You can describe a website to an AI tool and get something recognisable back remarkably quickly. That changes the starting point for a business owner. A rough idea can become a working layout before you have written a formal brief, and a small, straightforward project may be something you can manage yourself.
It does not follow that every business now needs a developer, or that nobody does. The useful question is what the website must do, what could go wrong and who will take responsibility for it. Looking finished and being ready to rely on are different stages of the same project.
Start with the job, rather than the tool
Imagine a self-employed gardener who needs a few service pages, photographs and a clear telephone number. A hosted builder, with carefully checked content, might be a sensible starting point. Now imagine a supplier with trade prices, different customer accounts, a stock system and orders arriving from several channels. Both need a website, but the decisions and failure points are very different.
Before choosing how to build, write down the journeys that matter. Can someone understand your offer? Can they book, buy or enquire? Where does the information go afterwards? Who changes the content? A website design project should start with those questions, even when AI helps produce the first layout.
AI can help with more than a mock-up
AI can draft content, produce code, suggest tests and help investigate problems. It would be misleading to dismiss everything it makes as a cheap imitation. Used with suitable tools and context, it can contribute useful work throughout development. The quality depends on the task, the information supplied, the system around it and how the result is assessed.
Anthropic’s discussion of trustworthy agents places human control, user expectations and security alongside useful autonomy. That is a more helpful way to think about business use than assuming that a convincing answer is an assurance that a connected system will behave correctly.
For an owner, the practical benefit may be a quicker way to explore an idea. For a developer, it may be faster investigation or a first implementation to review. Neither use removes the need to establish what success actually looks like.
The difficult part is often the detail behind the page
A contact form is an example. The visible fields are straightforward. The working feature also needs validation, spam handling, understandable errors and a dependable destination for the enquiry. If it connects to a CRM, someone must decide how duplicates are handled, what happens during an outage and whether a successful submission really created the intended record.
A developer helps turn those decisions into a system that can be tested. Our web development work can include the integrations and business rules behind the interface, rather than stopping when the screen looks right. The same principle applies to stock, bookings, customer accounts and payment providers.
These are not reasons to make every project complicated. Often the right decision is to remove an unnecessary feature or use an established service. Knowing when to keep something simple is a useful part of the work.
Someone still needs to check the result
Testing should follow real business tasks. On a shop, that means checking delivery options, variations, payment outcomes, customer emails and what the team sees afterwards. On an enquiry website, it means following a realistic enquiry from the first page to its destination. A desktop screenshot does not show whether a phone user can finish the same journey.
- Check the pages on a small screen and with a keyboard.
- Try incomplete, incorrect and repeated submissions in an agreed test environment.
- Review accessibility, loading behaviour and useful search settings.
- Check that the business facts, images and claims are accurate.
- Record the change and a working recovery route.
AI can help create and run some checks, but the tests need to represent the business. If the brief never explains a trade-pricing exception, an automatically generated test may confidently confirm the wrong behaviour.
A website is something you operate after launch
Once a site is live, software, services and content keep changing. Someone needs to know which integrations matter, where accounts are held and what the hosting arrangements include. A plugin update, an expired subscription or a change in an external API can affect a feature that worked on launch day.
This is where ongoing maintenance and clear ownership become valuable. Ask who receives alerts, checks important journeys and can restore a working version. Keeping access organised and maintaining a recovery copy are practical responsibilities, whether the original code came from a person, a generator or a mixture of both.
Even WordPress.com’s own guide to building themes and plugins with AI recommends testing bigger changes away from the live site and knowing how to recover. The platform’s guidance supports a careful workflow rather than treating generated code as automatically production-ready.
When doing it yourself can make sense
A simple requirement, a supported platform and time to learn can make a self-managed build a reasonable choice. Be honest about the time involved, including writing content, choosing images, checking mobile layouts and keeping the site current. A low initial software cost does not tell you how much attention the site will need.
You can also split the work. Explore your idea with AI, bring the working example to a developer and ask for a review of the important journeys. Or commission a sound structure and use an agreed editing process for routine updates. There is room between doing everything alone and handing over every small change.
Ask for responsibility you can understand
When comparing proposals, ask what is included in planning, testing, handover and support. Ask which accounts you will own and how the site can be moved or recovered. An answer about a particular tool should connect with the result you need.
Our AI and automation work starts with useful tasks and suitable boundaries. If you are deciding what to build yourself and where to bring someone in, talk the project through with us. A clear requirement is a better starting point than either fear of AI or a promise that it can do everything.