Most businesses do not need to start their AI journey by choosing a tool.
They need to start by looking at the work.
That may sound less exciting than comparing the latest AI platforms, but it is usually where the value is. A business can waste a lot of time asking which tool to buy before anyone has asked the more useful question: “Where is work currently slow, repetitive, risky or dependent on one person?”
That is what I mean by an AI audit. It is not a grand consultancy exercise with a huge report at the end. For a small or medium-sized business, it should be a practical review of how work actually moves through the organisation, where time leaks away, and where AI might help without creating unnecessary risk.
Start with repetitive work
The first thing I look for is repetition.
Not every repetitive task should be automated, but repeated work is often where AI can help. Examples might include summarising enquiries, turning meeting notes into actions, drafting first responses, checking documents against a template, producing standard reports, researching prospects, classifying support requests, or moving information between systems.
The important point is to be specific. “We spend too much time on admin” is too vague. “Every Friday, someone spends two hours copying figures from three systems into a management update” is useful. “A fee earner rewrites the same client explanation twenty times a month” is useful. “The operations manager is the only person who knows how to prepare the monthly compliance pack” is useful.
AI works best when the job is clearly described. If the process is vague in human terms, it will usually be vague in AI terms too.
Then look for bottlenecks
The next question is: where does work queue up?
In many Gloucestershire SMEs and professional firms, the bottleneck is not a formal process. It is a person. A business owner reviews everything before it goes out. A senior manager answers every unusual customer query. One administrator understands the old system. One partner holds the client context in their head.
AI is not a replacement for judgement, but it can often reduce the load around that judgement. It can prepare a first draft, gather the background information, flag missing details, compare something against a checklist, or produce a short summary so the human decision is faster.
That distinction matters. A sensible first AI project often sounds like: “help the expert get to the decision faster”, not “remove the expert from the decision”.
Check the data sources
Before recommending any AI tool, I want to know where the information lives.
Is it in email? A CRM? Excel files? Word documents? PDFs? SharePoint? A practice management system? A folder on someone’s desktop? A website? A database? Public sources?
This matters because AI is only useful if it can access the right information safely and reliably. If the business wants an AI assistant to answer client questions, but the knowledge is scattered across old proposals, staff inboxes and half-finished notes, the first project may be to tidy the source material, define what can be used, and create a controlled knowledge base.
It also matters for privacy. Some information should not be pasted into a public AI tool. Some data needs permission controls. Some work needs a human review step before anything leaves the business. An AI audit should identify those issues early, before enthusiasm turns into accidental risk.
Understand the existing software
AI should fit the business, not the other way round.
If a team already runs on Microsoft 365, Google Workspace, Xero, HubSpot, Sage, Clio, Salesforce, Airtable, WordPress or a sector-specific system, the first question is whether AI can improve the workflow around those tools. There is no point introducing a shiny new platform if it creates another place for staff to check.
I am looking for practical connections: enquiries summarised before they reach the CRM, documents checked before filing, or a weekly report drafted from existing data. A quiet workflow that saves time every day is often better than an impressive demo that nobody adopts.
Assess staff skills and confidence
An AI audit is not just technical. It is also about people.
Some staff will already be experimenting. Some will be worried about being replaced. Some will not trust AI at all. Some will trust it too much. All of that affects the right recommendation.
A useful AI project needs clear rules: when to use it, when not to use it, what must be checked, what data is allowed, and who is responsible for the final output. Staff do not need to become AI experts, but they do need enough confidence to use the system properly and enough scepticism to spot problems.
In practice, training and guardrails are often as important as the tool itself.
Look carefully at risk areas
The best AI opportunities are not always the safest first projects.
If a process involves legal advice, medical information, regulated financial work, confidential HR matters, sensitive client data or reputational risk, I would be cautious. AI may still help, but usually with tighter controls: internal drafting only, clear human approval, audit trails, restricted data access, and careful measurement.
For many businesses, a better first project is useful but low-risk: summarising internal meetings, drafting non-sensitive documents for review, building a staff knowledge assistant, improving research, creating a reporting workflow, or helping triage enquiries before a human responds.
A local example: Cheltenham Times
My Cheltenham Times experiment is a useful example of why the audit mindset matters.
The project was not simply “use AI to write articles”. The AI-assisted workflow had to check local sources, find leads, sort information, draft content, create images, categorise posts, update useful pages and suggest improvements. It also needed rules around geography, source quality, thin content, image suitability, cost and human oversight.
The interesting lesson was not that AI could do everything perfectly. It could not. Mistakes led to tighter checks: better routing between Cheltenham and wider Gloucestershire stories, safer image review, clearer draft and review stages, and more thought about which ideas should become useful tools rather than one-off posts.
That is how I think businesses should approach AI. Do not start with hype. Start with the work, the data, the risks and the people. Then choose the tool.
A simple next step
If you run a business in Cheltenham or Gloucestershire and are wondering where AI could genuinely help, start with a short list:
- three repetitive tasks that take too long;
- three bottlenecks where work waits for one person;
- the main systems where your useful information lives;
- the areas where mistakes would be costly;
- the staff who would actually use the workflow.
That list is a better starting point than a tool comparison.
If you would like help turning that into a practical AI plan, book a short AI consultancy call. We can look at your current workflows, identify a sensible first project, and work out what AI could do in your business without adding unnecessary complexity or risk.
