Most business owners have now tried AI in some form. They have asked ChatGPT to write an email, summarise a document, tidy up a proposal or suggest a few marketing ideas.
That is useful, but it is still a long way from the more interesting question: what happens when AI is given a proper job?
That was the point of the Cheltenham Times experiment. The aim was not to create another quick AI demo. It was to see what happened when an AI agent was asked to help build and run a local news and information website for Cheltenham, with real tools, real judgement calls and real mistakes to learn from.
The business lesson is simple: if you want AI to do useful work, treat it less like a magic chatbot and more like a digital employee. Give it a clear job, the right tools, training, oversight and sensible limits.
Start with a clear job
A vague instruction produces vague results. “Use AI in the business” is not a plan. Neither is “automate our admin” or “make us more efficient”.
In the Cheltenham Times case study, the job was specific enough to test properly: build a local news and information website for Cheltenham and start finding useful local stories and resources.
That job had a recognisable output. The AI-assisted workflow could search for local information, prepare draft posts, create featured images, add categories, handle basic SEO, monitor possible story leads and suggest future improvements.
For a Gloucestershire business, the equivalent might be: monitoring tender opportunities, keeping a market page updated, turning technical notes into draft client updates, checking competitor websites, collecting common customer questions, or preparing first-draft reports from approved sources.
The useful starting point is not “what AI tool should we buy?” It is “what job would be valuable if it happened reliably every week?”
Give the AI the right tools
A person cannot do a job properly without access to the systems, information and processes they need. AI is the same.
The Cheltenham Times workflow was not just a writing prompt. It involved source checking, WordPress publishing workflows, image generation, categorisation, local relevance checks and automated updates for useful sections of the site. Over time, it also helped create practical local resources, including pages around fuel prices, property information, planning applications, clubs, events and school catchment areas.
That matters because the most valuable AI work is often not “write me some content”. It is scanning, sorting, checking, updating, drafting and turning scattered information into something useful.
In many SMEs, jobs like this are hiding in plain sight. Someone checks the same websites every Friday. Someone manually copies information between systems. Someone keeps a spreadsheet alive because no one has had time to replace it. Someone answers the same customer questions again and again.
AI can help with some of that, but only if it is connected to a sensible workflow. A chatbot on its own is rarely enough.
Expect a training period
One of the biggest misconceptions about AI is that it should work perfectly as soon as it is switched on.
That is not how most real business systems work. A new employee needs induction. A new process needs refinement. A new CRM needs configuration. An AI agent is no different.
In the Cheltenham Times experiment, the AI needed steering. It had to learn what counted as Cheltenham and what belonged on a wider Gloucestershire site. It needed rules on which sources were reliable enough, which stories were too thin, what tone was acceptable and when an image was misleading.
That is a useful lesson for any business. AI implementation is not just a technology task. It is a management task.
You define what good looks like. You correct mistakes. You create examples. You add rules. You decide what needs human approval. The system gets better because the business takes responsibility for training it.
Keep human oversight where it matters
AI can move quickly. That is part of its value, but also part of the risk.
In the Cheltenham Times project, some posts and images needed review. A story could be technically well written but not locally relevant enough. An image could look plausible but give the wrong impression. A source could be too thin to support a useful article.
Those are not reasons to avoid AI. They are reasons to design the workflow properly.
For most SMEs, the right approach is to let AI gather, structure or draft the work; route anything important through a review queue; measure what improves; then tighten the rules before increasing automation.
This is especially important for customer communication, legal or financial content, recruitment, health-related information, sensitive client data and anything that could affect reputation.
The aim is not to remove judgement. It is to free up human judgement for the places where it matters most.
Measure usefulness, not novelty
AI projects often go wrong because they are judged by how impressive they look in a demo, not by whether they make the business better.
The Cheltenham Times experiment became more interesting when the AI was helping create useful assets, not just articles. A fuel price page, planning information or a school catchment map has a different kind of value from a one-off post. It can be updated, reused and improved.
That is a good way for business owners to think about AI. Did it save time? Did it improve consistency? Did it help us respond faster? Did it create something customers or staff actually use? Did it reduce a bottleneck?
If the answer is no, it does not matter how clever the tool looks.
The practical takeaway
The Cheltenham Times experiment is not a recommendation that every business should build a local news website. Most should not.
The useful lesson is that AI can now take on defined, repeatable, information-heavy jobs when it is managed properly. It needs a clear role, access to the right tools, examples of good work, review points and continuous improvement.
That is a much more practical way to think about AI than asking whether it will “replace people”. In most small businesses, the better question is: what useful work is not being done because it is too repetitive, too messy or too time-consuming?
That is where a well-designed AI workflow can help.
If you run a business in Cheltenham, Gloucestershire or the surrounding area and want to explore this sensibly, start with one job. Not a grand AI transformation programme. One clear job with a business reason behind it.
If you would like help finding that first useful AI opportunity, book a short AI consultancy call and we can look at what a digital employee could realistically do in your business.
