That's a feature, not a bug.
Frequently asked questions
Realistic 2026 ranges: $2,500-$5,000 for a focused operations audit (2-3 weeks, written deliverable). $5,000-$15,000 for a custom AI tool build (4-8 weeks). $2,000-$5,000/month for an ongoing retainer. Anything 3-5x above these ranges should have specific justification.
An audit assesses your operations and produces a written priority list of where AI would help and where it wouldn't. A build is the custom tool that solves one of those priorities. Most operators should do the audit first — many leave with the audit and don't need to commission a build at all.
Audit: 1-3 weeks. Build: 4-8 weeks. Retainer: month-to-month ongoing. If a consultant proposes a 6-month engagement before they've shipped anything, that's a red flag — they're padding.
Yes — both directions. Mutual NDA covers your operational data and their methodology. Avoid one-sided NDAs that protect them but not you. Mutual NDAs are standard and cheap (free templates exist).
Build kill-clauses into the engagement upfront. Month-to-month pricing, you own all deliverables, no early-termination penalties. Good consultants accept these terms because they're confident you'll keep them. Consultants who refuse these terms are protecting themselves against an outcome they expect to happen.