Choosing the right AI consulting firm is one of the highest-leverage decisions a business leader can make. The wrong choice wastes months and hundreds of thousands of dollars; the right one transforms your operations in weeks. Here are the 10 questions that separate credible AI consultants from expensive PowerPoint factories.
1. Can you show me production systems you've built?
The single most important question. Any firm can build a demo; the question is whether they've deployed systems that run in production with real users and real data.
Ask for specific examples: What does the system do? How many users does it serve? How long has it been running? What are the uptime metrics? A credible firm will show you live systems, not screenshots. Be wary of firms that only reference "strategy engagements" or "proof of concepts" — that often means they advise but don't build.
At AffixedAI, every engagement includes a deployed production system. Our showcase features live applications running on client infrastructure.
2. Who actually does the work?
At many consulting firms, the senior partner sells the engagement and then hands it off to junior consultants. Ask who will be doing the hands-on technical work on your project.
The AI field moves so fast that experience level matters enormously. A team of junior engineers building their first RAG system will take 3-6 months for what an experienced AI engineer deploys in 2 weeks. Ask about the specific people who will work on your project, their backgrounds, and whether the senior person who pitches stays involved through delivery.
3. What do I own when the engagement ends?
This question exposes vendor lock-in risks. You should own 100% of the code, data, and infrastructure — running on your own accounts.
Some firms build on proprietary platforms that require ongoing licensing fees. Others retain IP rights to code they develop. The best firms deploy everything on your infrastructure (your Vercel, AWS, or Supabase accounts), hand over all source code, and train your team to maintain and extend it. If the firm disappeared tomorrow, your systems should keep running.
4. What's your delivery timeline?
Be skeptical of both extremes. "We'll have a full AI system in a week" is unrealistic. "We need 6-12 months for discovery" is a red flag for scope creep.
A firm with pre-built infrastructure should be able to deliver a first production system in 2-4 weeks. Complex multi-agent systems may take 2-3 months. If a firm needs longer than that for an initial deployment, they're likely building from scratch — which means you're paying for their learning curve.
See our process page for typical AffixedAI timelines: Empowerment delivers in 2 weeks, Growth partnerships have production systems in the first month.
5. How do you measure success?
Vague answers like "improved efficiency" or "AI-powered transformation" are worthless. Demand specific, measurable outcomes tied to business metrics.
Good answers include: "We'll reduce your customer response time from 4 hours to 15 minutes," or "We'll automate 30 hours/week of manual data entry," or "We'll cut your support ticket volume by 40%." The firm should establish baseline metrics before the engagement starts and report against them regularly. Learn more about measuring AI ROI or try our AI ROI Calculator for a quick estimate.
6. What AI models and tools do you use?
A model-agnostic approach is essential. Firms locked into a single vendor (only OpenAI, only AWS) limit your options and increase your costs.
The best firms evaluate which models fit each use case. Some tasks need GPT-4o for speed, others need Claude for nuance, and some benefit from open-source models for cost efficiency. Ask about their model selection criteria and whether they'll help you switch models as the landscape evolves.
7. How do you handle data security and privacy?
Your data is your competitive advantage. Ensure the firm has clear protocols for data handling, model access, and compliance with your industry's regulations.
Key questions: Where does my data go? Is it used to train AI models? What encryption is used in transit and at rest? How do you handle PII? Are you SOC 2 compliant? Does the system meet GDPR/HIPAA/industry-specific requirements? A credible firm will have documented answers, not ad hoc assurances.
8. What happens when something breaks?
AI systems require monitoring, maintenance, and occasional intervention. Ask about post-deployment support, SLAs, and who handles incidents.
Does the firm offer ongoing support? What's the response time for production issues? Is there monitoring for model performance degradation, cost overruns, and accuracy drift? Will they train your team to handle common issues independently? The best engagements include knowledge transfer so your team can self-serve after the initial deployment.
9. What's the total cost of ownership?
Consulting fees are just the start. Ask about infrastructure costs, API spend, ongoing maintenance, and what happens when you need changes.
A transparent firm will break down: (1) their engagement fee, (2) estimated monthly infrastructure costs (hosting, databases, API calls), (3) ongoing support costs, and (4) the cost of future modifications. See our full cost analysis for detailed benchmarks.
10. Can I start small and scale?
The best AI consulting relationships start with a contained proof of value and expand based on measurable results. Be wary of firms that require massive upfront commitments.
A good firm offers tiered engagements. Start with a focused project to prove the methodology works for your business. If it delivers, expand scope with confidence. If it doesn't, you've limited your downside.
AffixedAI offers exactly this path: start with an Empowerment Partnership ($7,500, 2 weeks), scale to a Growth Partnership if the results warrant it, and your Empowerment investment applies as a credit.
Red flags to watch for
Beyond the 10 questions, here are warning signs that a firm may not deliver value:
- Strategy-only engagements: They produce reports and roadmaps but don't build or deploy anything.
- Black-box platforms: They deploy on their proprietary platform with no code access.
- No vertical expertise: They claim to serve every industry equally. Specialized knowledge matters.
- Buzzword-heavy proposals: Excessive use of "digital transformation," "synergy," or "AI-powered disruption" without concrete deliverables.
- No team training component: If they don't plan to train your team, they're optimizing for ongoing dependency, not your success.
Next steps: evaluating your options
Use these 10 questions as a scorecard when evaluating AI consulting firms. The firms that give clear, specific, verifiable answers to all 10 are the ones worth your investment.
For a broader perspective on the consulting landscape, see our comparison guides: AI Consulting vs. Big 4 Firms, vs. Freelance Developers, and vs. In-House Teams.
Or, skip the evaluation process and take our free AI assessment to see if AffixedAI is the right fit for your business — no commitment, no sales pitch, just an honest evaluation.