I built Cheltenham Times as an experiment, but not the sort of AI experiment that lives in a slide deck.
I wanted to know what would happen if an AI agent was given a real job: build and help run a useful local news and information website for Cheltenham.
That meant dealing with WordPress, story ideas, local sources, SEO, images, useful tools, categories, tone of voice, and all the awkward judgement calls that come with publishing anything in the real world.
The result became Cheltenham Times. More importantly, it gave me a practical view of what AI agents can and cannot do for a small business.
The useful bit was not “AI writing articles”
It is easy to reduce this sort of project to: “Can AI write content?”
That is the least interesting question.
The more useful question is: what recurring job could an AI-assisted workflow do every day if it had the right instructions, tools and supervision?
For Cheltenham Times, the job was not just writing. The early workflow could look for local story leads, assess whether they were relevant to Cheltenham, turn suitable items into draft posts, create featured images, add categories, handle basic SEO and publish into WordPress.
At the start, it was producing around five to seven posts a day. Some were useful. Some needed editing. Some should not have been published at all.
That is the first lesson for any business owner: AI is fast, but fast is not the same as right.
Treat an AI agent like a very fast junior assistant
The agent needed steering.
I had to teach it what counted as Cheltenham, which sources were reliable, what tone was acceptable, when an article was too thin, when an image was misleading, and when a story belonged somewhere else.
One early mistake was geographic. A story about Stroud may be relevant to Gloucestershire, but it does not automatically belong on a Cheltenham-focused site. Another issue was imagery: a generated or selected image can look plausible while still being wrong for the specific story.
That is exactly why I do not talk about AI agents as replacements for judgement. A better comparison is a fast junior assistant. It can do a lot of work, but it needs rules, examples, correction and review.
For an SME, this is where the value often sits. You are not trying to “automate the business”. You are trying to take a regular, messy process and make the first draft, first search, first sort or first check much faster.
The system mattered more than the prompt
A single clever prompt would not have built Cheltenham Times.
The useful progress came from building a workflow: sources to monitor, rules for what counted as useful, WordPress actions, SEO checks, image checks, and human review points.
The site does not rely on one source. The AI-assisted workflow can look at local websites, public feeds, council updates, events pages, business news, community sources and social signals. It treats those as leads, not guaranteed truth.
That distinction matters. AI is useful for scanning, sorting, summarising and suggesting. It is not a magic truth machine.
A practical business workflow needs questions built into it:
- Is this relevant?
- Is the source good enough?
- What should happen next?
- Does a human need to approve this?
- What would be embarrassing if it went out unchecked?
Those questions are not technical extras. They are the guardrails that make the system usable.
The biggest lesson: build useful assets, not just content
One of the best parts of the experiment was discovering that AI agents are not only useful for producing articles.
For a local site, ordinary posts have a short shelf life. Useful tools and reference pages can keep helping people for months.
Cheltenham Times added pages for fuel prices, property prices, rental prices, planning applications, school catchment areas, local clubs and events. The fuel price page briefly pushed traffic to around 150 visitors a day after being shared locally. It was a small result, but it proved the point: people respond when a site solves a real problem.
That lesson applies well beyond local news.
An estate agent might build a local moving checklist or market update workflow. A professional services firm might build a client onboarding assistant. A manufacturer might build an internal troubleshooting guide. A care business might build a rota-checking or enquiry-triage process.
The important question is not “what can we generate?” It is “what useful thing could we maintain, update or improve with AI support?”
Privacy, cost and control matter from the beginning
I set three rules for the Cheltenham Times experiment.
First, the setup had to be as private and secure as possible. Second, the running cost had to stay low enough to make sense for a small business. Third, the work had to create something with actual value, not AI content for its own sake.
Those three rules are a good starting point for most SMEs.
Before adding AI to a workflow, ask:
- What data will the system see?
- Which tools and accounts can it access?
- How much will it cost to run each month?
- Where does human approval sit?
- How will we know whether it is saving time, improving quality or creating new value?
If those questions are ignored, AI becomes a risk. If they are answered clearly, AI becomes much more manageable.
What this means for your business
Cheltenham Times is still an experiment. It still needs human judgement. It still changes when something goes wrong.
That is why it is useful.
It shows that AI agents can do real work when they are given a clear job, connected to the right tools and managed properly. It also shows that the messy parts — judgement, tone, quality, relevance, privacy and cost — do not disappear.
For most Gloucestershire businesses, the best first AI project will not be dramatic. It will probably be a practical workflow that removes friction from something you already do: responding to enquiries, preparing reports, checking documents, researching leads, updating content, or organising information.
Start there. Pick one repeatable job. Define what good looks like. Keep a human in the loop. Measure whether it helps.
If you run a business in Cheltenham or Gloucestershire and want to explore what a sensible AI agent or workflow could do for you, book a short AI consultancy call. We can look at where time is being lost now, what is safe to automate, and what a first useful project might look like.
