Once people understand roughly how AI tools work, the next question is always the same: which one should we actually use? The honest answer is less dramatic than the headlines suggest, but there are real differences, and there is a bigger shift happening underneath that is worth understanding now. Here is the plain-English version, as things stand in July 2026.
The four main names
ChatGPT, from OpenAI, is the household name and the most widely used. It moves fast, adds features constantly, and does a bit of everything: writing, analysis, images, voice, and research. For many people it simply is AI, in the way Hoover became the word for vacuum cleaners.
Claude, from Anthropic, has built its reputation on careful, natural writing and strong work with long documents, and it has become a favourite for business writing and for software development. Anthropic leans heavily on safety research, which shows in how the tool behaves with sensitive or ambiguous requests.
Gemini, from Google, is woven into the Google world. It is strong on questions that benefit from live search, it handles images and video well, and if your business runs on Google Workspace it turns up inside Gmail, Docs and Sheets, where you already work.
Copilot, from Microsoft, is less a separate destination and more AI threaded through Word, Excel, Outlook and Teams. For businesses that live in Microsoft 365, it is the path of least resistance: nobody has to open a new tool to use it.
There are others, and some are very capable, but for most UK small and medium-sized businesses these four are the realistic shortlist.
More alike than the marketing suggests
Here is the part the comparison articles rarely say plainly: for everyday business work, all four are good, and they leapfrog each other every few months. A league table written in January is out of date by summer. If you are choosing an assistant for drafting, summarising, analysis and general thinking work, you will get real value from any of them.
So the choice usually comes down to more practical questions than “which one is cleverest this month”.
Where does your business already live? If you are a Microsoft 365 business, Copilot arrives inside the tools your staff already use every day. If you run on Google Workspace, Gemini does the same. That convenience matters, because the best AI tool is the one people actually use.
What happens to your data? This is the question that should be asked before anything is rolled out. The business versions of all four offer proper data controls, including commitments not to train on your content, but those protections belong to the business plans, not the free personal accounts your staff may already be using quietly. Getting this right is a governance question, and it is exactly the ground an AI use policy covers.
Will your team actually use it well? The gap between businesses that get value from AI and those that do not is rarely about which provider they picked. It is about whether anyone showed the team what good use looks like. One well-adopted tool beats three ignored ones.
My honest advice is to pick one main assistant on a business plan, match it to the ecosystem you already work in, and revisit the decision once a year. Chasing every new launch is a hobby, not a strategy.
The bigger shift: from answering questions to doing the work
While the chatbots trade places in the league tables, the more important change is happening elsewhere. The newest AI tools are agents: instead of answering a question, you give them a whole task, and they work through the steps themselves.
Claude Cowork is the clearest example for non-technical staff. It sits in the Claude desktop app on the paid plans, and rather than chatting with it, you point it at a folder and give it a job: tidy up these files, pull the key terms out of these contracts, build a first-draft report from these documents. It then works through the task on its own while you do something else, coming back with the finished result. It is early days, and the same rules apply as ever, in that you stay in charge of the decisions that matter and you check the output. But for the repetitive, document-heavy work that eats admin hours, this style of tool is a genuine step change, and it is now spreading beyond the desktop app to the web and mobile.
Claude Code and Codex are the same idea applied to software. Both are, in effect, AI developers you can brief. Claude Code, from Anthropic, works alongside a developer in their own environment, reading a whole codebase and making coordinated changes under supervision. Codex, from OpenAI, leans further towards delegation: it can be briefed, go away and do the work in the cloud, and come back with finished changes for review.
If you never write software, here is why this still matters to you. Whoever builds or maintains your systems, whether that is an in-house developer, an agency or a freelancer, is very probably using one of these tools already. Used well, they make good developers meaningfully faster. But they also make it very easy to produce software that nobody has properly reviewed. If you commission software, the question to ask has not changed, but it has become more urgent: how is this code reviewed and tested before my business depends on it? I have written before about why due diligence still matters with AI-built software, and everything in that piece applies double now.
What I would actually do
If you are starting from scratch, the order is this. Decide where your business lives, Microsoft or Google or neither, and let that shortlist the assistant. Put it on a proper business plan so the data protections apply. Write down a few plain rules for staff use before rolling it out, not after. Show the team how to use it well, with real tasks from your own business, and then give it a few months of honest use before judging.
And keep half an eye on the agent tools. You do not need to adopt them today, but the businesses that quietly automate their document-heavy admin over the next couple of years will mostly be using tools like these, and it will not require a technical team to do it.
If you want the groundwork first, my earlier piece on AI models and tokens explains the machinery in plain English, and the free prompt library on this site is a practical place to start. And if you would rather talk through what this looks like for your business specifically, that is precisely what an AI discovery call is for.