Priya runs a six-person letting agency in Manchester. She’s got a ChatGPT tab open, a half-read newsletter about “agentic workflows,” and a nagging sense she’s behind.
But every time she sits down to begin, the choice feels too big, so she closes the tab and answers emails instead.
There’s a simpler way in, and it takes about a month. Across the UK, only around 1 in 6 businesses (16%) currently use any AI technology, and in the US the figure runs close to 17.7% of firms, so most owners stand where Priya does.
The plan below moves in four weekly steps:
- In week one, you map your week and pick one repetitive task.
- In week two, you choose one tool and run it on that task every day.
- In week three, you add human checks and a second use case.
- In week four, you measure what changed, write down what worked, and decide what to keep.
Most adopters who do this report higher productivity (75%), and the common starting points are marketing, administration, and IT (64-72%). Revenue tends to move later, so don’t expect it in month one.
Week 1: Map your time, pick one task
Before choosing any tool, track where your hours go for five working days. Note every recurring task: drafting listings, chasing invoices, writing the same three emails, summarising calls.
Then pick one task you do often—one that uses words more than judgement, and that won’t cause harm if a first draft needs editing.
This step addresses the most common reason businesses stall. In the UK survey, 71% of non-adopters named a lack of identified use for AI as their barrier, ahead of limited skills at 60%. So you’re not choosing a strategy yet. You’re choosing one job.
Most owners (71%) spend around a year thinking about AI before they deploy it. By starting from a task you already understand, you’re compressing that into a week.
Week 2: choose one tool, run it daily
Pick one general assistant and stay with it for the week. There’s no need to build anything custom. Among UK adopters of generative AI tools, many buy a ready-to-use tool rather than developing their own, and the off-the-shelf options cover most early needs.
Run your week-one task through the tool every working day. Draft the listing, then edit it. Summarise the call, then correct it. Keep a running note of the prompts that worked, so you build a small playbook as you go.
By Friday you’ll have five days of output and a clear sense of where the tool helps and where it slips.
| Recurring task | Good first fit | First prompt to try |
| Drafting listings or product copy | General writing assistant | “Draft a 120-word listing from these notes: [paste]” |
| Turning calls into notes | Assistant with summarising | “Summarise these notes into 5 bullets and 3 actions: [paste]” |
| Answering repeat customer emails | Assistant plus saved templates | “Write a polite reply to this enquiry in our tone: [paste]” |
| Cleaning a list or spreadsheet | Assistant that handles tables | “Standardise these rows into name, date, and amount columns: [paste]” |
Week 3: add oversight, add a second task
Now you slow down and add a check. In the UK 67% of AI users apply significant human review to outputs, and that habit keeps errors out of client-facing work.
Set one rule for yourself: nothing goes to a customer, a tax form, or a contract without a person reading it first.
Then add a second task from your week-one list. Keep it close to the first. If week one’s task was email drafting, week three’s might be turning notes into a summary. The goal is to widen use slowly.
Week 4: measure, document, decide
Compare week four with week one. How long did the task take before? How long now? You don’t need precise figures. A rough before-and-after will show you the direction.
Expect productivity to move before money does. In the UK survey, 77% of AI users reported no revenue change yet, even as most saw output rise.
US data points the same way, with more than 80% of small businesses using AI reporting productivity gains. So judge month one on time saved and quality held.
Then write a one-page note: the two tasks, the tool, the prompts that worked, and your review rule. That’s now your standard. Hand it to the next person who joins, and your first month becomes a system you can repeat.
| Week | Focus | Main output |
| 1 | Map your time, pick one task | A shortlist and one chosen task |
| 2 | Run one tool daily | Five days of edited output and a prompt note |
| 3 | Add oversight, add a second task | A review rule and a second use case |
| 4 | Measure and document | A before-and-after and a one-page standard |
Work with me
Making AI pay across the whole business takes three steps, and I can run them with you.
- Audit. I map every task in your business against the week-one test, then rank them by time saved and risk. You get a short list of the tasks to automate first (see how I’ve done it before here).
- Workflows. I turn the winning tasks into documented, repeatable systems, with the prompts, the tools, and the review rule built in.
- Team training. I train your people to use the tools well, with shared prompts and one standard for human checks.