“How can I use AI in my business?” is the question we are asked most often, and the candid first answer is: not in the way the advertisements suggest. The advertisements begin with a tool. The businesses that obtain real value begin with a specific task.
This article is therefore not a list of tools. It sets out the method Team Agentic uses with owners of businesses of 10 to 250 people, in a form you can follow yourself. There are five steps, and we will take each in turn.
Step 1. Identify the tasks, not the technology
Ask each team one question: which task recurs every week and takes longer than it should? The question is not “where could we use AI”, which invites speculation, but which task consumes a disproportionate share of the working week. The answers are generally routine, and that is precisely the point: answering the same enquiries individually, rebuilding the monthly report from the same spreadsheet, writing every proposal from scratch, or re-entering supplier invoices into the accounts.
List them with an approximate figure beside each: hours per week, and who spends them. Precision is not required at this stage. What is required is a short list.
Step 2. Question each task before you automate it
It would be easy to take the most time-consuming task and hand it to AI. Automating a task exactly as it is performed today, however, simply means the unnecessary steps are carried out more quickly. For each item on the list, ask three questions. Does this step need to exist at all? Could a simpler change – a template, a form or an automatic acknowledgement from your own email system – remove most of it? And if AI prepared the first draft, who would check it?
Some tasks will disappear at this stage, which is a welcome outcome. There is nothing more wasteful than automating work that should not exist.
Step 3. Select one task and define success
Select one task, not three. Choose the task where the time cost is clearest and the output is easiest to check. Then define a good result using figures you already hold: hours per proposal from brief to sign-off, hours to first reply on an enquiry, or days to produce the management pack.
That figure is your starting point. Record it before anything changes, so that you can later judge whether the change was worthwhile. Most AI projects omit this step, and six months later nobody can say what they have achieved.
Step 4. Test on your own examples before relying on it
Whichever tool you use, provide it with your actual past work: last year’s proposals, a month of enquiries, three months of invoices. Compare its output with the work your people produced. Where it is wrong, establish the cause and correct the instructions or the source material, rather than hoping it will improve. If the drafts are accurate but do not sound like your business, a brand kit gives the tool your tone of voice and your templates.
Then run a short pilot with the people who carry out the task, with a named person checking every output before it is used. Your people own the first and last step. AI does the heavy lifting in between. If the pilot does not improve on the starting point, stop: you will have spent a few weeks rather than a year.
Step 5. Measure the result and decide the next step
After the pilot, compare the figures with the starting point. If the task is faster, better or less costly, retain the agent, assign it an owner and write a one-page guide so that it continues to operate through any change of staff. If not, you have learned something at little cost. In either case, the next task on your list remains, and you now have a method for approaching it. Once an agent is in daily use, our Managed service can keep it working as the tools change.
Two matters to settle first
Data, and tools. Before the first pilot, decide which AI tools your business approves and what data may and may not be entered into them, and inform everyone. A single page is sufficient. It is likely that someone in your organisation is already using a free tool containing customer details, and this is the moment to address it. We have written separately about a simple AI governance framework for SMEs.
How Team Agentic can help
If you would like to apply this method with someone who has run a business, our AI Quick Scan is designed for that purpose: a 90-minute conversation about how your business operates, followed by a two-page summary of up to five areas where AI, or a simpler change, could help, including what is not yet worth doing. If you have already identified the task, we will omit the scan and scope the build. An initial call is free of charge, and if you do not need our help, we will tell you so.
