In this article
- Gartner reports that at least 50% of generative-AI projects are abandoned after proof-of-concept: usually for poor use-case selection, missing data readiness, or cost overruns.[1]
- The questions below are filters for the consultants who keep projects in the surviving 50%, not the ones who’ll write you a slide deck.
- The single biggest red flag isn’t price: it’s vagueness. Anyone who can’t answer concretely on day one isn’t going to answer concretely on day 30.
AI consulting is a booming industry, which means the range of quality varies wildly. Some consultants deliver real results. Others deliver slide decks and invoices. Here are seven questions that separate the two: and what the right answers sound like.
GenAI project survival funnel
Gartner: organizations abandon ~50% of GenAI projects after PoC; 45% of high-maturity orgs sustain 3+ years
Source: Gartner press releases on GenAI project abandonment (2024) and AI maturity (2025)
1. Do They Know Your Industry?
AI for a law firm is fundamentally different from AI for a manufacturer or a healthcare practice. The compliance requirements, the software ecosystem, the decision makers, and the pain points are all different.
A good consultant should be able to name specific tools you use (Clio for law firms, athenahealth for healthcare, Epicor for manufacturing), describe common pain points without you explaining them, and know the compliance landscape of your industry inside and out.
Red flag: They talk about “AI in general” but can't name a single software platform in your industry.
2. Are They Transparent About Pricing?
You should know exactly what you're paying before you sign anything. The right consultants will give you a real number on a discovery call (after enough conversation to understand your scope) and follow it up with a written, fixed-scope quote. Beware of consultants whose pricing is structured as open-ended hourly billing, or who need “a few weeks of paid discovery” before they'll commit to a number.
Fixed-price packages are ideal for small businesses because they cap your risk: you know the total cost before work begins. For more on what to expect, see our guide to AI consulting costs.
Red flag: Hourly billing with no cap, vague scope, paid “discovery” phases that should be free, or pricing that requires a lengthy “custom proposal” for even simple engagements.
3. What's the Implementation Timeline?
A targeted automation (like client intake or document processing) should take 2–4 weeks to implement. A broader engagement covering multiple workflows might take 4–12 weeks. If someone tells you it will take 6 months to automate one process, they're either over-engineering it or under-experienced.
Red flag: No clear timeline, or a timeline that stretches into months for a single automation.
4. How Do They Handle Data Security?
This is non-negotiable, especially for regulated industries. Ask specifically: Where does my data go? Is it encrypted? Who has access? Is data used for model training? Do you have a BAA (for healthcare)?
The right consultant should have clear, documented answers to all of these: not “we use industry best practices” hand-waving. (For healthcare specifically, our HIPAA AI Implementation Checklist walks through what those answers should sound like.)
Red flag: They can't explain the data flow, or they suggest using consumer AI tools (like standard ChatGPT) for sensitive business data.
5. What Does Training Look Like?
Technology adoption fails when the team isn't trained. Ask: Will you train my staff? How? Is it hands-on or a manual? What happens when someone has a question after go-live?
The best consultants provide hands-on training with your actual workflows, not generic slide presentations. They train your team on the specific automation built for your business, with real examples and real data.
Red flag: Training is a PDF manual emailed after the project. Or there's no training at all: just a handoff.
6. What Ongoing Support Do They Offer?
Every automation needs adjustments after go-live. Workflows change, software updates, and edge cases appear. Ask: What support do I get after the project? For how long? What does ongoing support cost?
A minimum of 30 days of post-launch support should be included in the project. Beyond that, tiered monthly retainer options provide continuous optimization and new automations as your needs evolve.
Red flag: Support ends the day the project is “delivered,” or ongoing support requires a completely new contract.
7. Can They Show Results?
Ask for specific examples: hours saved, error reduction, revenue impact, client satisfaction. The best consultants can describe concrete outcomes from similar engagements: even if client names are confidential, the numbers and context should be specific.
Be cautious of consultants who can only point to case studies from Fortune 500 companies or whose results are framed in vague terms like “improved efficiency.” You want numbers.
Red flag: No concrete examples, only theoretical benefits and aspirational language.
Why GenAI Projects Fail: And Which Question Catches Each
Map the questions to the failure modes Gartner sees most often:
| Failure mode (Gartner) | Catches it | What good looks like |
|---|---|---|
| Poor use-case selection | Q1, Q7 | They name 2–3 specific automations relevant to your firm with hours-saved estimates |
| Missing data readiness | Q1, Q4 | They walk you through your data flow before quoting |
| Cost escalation | Q2 | Fixed-scope written quote, ongoing-support pricing in writing |
| Weak risk controls | Q4 | BAA / DPA on file with named subcontractors |
| Adoption stall after launch | Q5, Q6 | Hands-on team training + 30+ days post-launch support included |
The Bottom Line
The right AI consultant makes the process feel simple. They know your industry, they're honest about pricing, they work fast, they take security seriously, they train your team, they stick around after go-live, and they can prove their work delivers results.
If you're evaluating your options, start with a free AI Readiness Assessment. It gives you a clear picture of where AI fits in your business: which makes it much easier to evaluate whether a consultant's proposal actually addresses your needs.
Frequently Asked Questions
Not always: but for small businesses it's usually the wrong fit. Hourly billing transfers all of the scope-and-timeline risk from the consultant to you. Fixed-scope contracts force the consultant to actually understand your situation before quoting (which is the work that should happen anyway), and cap your downside if the project takes longer than expected. Hourly is reasonable for ongoing partnerships where scope is genuinely fluid; for first engagements, fixed.
Three questions that get past the “they were great” reflex: (1) “What did they get wrong on the first attempt?” (2) “What did you wish you'd known before you started?” (3) “If you were doing it again, what would you do differently?” Bad consultants' references can't answer these without revealing real problems. Good consultants' references answer them in a way that builds your confidence.
Generalist AI shops can deliver value: if they pair their AI fluency with someone who knows your industry deeply (often you, on the project team). The risk is that they design solutions for a generic business instead of yours. The hybrid model (industry expertise + AI execution) is usually safer than either pure generalist or pure industry-veteran-but-light-on-AI.
Generally not in the first year. Consolidating with one partner means consistent integration, single-throat-to-choke for support, and tooling that works together. After 12–24 months, when you have multiple mature systems and want specialized depth in one area (e.g., a niche tax-research AI vendor), branching out can make sense. Premature multi-vendor strategies usually cost more in coordination overhead than they save.
For Quick Win-tier projects: 50% on signing, 50% on delivery is standard. For longer engagements (Practice Accelerator, Full Transformation): a 30/40/30 split tied to milestones (kickoff, mid-engagement deliverable, completion) keeps incentives aligned. Avoid 100% upfront and avoid 100% on delivery: the first transfers all risk to you, the second transfers all risk to the consultant. Both produce bad behavior.
- Gartner press release, “Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025” (July 29, 2024); supplemented by Gartner's research on agentic-AI project failure rates and AI-ready data requirements. gartner.com/.../gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned
Ready to Evaluate Your Options?
Start with our free AI Readiness Assessment. Know exactly where AI fits in your business before talking to any consultant.