Skip to main content
AI9 min read

Using AI tools to run a leaner small business

The highest-impact uses of AI are drafting, research, summarisation, and extraction. A practical UK guide with named tools and honest boundaries.

The highest-impact uses of AI in a small business are drafting, research, summarisation, and structured extraction. That is the shortlist. Anything else is secondary, and some of it is a distraction. If you want to run a leaner business using AI, focus the investment on those four capabilities, pick a small number of tools that do them well, and resist the urge to adopt every new product that launches on a Tuesday.

This article names the tools worth using, explains what each is genuinely good at, and is honest about where AI still cannot help.

The four capabilities that pay back

Drafting. The blank page is expensive. Emails, proposals, blog posts, social captions, job descriptions, internal updates. AI will give you a workable first draft in a minute or two. Your job is to rewrite it into something that sounds like you and contains specifics only you know. The time saving is real. The quality ceiling, if you edit properly, is higher than what you would produce alone under time pressure.

Research. Understanding a new topic, summarising a competitor landscape, finding recent data, or preparing for a client call used to take an hour of tab-hopping. Now it takes ten to fifteen minutes of focused prompting, plus a quick verification pass on key claims.

Summarisation. Long documents, calls, emails, and meetings used to require someone to sit through them and write up what happened. AI handles this in seconds. The value is not just the time saved. It is the fact that things now get summarised at all, whereas before they were lost.

Structured extraction. Taking unstructured input, a call transcript, a long email, a customer survey, and pulling out specific information in a consistent format. What objections came up? What dates were discussed? What action items were agreed? AI does this reliably and quickly.

Everything else, generating images, building websites, writing code, acting autonomously, is either narrower in value or more specialised than the four above. Start with what has the best return, then experiment outward.

The tools worth paying for

You do not need a stack of fifteen AI tools. For most UK small businesses, four or five is enough. Here are the ones that earn their place.

ChatGPT is the most capable general assistant, particularly the paid tier at around twenty pounds a month. It is good at drafting, research, structured extraction, and chatting through a problem. The custom GPT feature lets you set up persistent assistants for recurring tasks, such as a proposal draft GPT trained on your tone of voice, or a client intake GPT that turns notes into a standard summary. This is one of the practical applications of the broader shift we describe in how AI is changing small business marketing.

Claude is a close competitor, particularly strong on long documents, nuanced writing, and careful reasoning. I use Claude when the writing quality matters more than speed, and when the input is a long document that needs careful analysis. Many teams pay for both. At twenty pounds a month for each, the combined cost is still less than a single hour of a junior employee.

Perplexity is built for research. It searches the live web and cites its sources, which matters when you are trying to find something specific or recent. It is not a replacement for ChatGPT or Claude in drafting, but it is the right tool for research queries that need current information.

Fathom, or alternatives like Granola, Otter, and Tl;dv, handles meeting recordings, transcription, and post-call summaries. For any small business that runs client or sales calls, this is transformative. You leave a call with a tidy summary, action items, and a searchable transcript, all inside your usual tools. Fathom has a generous free tier that is often enough.

Notion AI or Superhuman AI. If you live in those tools, are less standalone and more embedded. They speed up specific workflows, writing inside Notion, triaging inside an inbox. They are worth it only if you were already paying for the underlying tool.

That is the basic kit. Four to five tools, under a hundred pounds a month combined. More than enough for most UK small businesses.

The tools that sound important but often are not

Short lists of what to be cautious about.

AI phone answering services. Some work, most do not. Test carefully before letting one near customers. The broader rule of keeping a human on first contact is covered in how to automate your small business without losing the personal touch.

AI social content schedulers that promise to run your LinkedIn or Instagram. The output is usually generic and your audience will notice within a month.

Bespoke AI chatbots for your website. Often the investment outweighs the benefit for small sites. A good FAQ page plus a simple contact form converts better in most cases.

Image generators for brand visuals. Useful for mood boards or quick concept art, but do not replace your designer for anything client-facing unless you are genuinely skilled at directing the output.

None of this is to say these tools have no value. They have value in specific contexts. For a first adoption, focus on the high-return categories, and revisit the others once those are embedded.

Concrete workflows that save hours

Five workflows that are quick to set up and return the investment within a month.

The proposal draft. Paste your notes from the client call into Claude or ChatGPT. Prompt, draft a proposal covering scope, deliverables, timeline, pricing, and assumptions. Use our usual structure. Wait. Edit. You turned an hour of writing into twenty minutes.

The meeting summary. Use Fathom or Granola on every client and internal call. At the end of the call, the tool sends a summary, action items, and a transcript to your email or chosen inbox. Time to file a proper meeting note drops from thirty minutes to two.

The competitor scan. Once a quarter, ask Perplexity to tell you what three named competitors are doing right now, what services they are promoting, and what recent changes have appeared on their websites. Twenty minutes of prompting gives you a structured view of the competitive landscape.

The customer research extraction. Collect customer feedback, survey responses, or call transcripts from the last quarter. Feed them to Claude with a prompt, identify the top five concerns, the top three decision drivers, and the exact phrases customers use. The output becomes the basis for landing pages, ads, and sales conversations, which pairs well with how to write website copy that converts.

The tidy-up pass. Before sending any important email, press a shortcut, paste into ChatGPT, and prompt, tighten this, keep the tone, remove filler. The difference in clarity is noticeable, and over time your own writing improves by osmosis.

Where AI still falls down

Be honest about the limits, because they are real.

It confabulates. AI will state things plainly that are wrong. Statistics, dates, case law, medical guidance, financial figures. For anything that matters, verify. Two minutes of checking is worth the insurance.

It lacks taste. AI can produce a competent blog post, but it cannot tell you which of three ideas is the right one for your audience. The editorial judgement stays human.

It does not know you. Without context, AI produces generic work. You must feed it your specifics, your tone of voice, your past examples, your actual situation. The more context you supply, the better the output.

It performs worse on recent events. Training data has a cutoff. For current UK events, legislation, or pricing, use the tools with live search, Perplexity, ChatGPT's search mode, Claude's web search. Do not trust a plain chat for fresh information.

It struggles with your data safely. Unless you are careful, anything you paste into a free tier of an AI tool may be used for training. For sensitive client data, either use business tiers that contractually exclude training, or strip identifying details before pasting.

Setting up AI in your business without chaos

A simple four-step adoption path.

First, pick two tools. One general-purpose, ChatGPT or Claude. One specialist, Fathom for meetings, or Perplexity for research. Pay for them. Use them personally for two weeks until they are habit.

Second, write down five tasks where AI could save you an hour a week. These are your beachheads. Turn each into a short workflow, documented, with the exact prompt you use.

Third, share the workflows with your team. Teach by doing, not by lecture. Sit with one person at a time for thirty minutes and show them the workflow. Uptake from peer-led adoption is twenty times higher than uptake from top-down emails announcing a new tool.

Fourth, review monthly. What is working? What is not? What new tool is worth trying? Cancel anything not earning its place. Most AI subscriptions have a shelf life, and pruning is part of the hygiene.

A realistic expectation of gains

A small UK business that adopts AI thoughtfully, targeting the four high-impact capabilities, can reasonably expect to save between five and fifteen hours per person per week. Not from any single dramatic workflow, but from dozens of small reductions in friction. Faster drafts. Quicker research. Better meeting records. Cleaner customer data.

Those hours are not automatically repurposed into growth. If you do not deliberately redirect them to higher-value work, they get absorbed into more email, more meetings, more admin. Plan what to do with the time you are about to save, or the saving disappears into the general fog of working life. This is one of the most common failure modes we describe in what small businesses get wrong about digital transformation.

The honest summary

AI is a productivity lever, not a strategy. It lets a small team punch above its weight on the tasks that used to be disproportionately slow. The businesses that benefit most are the ones that stayed ruthlessly focused on the high-impact uses and the handful of tools that deliver them.

If you are weighing up where AI fits best in your specific setup, or you want a second opinion on what to prioritise, our business audit is a quick starting point. Or book fifteen minutes and we will look at your situation directly.

Steffen Hoyemsvoll

About the author

Steffen Hoyemsvoll

Founder of Voll. Oxford Physics, ex-fintech co-founder, Chartered Wealth Manager. Writes about what he actually uses to grow small businesses.

Work with Steffen

Keep reading