How can I automate my business using AI?

Most attempts at automation stall because they begin with a product: someone signs up for an AI assistant and then looks for something for it to do. It works better the other way round. Pick a piece of work that already happens over and over, costs someone real time, and follows a recognizable pattern.

Good starting points tend to sit between systems or between people: an enquiry that arrives by email and has to end up in the CRM, a booking that needs a reminder, a question staff answer by digging through shared folders. If you are unsure which process to pick, what tasks to automate first sets out the criteria.

1. 01Choose one repeated process. Name it precisely: "new website enquiries", not "sales".
2. 02Map how it runs today. Write down where the work arrives, who touches it, what they look at, what they decide and where the result goes. Note the exceptions, because they are where automations break.
3. 03Decide what stays human. Mark the steps that need judgement, carry risk or involve a sensitive conversation. The system prepares and routes; a person approves or handles those steps.
4. 04Connect to the tools you already use. The automation should read from and write to your existing email, forms, calendar, CRM and documents rather than adding another place to check.
5. 05Pilot one workflow. Run it on real work for a limited period, with a person checking the output, and keep a record of what it got wrong.
6. 06Measure the result. Compare time spent, response speed and error rates with how the process ran before, then decide whether to keep, adjust or drop it.

Only after one workflow is running reliably is it worth moving to the next. Each new workflow is easier because the connections and review habits are already in place.

Enquiry handling

Reads each incoming message, works out what the person wants, pulls out the useful details (name, service, location, urgency) and sends it to the right person with a short summary and a suggested reply for review.

CRM updates

Creates or updates the contact record from a form, email or call note, logs the interaction and sets the next step, so the CRM stays current without someone retyping information.

Scheduling follow-ups

Watches for events such as a quote sent or an appointment completed, then schedules a reminder, a check-in email or a task for a staff member at the right time.

Internal document search

Lets staff ask a question in plain language and get an answer drawn from the business's own policies, manuals and past documents, with a link to the source so they can check it.

Several of these are plain automations with a small AI step in the middle, usually the reading and sorting of free text. Our business automation work is mostly this kind of combination.

Imagine a physiotherapy clinic whose front desk spends much of the morning answering web enquiries and phone messages. Many are booking requests, some are insurance questions and a few are clinical questions that a physiotherapist should answer.

A first workflow could read each enquiry, sort it into one of those groups, add the contact to the clinic's system and draft a reply for booking and insurance questions using the clinic's own information. Front-desk staff review and send the drafts. Clinical questions are never answered by the system; they are flagged and passed to a clinician. After the pilot, the clinic compares how long enquiries took to handle before and after, and checks how often the drafts needed changes.

• –Automating a messy process makes the mess faster. If the current process is unclear or changes weekly, fix it on paper first.
• –Keep a person in the loop for anything that goes to customers, involves money or changes records permanently, at least until the workflow has proven reliable.
• –Give the system access only to the data and tools that the task needs.
• –Plan for upkeep. Connected tools change, and prompts and rules need adjusting as the business changes.
• –Most of the effort is usually in connecting the tools you already use, not in the AI step itself, so check early that your systems can be connected.

Before the pilot starts, write down what you will measure and how the process performs today. Without that baseline it is hard to tell whether the workflow helped. How to measure AI automation ROI covers what to track.