Guides
The first thing an AI consultant sells you is a document. It says what is worth doing in your business, what each change would cost to carry out, and what it would produce. Before you decide who to hire, decide what has to be inside that document. Nobody tells you how to judge one.
Search for an AI consultant, an AI advisor or an AI implementation consultant and you get three names for overlapping work. The first deliverable is almost always the same. It is an assessment, sold as a paid engagement and priced in the thousands. We give that document away. This guide is the reason why, and what to look for if you take it somewhere else.
Strip the names away and the first stage of nearly every AI engagement has the same shape. Somebody talks to you, looks at how your business runs, and writes down what they think you should do next. What varies is how much of your actual business gets into it.
That is why the assessment is where the money sits in this industry, and why so many firms lead with it. Billable, fast, and it produces something the client can hold. A good one changes what you do on Monday. A weak one is a tour of tools that already exist, reformatted around your industry, and you can usually tell which of the two you are holding by the second page.
Here is the test. Open the document and look for a number that came out of your business rather than out of a brochure. If every figure in it is an industry average, nobody read your data.
Between the conversation and the recommendation there is a step that costs real time and is the first thing dropped when an assessment is being produced at volume. Somebody has to read your business's own history.
Appointments, cancellations, invoices, customer records, whatever your systems have been collecting while you were working. It is rarely tidy. It is usually more revealing than anything said in the room, because the calendar remembers things an owner has stopped noticing.
We have done this twice in public and written up both. At an acupuncture practice the staff read seven years of appointment history before any conversation about platforms, and what that reading turned up set the order of everything afterwards. At a chiropractic practice the review came before a migration, and the finding was that the complaint and the cause were two different things.
An assessment that can only arrive at one conclusion is a quotation with extra pages. A review worth its name has three places it can land, and the third one is the reason to trust the other two.
Build something. There is work in your business that runs on somebody remembering, and it can be handed over. This is the answer people expect and it is the least common of the three in the first year.
Fix what you already own first. The most frequent finding by a distance. You are paying for a system whose setup was never finished, and putting AI on top of it would automate the mess rather than clear it. This answer costs the least and annoys the most, because it means the problem was never the software.
Do nothing for now. Sometimes the honest reading of a business is that the timing is wrong, the volume is too low for the work to pay back, or the owner has something more pressing. Writing that down costs us the engagement. It is also the only thing that makes the other two worth believing.
For some businesses the document is enough on its own. For others, handing it over and walking away would leave the owner holding a plan they have no way to run. Size has little to do with which is which.
One system that mostly works, a short list of changes, and somebody in the business who enjoys this kind of thing. Take the plan, do it in the order it recommends, and call someone if a step stalls. Paying a monthly fee to be told what you are already able to do is a bad trade and we will say so in the document.
Several systems that disagree with each other, work that has to keep running while things change underneath it, and no one whose job this is. Here the order of operations decides whether it works, and the order is the expensive knowledge. This is where an implementation engagement earns its price, and where we would rather be judged on the first month than on a contract.
One conversation, a review of your own data, and a written plan addressed to you. It covers what is worth doing, every realistic option weighed against the others including staying exactly where you are, a timeline, and every cost involved including ours. If the reading says do nothing yet, the document says so plainly. It is yours to keep whether you engage us or not.
You need the document one produces. Whether you need the person depends on what the document says. A business with four staff and one system that mostly works often needs a short list of changes and nobody to manage them. A business running five tools that disagree with each other has a sequencing problem, and sequencing is the part worth paying for. Get the plan first. It tells you which of those you are.
Yes, and it is worth knowing what to ask them for. The useful version reads your own business history before recommending anything: appointments, invoices, customer records, whatever exists. The other kind arrives holding a list of tools. Ask which of your files they want to see, and how early in the process they want them. The answer separates the two.
The assessment is commonly sold as a paid engagement priced in the thousands, and what it produces is a document. After that, implementation is usually a project fee or a monthly retainer, and the honest ones calculate both from your own numbers rather than quoting a package before they have seen anything. Here the plan is free, including every cost written inside it. If a firm will not put its own price in the document, that is worth asking about before you engage.
Advice against building. A consultant tells you what is worth doing and hands you a plan. An implementation consultant builds it and stays to see it run. Some firms do one, some do both, and a few sell the first as a route to the second without saying so. Ask who maintains the work at seven in the morning when it stops.
Often, and for some businesses that is the right answer. The tools you already pay for have added AI features, and a fair number of owners are paying for capability they have never switched on. What you cannot easily do alone is read your own history against what the alternatives would actually cost, because that means exporting your data, cleaning it, and comparing systems you have no reason to know. If the answer turns out to be that you already own what you need, a plan worth reading says so.
Six things. What your own data shows. What the current system does for you and where it falls short. Every realistic option weighed, including staying where you are. What each change would produce, worked from the figures your business already has. A timeline. Every cost, including the cost of the firm writing it. A document missing the last one is a sales sheet with a plan's cover on it.
Next: the AI features your software is already adding, and whether they are worth the bill
