Treat adoption as a change to work.
Oliver Wyman’s analysis argues that scaling agentic AI requires attention to workflows, ownership and controls, beyond selecting a platform. That is an enterprise perspective, not evidence of what a particular small business should spend or achieve.
The UK government’s 2026 AI adoption survey examines how firms use AI and the barriers they report. Its findings are a starting point for questions, not a forecast of your business’s results.
Our application for a smaller team is deliberately bounded: choose one workflow, describe its inputs and outputs, identify the person responsible and compare a trial with the current way of working.
Start with the task and its exceptions.
For an enquiry-handling workflow, ask where information arrives, how it is checked, who writes the response and which exceptions require judgement. If the real problem is incomplete inputs or unclear ownership, document that before choosing a tool.
Include the time spent checking and correcting output in the review. A faster first draft is only one part of the work. The useful comparison is the complete workflow, including preparation, review, rework and handover.
A small trial with a clear review.
| Planning question | Illustrative answer |
|---|---|
| Which task? | Drafting a response to a common customer enquiry. |
| Which inputs? | An agreed enquiry example and approved product information. |
| Who checks it? | A named team member reviews accuracy, completeness and tone before use. |
| What will be compared? | Total preparation and review time, corrections, unresolved questions and tool costs. |
| What ends the trial? | A review against agreed criteria, followed by a decision to stop, revise or extend. |
This is an original planning illustration, not a client implementation, benchmark or claim of savings.
Define a useful output from the review.
A first piece of work could be a workflow map, requirements brief or options assessment. It should explain the problem, the proposed trial, who is responsible and what would justify the next investment. Implementation should follow the scope and support the business actually needs.
Choose one repeated task. Write down its inputs, owner, exceptions and review step. Then define what a useful trial would need to demonstrate before you buy or expand a tool.
Sources and interpretation
Oliver Wyman: Why IT operating models need to change before agentic AI ↗
GOV.UK: AI adoption survey report 2026, executive summary ↗
Accessed 8 November 2026. The small-business application and worked trial plan are Tresswell editorial analysis. Source links do not imply endorsement or affiliation.