AI for Business
Without the Sales Pitch
Most businesses are not short of AI opinions. They are short of anyone willing to say which parts apply to them and which do not. This is everything we tell clients, written down.
Start Here
If you have half an hour, read what AI can actually do for a business like yours, then take the five-minute readiness check. If you have five minutes, take the check first. It asks fifteen questions, needs no email address, and tells you whether the groundwork is in place before you spend anything.
The Four Questions Every Business Asks
In roughly this order, in our experience.
What could it actually do for us?
Function by function: admin, customer enquiries, sales, finance, operations and engineering. What works today, what it saves, and what still needs a person.
What is this going to cost?
Licence prices are the small part. The real bill is time, checking and change - and it is the part nobody quotes you.
Is our data safe in these tools?
What happens to what you paste, the three questions to put to any vendor, and when keeping the model in your own building is the right answer.
How do we start without wasting money?
The ninety-day method: find the candidates, pick one, prove it with a number, then decide. Stopping is a valid answer.
Our Position, Briefly
AI is neither the thing that will transform your business nor a fad you can wait out. For most established companies it is a modest, real improvement to a handful of tasks, arriving with a great deal of noise attached. The businesses that do well with it are unromantic about which tasks those are.
The government's own assessment of why smaller firms hold back names three barriers - capability, cost, and awareness of what the benefit would even be.1 That last one is not ignorance. Nobody has connected the technology to your particular business, because the people offering to do so are usually selling something with a fixed shape.
Meanwhile the failure rate is real. Gartner expects more than 40% of agentic AI projects to be cancelled before the end of 2027, and attributes it to escalating costs, unclear business value and inadequate risk controls rather than to the technology falling short.2 Every one of those causes is a decision made before anything was installed.
So we start from your work rather than from a product, we test one thing with a number attached, and we tell you when the answer is that nothing here is worth doing yet. That answer is more common than the industry admits and it costs you a fraction of the alternative.
If You Already Know Where You Are
You have no idea where AI fits
Start with what AI can do by business function, then the readiness check. If you would rather talk it through, the opportunity review is a fixed two to three weeks and ends with a short list and a recommendation.
You bought something and nothing changed
Usually the technology was fine and the working practice around it never changed. What AI actually costs covers why that happens, and the ninety-day method covers how to test the next one properly.
Your staff started using it without asking
Common, and not a disciplinary matter. Read is our data safe, then put a one-page rule in place. The governance policy guide has a full template if you want to go further.
You have engineers and the question is about the product
Different problem, different content. See introducing AI coding tools to an engineering team, measuring whether the tooling pays, and our software development advisory.
You are a professional services firm
Client confidentiality changes the answer to nearly every question here. See AI and technology advisory for professional services firms.
Your data is stuck in systems from 2009
Then that is the project, not AI. Legacy modernisation comes first, and this piece on when a codebase becomes a business risk explains how to tell.
Further Reading From Us
🖥️ Local AI vs Cloud AI
When running the model on your own hardware is worth the trouble, with the hardware and the trade-offs spelled out.
📈 Digital Transformation for SMEs
The same argument applied more broadly: change one thing, measure it, then decide on the next.
🤔 What You Actually Need From Advisory
Six places where what businesses ask for and what they need come apart, and what it costs when they do.
Sources
- Department for Business and Trade - SME Digital Adoption Taskforce: 2026 update.
- Gartner - Over 40% of agentic AI projects will be canceled by end of 2027.
Want the Short Version for Your Business?
A free half-hour. Describe the business and we will tell you the two or three places worth testing, or that there are none yet.