AI consulting services for small and medium businesses
AI consulting services that start from where your hours go rather than from what the technology can do. The output is a written recommendation naming what to build, what to buy as ordinary software, and what to leave alone.
Priced separately from building, so the recommendation is independent.
- Two to three weeks
- Credited against a build if you go ahead.
- Three costs, not one
- Build, run, and the staff time spent checking output.
- A portable document
- Written so another firm could execute it.
Much of what is sold as AI is ordinary software
Most businesses have been told they are behind without being told behind on what. The pitch is usually a demonstration with no clear connection to a cost the business actually carries.
The useful question is narrower: where does your team lose hours to work a computer should be doing, and is AI the cheapest way to recover them? Often it is not. A form, a rule or an integration is more predictable, cheaper to run and needs no supervision.
That is what our AI consulting services do: establish where the time goes, sort it into what AI helps with and what it does not, and put a number against each.
The most common recommendation this produces is a chatbot that books work. The second most common is ordinary software rather than AI at all, which is software development. If the real problem is that too few people find you, the answer is marketing, not AI.
The assessment starts with your hours, not the technology
The first week is spent on what your people do all day, which means sitting with them rather than asking in a meeting. The output is a list of tasks with hours attached, produced before any tool is discussed.
Everything after that is arithmetic. A task costing four hours a week is worth roughly five thousand dollars a year, and that figure decides whether a given fix is worth building.
What the assessment covers
Four steps, ending in a document you can act on without us.
Finding where the hours go
Week one · observationWe look at what your people actually spend the day doing: re-typing, chasing, reading documents to extract a few numbers, and answering the same question dozens of times a week.
That list is the basis of the recommendation. Most failed AI projects start from the technology and look for a use, rather than starting from a cost and looking for a fix.
Sorting it into three categories
AI, ordinary software, or no changeSome of it suits AI well: reading unstructured text, answering repetitive questions, summarising, and drafting a first version a person then edits.
More of it is ordinary software — a form, a rule, an integration — which is cheaper, more predictable and needs no supervision. Some should be left alone, because a process costing two hours a month will not repay the cost of automating it.
Costing each option
Build cost · running cost · supervision costThree numbers rather than one: the build cost, the monthly running cost, and the cost in staff time of checking its output. AI that requires a person to review every result has not replaced a person.
Against that sits the hours it returns, measured from the first stage rather than estimated from a vendor case study.
A written recommendation
The deliverable, written to be portableThe document names what to do first, what to do next, what not to do at all, and what each costs. It is written so another firm could execute it, and you are free to take it to one.
That is why advice is priced separately from building. If the only recommendation we could make were a project we sell, the advice would be worth nothing.
The scope, written down before you pay
Scope goes in the agreement before any money changes hands. It is written out in full below.
In the assessment
- Time spent with the people doing the work
- Tasks listed with real hours against them
- Each sorted into AI, ordinary software, or no change
- Build, run and supervision costs for each option
- A written recommendation, in priority order
What you can do with it
- Act on it yourself where it is simple enough
- Take it to another firm. It is written to be portable
- Hire us to build, with the fee credited
- Use it as a staff policy starting point
- Do nothing, knowing what you are declining
Not included
- A recommendation shaped by what we would prefer to sell
- A strategy deck with no costs in it
- Tool licenses resold at a markup
- Anything requiring a person to check every output
Tell us where the hours go. Our AI consulting services will tell you whether AI is the answer.
Three things you can check yourself
- Ask what we would not automate
- A consultant with no list of things to leave alone is selling rather than advising.
- Ask for the supervision cost
- The staff hours spent checking output. It is the figure most often omitted.
- Ask if the document is portable
- You should be free to hand it to another firm. Ours is written for that.
Questions we get on the first call
What do AI consulting services actually produce?
A written recommendation covering where AI would save measurable hours, where ordinary software is cheaper, where to do nothing, and what each option costs to build, run and supervise. You can act on it without hiring us.
What does it cost and how long does it take?
Two to three weeks, quoted against what you need assessed. If you hire us to build something afterwards, the fee is credited against it. Starting figures are on the pricing page.
Is AI actually worth it for a business like mine?
Sometimes, and less often than the marketing suggests. It is good at reading unstructured text, answering repetitive questions and drafting material a person then edits. It is an expensive substitute for a rule, a form or an integration, which is what many businesses actually need.
Will you just recommend a project you can sell?
The recommendation regularly comes back as ordinary software, or as no change. That is why consulting is priced separately: if we were only paid for building, the advice would not be independent.
What about our data and privacy?
It forms part of the assessment: which tools would see what data, whether that is acceptable in your industry, what must stay in house, and what your existing customer contracts commit you to. For anyone handling health or financial data this is usually the deciding factor.
Everyone says we will fall behind. Is that true?
Not in the way it is usually meant. For most local service businesses the advantage is narrow and specific: answering faster, quoting faster, and losing fewer leads. Those are worth pursuing; a general AI strategy usually is not.
Can you help us write a policy for staff using AI?
Yes, and it is worth doing before someone pastes a customer list into a chatbot. It is usually a one-page document: which tools may be used, what must never be entered, and who to ask.
Related: AI overview · Chatbots · Software development · Enterprise
Where privacy is the deciding factor we work from the vendors' own terms, such as OpenAI's enterprise privacy commitments, rather than from marketing material.
Tell us what you want AI to do
We will tell you whether it will work, what it would cost, and when ordinary software is cheaper.