How to Choose an AI Consultant for Your Small Business (a Practical Checklist)

How do you choose an AI consultant for a small business?

Pick an AI consultant the way you would pick a contractor to renovate your kitchen: ask to see finished work, get a fixed-fee proposal for a small scoped piece before you commit to anything big, and confirm you own everything they build. The single best filter is "show me a live product you built that real people use," because it separates people who ship working software from people who sell strategy decks. The right consultant starts with your messy operational data and a narrow workflow, not a 12-month transformation roadmap.

This is a checklist you can run through on a single fit call. It is written for owners and operators with no in-house data team who cannot afford to spend $25,000 finding out a vendor only knows how to make slideshows.

Who is this for?

This is for a small or mid-sized business with a real, specific problem ("our quoting takes three days," "we re-key the same data into four systems," "we cannot search our own contracts") that wants working software, not a workshop. If you only want someone to teach your team prompt basics, that is a different and cheaper purchase. This checklist is for when you want something built and running.

What questions should you ask an AI consultant?

Ask these in the first fifteen minutes of any call. The answers tell you almost everything.

• What will I be able to do at the end that I cannot do today? A good answer is concrete ("your team pastes an invoice and gets structured line items in QuickBooks"). A weak answer is a strategy document or a "roadmap."
• Can you show me a live product you built that real people use right now? Ask for the URL and watch them open it. Anyone can describe a project; far fewer can show one running in production.
• Where does your scope end? A consultant who claims they can do anything ("blockchain, agents, computer vision, whatever you need") is a generalist guessing. Real specialists name what they do not do.
• Do I own 100% of the code, data, and accounts? The build should run on your accounts (your cloud, database, API keys), not be held hostage on theirs.
• What happens to my data? They should raise data quality and privacy early and unprompted, especially if your data is regulated.
• What is the fixed price for the first small piece? If they cannot scope a starter engagement at a fixed fee, they are protecting their hours, not your budget.

If a consultant leads with the technology before they understand your business, that is a flag. The best ones ask about your workflow first and reach for a tool second.

Generalist vs specialist: which is better for AI work?

For a small business buying a working build, a specialist who knows your kind of data beats a generalist every time. AI projects rarely fail on the model. They fail on the messy, real-world data feeding it: the inconsistent invoices, the legacy export with sixteen date formats, the regulated claims file with PHI in it. Someone who has shipped against that kind of mess hits fewer surprises and quotes you a tighter price.

A generalist who talks fluently about every AI buzzword but has never wrangled data like yours will spend your budget learning on the job. Ask what kind of data they have actually worked in. At Palavir that has meant healthcare and Medicare claims, fraud signals, government and public records, legal research, and dealership and operations data, which is messy, regulated, operational data where the hard part is never the prompt. A Southeast Michigan AI consultant who can sit in your office and look at your actual systems also closes that gap faster than a generalist on a screen-share who has never seen your industry.

Build-it vs resell-it: does the consultant actually build?

There is a real divide in this market between people who build custom software and people who resell someone else's platform with a markup. Both can be legitimate, but you should know which one you are buying.

A reseller plugs you into a third-party SaaS tool, configures it, and bills you (often a recurring cut). You are renting their relationship with a vendor, so when the tool changes pricing or shuts a feature, that is your problem. A builder writes the pipeline that runs on your accounts and solves your specific workflow, and you own it.

Palavir builds; it does not resell. The working pipeline ships on your infrastructure and is yours to keep, modify, or hand to another developer later. Ask the question directly: "Are you building this for me, or reselling me a platform?" Neither answer is wrong, but a vague answer is.

How do you verify an AI consultant actually ships?

Talk is cheap in this market, so verify with evidence, not adjectives.

• Ask for live URLs, not case studies. A case study is a story; a live product is proof. Ask to see two or three things running today.
• Look for production track record, not proof-of-concept. A demo that worked once in a controlled setting is not software handling real users in a daily workflow. Ask how long their builds have been live.
• Ask for a reference who can name a real output gain. Not "they were great to work with," but "quoting went from three days to twenty minutes."
• Start with a paid audit before a big build. A small scoped engagement shows you how someone works before you spend implementation money. Palavir's AI Opportunity Audit does this: $2,500 for one workflow or $5,000 for a multi-workflow operation, scoped on a fit call, then a fixed-fee proposal. You see how they think before committing to a build.

For a fast read on whether your business is even ready, the free AI Readiness Scorecard walks you through it in a few minutes before you talk to anyone.

What are the AI consulting pricing models?

There are four common ways AI consultants charge. Knowing them keeps you from overpaying for the wrong shape of work.

• Hourly ($150 to $500/hr in 2026). Fine for short exploration. Risky for builds, because the meter runs and slow gets rewarded.
• Fixed-fee project. A set price for a defined deliverable. Best for small businesses: you get budget certainty and the consultant is rewarded for efficiency. A scoped build commonly runs $10,000 to $35,000.
• Monthly retainer. Ongoing access for continuous work. SMB retainers typically run $5,000 to $15,000 per month.
• Outcome-based. Payment tied to a result. Rare and hard to define cleanly for most small businesses.

For most owners the right path is a small fixed-fee audit, then a fixed-fee build, then a retainer only if there is ongoing work. As a benchmark, Palavir's AI Implementation Setup is a fixed $25,000 and ships one working Claude pipeline in ten business days, and the ongoing Implementation Partner retainer is $7,500 to $10,000 per month. Fixed fees mean you know the number before you sign.

What are the red flags when hiring an AI consultant?

Walk away, or at least slow down, if you see these.

• They promise specific results before seeing your data. Nobody can promise an ROI number on data they have not looked at.
• They promise "immediate" transformation. Real implementation takes weeks of working against your actual data. Instant transformation is a sales line.
• They lead with technology, not your workflow. If the first ten minutes are about their stack and not your problem, they are selling, not solving.
• They claim unlimited scope. "We do everything" means they specialize in nothing.
• They cannot show a live product. If every example is a slide or an NDA they cannot describe, assume nothing is running.
• They will not fix-price a small first step. Protecting their hours over your budget is the tell.
• You will not own the code or accounts. If the build lives on their infrastructure, you are renting, not buying.

The one-line version

Hire the consultant who can open a live product they built, scope a small fixed-fee first step, and hand you something you own that runs on your accounts. Everything else here just confirms those three things are true.

To start small and low-risk, the AI Readiness Scorecard is free, an AI Opportunity Audit is $2,500 to $5,000, and you can see how Palavir works with small and mid-sized businesses before committing to anything. Based in metro Detroit, serving Detroit, Ann Arbor, Troy, Southfield, and Southeast Michigan, plus remote nationwide. Call (248) 665-5757 or email josh@palavir.co.