AI for smaller organisations
Most AI projects fail. Here's how to not be one of them.
You are being sold AI by everyone. Very little of it comes with an honest account of what goes wrong, what it costs, or when the right answer is to do nothing yet. This page is that account — written for business owners, not technologists, by someone who has run transformation programmes for twenty-five years.
The one thing worth understanding
AI doesn't fix a broken process. It automates it.
Our position, plainly
The most common way an AI project fails is that a tool gets bolted onto ways of working that haven't changed. The result isn't a fixed problem — it's the same problem running faster, more visibly, and often in front of the customers you least wanted to show it to.
The alternative is unglamorous: change the process and the technology together, in small stages, each one proving its value before the next begins. It is slower to promise and considerably quicker to deliver something that lasts. It is also how you keep the confidence of your board, your staff and your customers while you do it.
This isn't a contrarian pose. It is what the evidence now shows, and the numbers have got worse rather than better as adoption has risen.
Read those together and the conclusion is uncomfortable for anyone selling AI tools: the tool is rarely the hard part. The hard part is knowing which process is worth changing, whether your information is in good enough shape to support it, and whether your people will actually work the new way.
Filtering the noise
What we'd tell you to be wary of, and what actually works
We have no product to protect here. These are the patterns we see succeed and fail in organisations of twenty to two hundred and fifty people.
Be wary of
- "AI transformation" quoted as a single large programme with one go-live date
- A customer-facing chatbot as the first thing you do
- Any proposal that doesn't ask hard questions about your data first
- Buying a tool because a competitor mentioned one at a networking event
- Pilots with no success measure agreed in advance, and no decision point at the end
- Anyone who won't tell you what AI can't do for you
What tends to work
- One internal process, chosen because it's repetitive and well understood
- A small, contained pilot with a number attached and a stop date
- Starting with the tools you already pay for, before buying anything new
- Fixing the information the process depends on before automating it
- Training the people who will use it, not just installing it
- Stopping at the end of a stage if the value isn't there
Three-minute check
Are we actually ready? Work down this list.
No technical knowledge required. Answer honestly — "I don't know" is a legitimate and very common answer, and it is itself useful information.
The risk nobody has told you about
Your team is probably already using AI. That's the exposure.
Whether or not you have adopted anything formally, there is a good chance staff are pasting work into free AI tools to save time. They are trying to be helpful. But client details, staff records, financial information or case notes entering a public tool can constitute a personal data breach — and under UK GDPR the penalties are calculated against turnover, which is a serious matter for a smaller organisation.
This is the cheapest problem on this page to solve and the most commonly ignored. A short acceptable-use policy, a list of what must never be pasted anywhere, and a twenty-minute briefing to staff will resolve most of it. We include this in the free assessment because it would be negligent not to.
Where it actually earns its keep
Unglamorous, internal, measurable
The successful projects we see are rarely the ones that make a good press release. They tend to look like this.
Trades & contractors
Getting the quote out the same day
For most trades the deciding factor isn't price, it's who replied first. Turning a site visit and a set of notes into a professional quote in twenty minutes rather than three evenings is worth more than any marketing spend.
Why it works
Internal, repetitive, and you can measure win rate before and after.
Professional services
First drafts, not final documents
Proposals, reports, file notes and correspondence drafted in minutes and then properly reviewed by the person accountable for them. The time saved is in the blank page, not in the judgement.
The boundary that matters
A human remains accountable for what goes out. That is not negotiable in a regulated practice.
Charities & not-for-profits
Funding bids and impact reporting
Small teams spend disproportionate time on bid writing and monitoring returns. AI can take the first pass at structure and draft, leaving scarce staff time for the parts that need genuine knowledge of the beneficiaries.
Handle with care
Never put beneficiary data into a public tool. Governance first, always.
Retail, hospitality & service
Enquiries triaged, not answered by a robot
Rather than a chatbot fielding customers, use AI behind the scenes to sort and prioritise incoming enquiries so a person answers the urgent ones first. Same benefit, none of the reputational risk.
Why start internal
If it gets something wrong, a colleague notices — not a customer.
Any organisation with a back office
The policies you never have time to write
Health and safety, data protection, EDI, environmental. Drafted in a fraction of the time, then reviewed and localised properly. Unglamorous, and it clears a job that has sat on the list for two years.
Why it works
Low risk, immediate output, and it builds confidence for the next step.
Anyone bidding for public sector work
Tender responses that actually score
AI drafts against the question; the judgement about what an evaluator rewards comes from experience. Social value sections in particular are routinely lost on evidence, not on effort.
The caution
Never submit a number you cannot substantiate. Commitments in a winning bid become contractual.
How we'd approach it
Stages you can stop at
Every stage produces something useful on its own. You can end the engagement at any point and still be better off than when you started.
STAGE ONE
Find out where you stand
What's already being used, where your information lives, what the obvious candidates are, and what needs fixing first. This is the free assessment.
STAGE TWO
Cover the basics
An acceptable-use policy, a short staff briefing, and a clear line on what never leaves the organisation. Cheap, quick, and it removes your biggest exposure.
STAGE THREE
One process, one measure
Pick a single internal process. Agree what success looks like as a number before starting. Run it for a defined period. Then decide honestly.
STAGE FOUR
Only then, widen it
Expand only what has demonstrably worked, and change the way people work alongside it rather than after it. This is where most projects go wrong, and where we're most useful.
No cost · No obligation · Nothing to buy
Book a free AI readiness assessment
One conversation, about ninety minutes, and a written summary you keep whatever you decide to do next.
- Where you stand today, honestly assessed
- What your people are already using, and the exposure that creates
- The two or three processes actually worth automating first
- What to fix before spending anything
- A written summary that is yours to keep
If the assessment concludes you should do nothing yet, we will tell you that plainly. We would rather be the people you call in a year than the people who sold you something you didn't need.
Also worth your time
Two other things we'd want you to know
Deadline: 31 January 2027 The copper phone network is being switched off Not just desk phones — alarms, lift emergency lines, card machines and older broadband run on the same network. Most businesses don't know what they're still using. Am I affected? Free guide · for charities & not-for-profits Implementing a new CRM — without the usual regrets Before you automate anything, your information has to be in a fit state. Seven phases, twelve lessons and four templates you can fill in, from real programme delivery. Read it freeStraight answers
Questions we get asked most
Is AI going to replace my staff?
In smaller organisations, overwhelmingly not — the evidence so far is that AI is being used to support people rather than replace them, and most firms using it report no change to headcount at all.
What it does change is what your people spend their time on. If someone currently spends six hours a week assembling a report, that time goes somewhere more useful. That's the honest version of the productivity claim.
We're a small business. Isn't this only for big companies?
The opposite, in one important respect: a small organisation can change a process in a week where a large one needs a committee. Your disadvantage is that you cannot absorb a failed project, so the discipline matters more, not less.
Start with what you already pay for. A competent tool costs tens of pounds a month. The cost that hurts is the implementation programme around it, which is exactly what we help you keep small.
What should this cost?
The assessment is free. Beyond that, a first sensible stage for a smaller organisation is a contained pilot on one process, not a six-figure programme. If a proposal you're considering doesn't have a measurable outcome within ninety days, that's worth questioning.
We will always tell you when the answer is a tool you can configure yourself rather than a project you pay us for.
Is it safe to put our information into these tools?
It depends entirely on the tool, the settings, and the information. Free consumer tools and paid business tools have materially different terms about what happens to what you type. Most organisations have never checked which one their staff are using.
If you handle sensitive personal data, this needs deciding before adoption, not after — including whether a data protection impact assessment is required.
Everyone says we're behind. Are we?
Probably less than you fear. Adoption figures range from roughly 16% to 78% depending entirely on whether the survey counts strategic deployment or someone once using a chatbot. The gap between those two numbers is where most of the anxiety lives.
The businesses genuinely getting value from AI are a much smaller group than the headlines suggest, and they got there by being deliberate rather than early.
Do you run sessions for groups or chamber members?
Yes. We've delivered practical workshops to chamber members before — including a GDPR readiness session for Merton Chamber of Commerce members ahead of the 2018 deadline — and we're happy to run the AI equivalent for a business network, a chamber, or a group of charities. Get in touch and we'll arrange it.
Figures on this page are drawn from published research current at the time of writing, including S&P Global and MIT analyses of AI project outcomes. This page is general guidance and not a substitute for advice on your specific circumstances. We are not solicitors and this is not legal advice on your data protection obligations.