Every proposal sounds similar and none of them is cheap. The questions below separate the engagements that leave something running in your business from the ones that leave a document.
Quick answer
Decide first whether you are buying advice, implementation, or both. Most disappointment comes from paying for one and expecting the other.
The question that sorts the field: what will actually be running in the business on the day you leave?
A good diagnosis names where AI pays for itself first and why, in the language of your business rather than the vendor's.
If nobody on your team is trained to run it, you have bought a dependency rather than a capability.
The proposals arrive and they read alike. A discovery phase. A strategy phase. Some tooling. A number at the end that is large enough to need a conversation with your accountant.
What is hard to tell from the outside is which of them ends with something working. Established owners are rarely naive about this. They have bought consulting before, and they know the failure mode: a thorough document, a sensible list of recommendations, and six months later nothing in the business runs differently.
Be Clear About What You Are Buying
There are three different purchases hiding under the same phrase, and they have different prices for good reasons.
Advice tells you what you should do. You get a diagnosis, a set of options and a recommendation, and your team does the work. This is genuinely valuable if you have the capacity to act on it and a waste of money if you do not.
Implementation means somebody builds and installs the thing. The output is not a document, it is a working process inside your business, with the unglamorous parts done: the connections, the knowledge written down, the edge cases argued about.
Ongoing support keeps it alive after the first month, when the exceptions show up and the people who built it have moved on.
Most engagements that disappoint were sold as the first and expected to deliver the second. Ask which one you are buying, in those words, before the price conversation.
The Question That Sorts the Field
If you ask one thing in a first call, ask this: what will be running in my business on the day you leave?
Listen for specifics. A named process, a place it runs, a person on your side who operates it. Vague answers here are the single most reliable warning sign available to you, because everything else in a sales conversation can be polished and this cannot.
The honest version of the answer often includes limits. One process rather than five. A step that still needs a person. A thing that will not work until your knowledge is written down. Those limits are evidence that somebody has done this before, not evidence of weakness.
Six Questions Worth Asking in the First Call
These are the ones that produce different answers from different firms.
What would you need to understand about my business before recommending anything? The answer tells you whether the diagnosis is real or a template.
Which job would you install first here, and why that one? A good answer refers to your operations. A weak one refers to what the industry usually starts with.
What has to be written down before this behaves? Anybody who says nothing has not built one of these.
Where does a human stay in the loop, and who is it? Especially on anything a customer sees or anything involving money.
What does my team need to learn to run this without you? If the answer is nothing, that is not reassurance.
What happens when it breaks? Everything breaks. The question is whether there is an answer rehearsed or improvised.
What a Real Diagnosis Looks Like
A diagnosis is not an audit of your software. It is a reading of how the business actually runs, where growth is held up, and which of those constraints a system could take off a person.
Strategic Marketer calls that first step the AI Business Assessment, and what it produces is deliberately narrow: a clear statement of what is holding AI back in the business, the two or three places it would pay for itself first, and a prioritised roadmap you keep regardless of what you decide afterwards. The reason it stays narrow is that most businesses do not have an AI problem, they have a sequencing problem, and a roadmap that names what comes first is worth more than a list of everything that is possible.
If a diagnosis rather than a pitch is what you want next, that is what the assessment is for: https://strategicmarketer.com/#contact
Whoever you hire, insist that the diagnosis is yours to keep. A diagnosis that only makes sense inside their proposal was a sales document with a different name on it.
Who Owns It When the Consultant Leaves
This is the question owners regret not asking.
A system that only the supplier understands is a dependency. It is not necessarily a bad deal, plenty of businesses are happy to outsource permanently, but it should be a decision rather than a surprise that arrives with the second invoice.
The alternative is that somebody on your team is trained to operate it, knows where the knowledge lives and can change it when the business changes. That person does not need to be technical. They need to understand the process well enough to notice when the output is wrong, which is a different and more important skill. The approach that leaves something installed and a team able to run it is the one worth paying for, because it is the only version where the value stays after the engagement ends.
What the Shape of the Price Tells You
You will see a few different structures, and the structure itself is informative.
A fixed project fee suits a defined installation with a clear end. A monthly figure suits ongoing support, where the work continues after the build. A combination is common: something to install it, something to keep it running and looked after.
What should make you pause is a price with no scope attached to it, or a scope that lists tools rather than outcomes. Ask what the number buys in terms of processes installed and people trained, and ask what happens to it in month four. Treat any specific result quoted from another client as that client's account rather than a forecast of yours. Those numbers are real for the business they happened in, and they are not a prediction about yours.
Warning Signs Worth Taking Seriously
Some patterns repeat often enough to be worth naming.
A recommendation arrives before anybody has asked how the business runs. The proposal is a list of tools to subscribe to rather than processes to install. Nobody mentions writing anything down. A specific outcome is promised in a timeframe. Training is absent from the scope entirely. And the contrast the industry itself keeps producing: a thorough list of AI tools, handed over, with the hard part left for you.
None of those is automatically disqualifying. All of them are worth asking about directly, and the quality of the answer tells you more than the slide deck did.
Before You Hire Anyone, Know the First Job
There is one piece of work worth doing before any of these conversations: deciding what you would want installed first if you could only have one thing.
You do not need a consultant to form a view on that, and arriving with one changes the conversation entirely. It turns a sales pitch into a working session, and it makes the proposals comparable, because they are all answering your question instead of their own. If you have not formed that view yet, choosing the first job to give it is the place to start, and the conditions that have to be true before any of it works are set out in what has to be true before it works.
Get the Diagnosis Before You Buy the Build
Ask the questions above, keep the roadmap whoever you hire, and start from how your business actually runs rather than from a list of tools.
Schedule Your AI Business Assessment →
One conversation. A clear diagnosis of what is holding AI back in the business, the two or three places it would pay for itself first, and a prioritised roadmap you keep either way. No pressure and no obligation, and if it is not a fit, they say so.
Questions Owners Ask Before Hiring AI Consulting
Do I need a consultant at all, or can my team do this?
If somebody on your team has the time, the authority to change a process and the patience for the written down part, they can make real progress. Most owners discover the constraint is the second one. A consultant is often being paid for decisiveness and sequence rather than technical knowledge.
How long should a first engagement be?
Long enough to install something and watch it survive contact with a busy week. Engagements that end at the recommendation stage tend to end there permanently, because the business returns to its normal workload the following Monday.
What should I prepare before the first conversation?
A plain description of how the work currently flows, the jobs that eat the most time, and the two or three questions your customers ask constantly. That is more useful than any document about AI.
Is it a problem if they do not know my industry?
Less than you would expect. What matters is whether they ask about your operations before recommending anything. Industry familiarity speeds up the conversation, but it also produces assumptions, and assumptions are what put the wrong system in the wrong place.
How do I compare two proposals that look the same?
Compare what is running at the end, who operates it, and what has to be written down first. Those three answers differ far more between firms than the phases in the proposal do.
What if I am an agency doing this for clients rather than for myself?
The same questions apply, with one addition about how the work is delivered repeatedly rather than once, which is covered on the for agencies side.