I run five small businesses on a roughly $600 a month AI stack. I sit on the operator side of the table every day and I also sit on the consulting side. That means I get a strange view: I see the pitches my own competitors send out, and I see the bills they send afterward.
This is the post I wish a buyer had given me five years ago when I was getting sold PowerPoint by people who could not log into the system they had built me. Five concrete red flags. Each one shows up in pitch decks, sales calls, or pricing pages. You can spot them in the first 20 minutes if you know where to look.
Ask one question on the first call: show me a live system you built and you still touch every week. Not a client case study. Not a logo wall. A real screen.
If they can screen share an actual dashboard, a Slack alert that fired this morning, or a log file from a voice agent that took a call yesterday, you are talking to an operator. If they pivot to a slide that says "AI Transformation Framework" you are talking to a strategy seller. There are uses for strategy sellers. Building a working system for a $1M to $10M operator is not one of them.
The 5+ businesses I run all break in small ways every week. The voice agent at the stone distribution business broke twice last month. I fixed both from my phone. If your fractional AI officer cannot do that, they are not running anything, they are advising on it.
Three of the five top competing services I reviewed this month offer a free AI audit. Read the fine print. In most of them the audit is a sales call recorded as a deliverable. The actual deliverable is a recommendation to start a $5,000 a month retainer.
The structure tells you what they want from you. An audit with no standalone deliverable pays back only if you sign up for the recurring service. An audit that ships a leak map with dollar estimates pays back the moment it is delivered, whoever does the work after. The deliverable aligns the consultant to spend real time on your problem rather than steering you to the next package.
I run a productized 5-day audit, free for qualified operators, with a deliverable that stands on its own: three prioritized leaks, dollar estimates per finding, a 20-page PDF you keep either way. There is no retainer here to funnel into, so the audit has nothing to sell you except the findings. If it surfaces nothing worth automating, you spent a week, not a dollar. The risk sits on me, not you.
A $5,000 a month retainer with "ongoing strategic guidance" written in the contract is a subscription, not an engagement. By month four nobody can tell you what was delivered. The consultant gets paid whether the AI system you wanted shipped or not.
The right structure is a fixed-scope Sprint at a fixed price for a fixed deadline. One leak, four weeks, one number that moves. When that ships, you decide whether to start a new Sprint or not. The day the value stops, the payment stops. The day they ship, they get paid. That is the right alignment for a $1M to $10M business buying AI systems for the first time.
If you want the math behind why retainers underperform Sprints, the real pricing guide for fractional AI officer engagements breaks down the four common models and what each one actually delivers.
This one is subtle. The pitch deck mentions a proprietary system: M.A.P., WINS, Blueprint to Business, Five Pillar Framework. Frameworks make sense at a Fortune 500. They make zero sense at a paint distributor with eight people and a forklift.
That branded system is for the consultant's marketing, not for your business. Real implementation at small-business scale is messy. It looks like: "the customer service inbox has 40 messages, here is a Claude prompt that drafts replies, you approve them, we measure response time tomorrow." That fits on one page. A branded methodology that requires onboarding workshops and stakeholder maps is a sign you are buying the wrong product.
This is the cleanest test of all. Ask: what does your team spend per month on AI tools and API calls across all the systems you run yourselves? If the answer is a hand wave, they do not operate. If the answer is precise, they do.
My answer is a tight operating stack across 5+ businesses (a Mozabrik photo mosaic ecommerce store, Kompozit USA paint distribution, OD Granite stone import, this consulting practice, and a Gifted Kids education retainer). It buys voice agents, content drafting, daily P&L reports across the portfolio delivered to Telegram at 9:00 AM, and customer service drafting for Amazon. I can break it down line by line in the operator stack post.
A real operator can answer this question without checking notes. They live with the bill. They watch it grow when Claude Sonnet usage spikes during a Sprint and they trim it when a customer service automation gets retired. A consultant who outsources every build cannot answer this. They have no bill. They sell hours.
A good fractional AI officer has at least three of the following. They show one production system live in the discovery call. Their audit ships a standalone deliverable with dollar estimates, free or flat-fee, never metered. They scope every Sprint to one measurable leak with a baseline and a target. They publish their AI cost openly. They have published lived case studies with specific numbers, not anonymous client logos.
Forbes featured my operator playbook in a Gene Marks piece for exactly this reason. The category is overrun with strategy theater. Outside press calls out the operators who actually run what they sell. If your candidate has been covered by name in a credible outlet, that is third-party signal you cannot fake.
You do not need a long checklist. Three questions surface ninety percent of the truth.
If any answer is vague, you have your signal. The consultants who pass these three questions are a different category from the ones who fail them. The price difference between the two groups is usually zero.
Last year I hired a different fractional consultant to look at the bookkeeping workflow for OD Granite. He had a deck with a five-step methodology. He had references. He could not show me a single live system he had built. Three months and $9,000 later I had a 40-page strategy document and zero working automation.
I built the replacement myself in two weeks using Claude API and n8n. Total cost to operate: under $50 a month. Hours saved per week: 12. I logged the full nine-month run in the bookkeeper replacement case study, including the months it drifted and what fixed it. The lesson was not that consulting does not work. The lesson was that consulting without an operator behind it is a $9,000 PDF.
If you are evaluating fractional AI officers and want to put these red flags to use, three options exist.
One: take the AI Readiness Quiz. Eight questions, 5 minutes, you find out if your business is even ready to build now or needs to clean up data and process first. Free.
Two: book a 30-minute call. I will spend the first 15 minutes asking the questions any good fractional officer should ask you, and the last 15 minutes telling you what your top automation candidate would look like with a dollar estimate attached. Calendly link here.
Three: read the services page. The productized AI Growth Audit is free for qualified operators. Sprint starts at $5,000 for one leak in four weeks. Full Install starts at $20,000 for three to five systems in 8 to 14 weeks. No retainer ever.
Want a buyer-side opinion on a fractional AI officer pitch you are evaluating?
Book a 30-min callFor a mid-market operator, a project Sprint should land at $5,000 to $8,000 and a full multi-system install from $20,000. Anything above $15,000 per month in retainer for that size is enterprise pricing pointed at the wrong customer. An audit is the special case: judge it by the deliverable, not the invoice. A real one ships a standalone leak map with dollar estimates (ours is free and does exactly that); an audit with no deliverable is a sales call, whatever it costs.
Depends on what the free audit is. Most exist to convert you to a retainer: the audit is a sales call dressed up as a deliverable. The test is simple. Does it ship a standalone deliverable with dollar estimates you can act on without hiring anyone? Is there a retainer waiting behind it? Our 5-day audit is free, ships a 20-page leak map with dollar estimates per finding, and there is no retainer to funnel into. A free audit that fails that test is not an audit, walk away.
Retainers create the wrong incentive. The consultant gets paid whether the system you built ships or not, and whether the business outcome moved or not. A project Sprint with a defined scope and a fixed price gets paid for shipping something measurable. The day you do not need a Sprint, you stop paying. That is the right structure for the buyer.
Ask them to show one production system they personally built and operate. Not a case study with a client name. A system they touch. If they cannot screen share an actual dashboard, log, or interface, you are buying slides. The 5+ businesses I run on AI exist precisely because I refused to outsource the build.
Three questions surface ninety percent of the truth. One: show me a system you built and operate yourself, not a client deck. Two: what is your monthly AI cost across all the systems you run, and what does that buy. Three: what does my deliverable look like on day five, and what does it look like on day thirty. If the answers are vague on any of the three, that is your signal.