How to choose your first AI project

Find a task that regularly holds up the team

Begin with work people keep raising: entering email attachments into a system, organising repair requests, chasing order updates or finding the same detail across several documents. Describe the task before discussing software.

“Improve operations” is too broad to scope. “Reduce the time needed to turn repair messages into complete work orders” gives the team something to observe. You can identify the information to read, the fields to enter and the person who receives the completed record.

Collect ordinary examples and awkward ones. Record how often the work occurs, how long it takes, which details are usually missing and how staff handle exceptions.

Check the software you already have

Ask the people doing the work to demonstrate the current process. Then inspect the existing system’s features. A form, template or reminder rule may remove repeated entry without a new AI application.

AI may be useful where the task involves interpreting different wording, extracting information from documents or preparing a draft. Fixed-date reminders and routing based on clear conditions are good places to check ordinary automation first.

Include checking in the comparison. Faster preparation is of little use if staff spend longer correcting the output. Time the full task through to a record or reply that someone can use.

Make the first scope easy to check

The first release might cover one document type, one enquiry channel or one handover. For example, organise a single type of purchasing document into a record for review before expanding to other formats.

A business owner needs to confirm fields, answer process questions and arrange staff testing. The technical team needs to establish system access and permissions.

Before work begins, answer these questions:

• Which task takes the time, and how many times does it occur each month?
• Where does the new workflow stop, and which cases go to a staff member?
• What are the project fee and the continuing operating costs?
• Which records will show whether the change is worth keeping?

Those answers support a choice between implementation, a smaller scope or improving the current process first.