1. Choose a business problem you can describe
“We need AI” is difficult to act on. “We spend time drafting similar replies, and they still need checking” gives you something to investigate. Start with a real example of work that takes too long, repeats frequently or gets delayed.
Write down who does the task, what information they need, which systems they use and what a correct result looks like. Ask the people doing the work where the friction occurs.
2. Decide whether AI is needed at all
If the task follows clear rules, a template, a reminder or an existing software feature may solve it. AI is more relevant where language or interpretation is involved, but its output may be incomplete or wrong.
Compare the simplest viable options. Include the time to prepare information, review outputs, handle exceptions and maintain the workflow. The most impressive demonstration is not necessarily the best investment.
3. Agree information, permission and review
Use dummy or redacted examples while exploring an idea. Before using real business or client information, agree approved tools, permissions and handling requirements. Do not assume an ordinary consumer account is suitable for confidential data.
Keep human approval where errors could affect customers, payments, safety or important decisions. If professional advice is required, involve the relevant adviser.
4. Measure a small trial
Record the current frequency, time taken and error rate. Test with a manageable set of examples, including awkward cases, and record both the benefit and the effort needed to check the result.
Illustration only: a ten-minute task done twenty times a month represents 200 minutes of existing work. Even a successful trial cannot save more than that baseline, and setup, checking and running costs still need to be counted. This is not a client result or a savings promise.
5. Keep, change or stop the idea
A useful trial ends in a decision. Keep it if it improves the task at an acceptable cost and risk; adjust it if the problems are fixable; stop it if the business case is weak. Document who owns the workflow and how the team falls back to the manual process.
If you want someone to work through this with you, I offer a free first conversation followed by a separately quoted review or implementation. You remain in control of whether to proceed.
Your questions, answered
What should I prepare for a first AI consultancy call?
Bring one task, a description of how it currently works and the outcome you want. You do not need to send confidential customer or staff records with an initial enquiry.
What is a sensible AI budget?
It depends on the task, risk, systems and support needed. Establish the potential benefit and compare setup and recurring costs before setting a project budget. Director’s Edge quotes paid work before it begins.
Can I share this with a business owner I know?
Yes. It is designed to help start a practical discussion. The owner should decide whether to contact an adviser and what information they want to share.