What should you automate first?

List what takes the time

Ask each team for the three tasks they would happily give away. Note how often each one happens and roughly how long it takes. An hour of asking gets you a better list than a month of guessing.

Score each task out of ten

Give each task up to two points for each of these:

• Volume: it happens every day or every week.
• Clear rules: a new hire could learn it from a written page.
• Examples: you can show a hundred real cases.
• Reachable data: the systems involved can export or connect.
• Safe mistakes: an error can be caught before it reaches a customer.

Good first projects

These tend to score well in most companies:

• Entering orders from email into an ERP.
• Reading supplier invoices and matching them to purchase orders.
• Sorting a shared inbox and drafting replies.
• Moving data between two systems that do not talk to each other.

What to leave for later

Some things score badly as a first project, however tempting they look:

• Tasks that happen a few times a year.
• Work where the rules live in one person's head.
• Anything where one mistake would be costly and hard to spot.
• Processes that are about to change anyway.

Fix the process before you automate it

Automating a confusing process only makes the confusion faster. Take out steps nobody needs first. You may find that the problem shrinks before anyone builds anything.

Keep the first milestone small

Aim for something you can test with real work within a few weeks. Run it alongside the manual process at first, then compare the results. Our savings calculator gives a rough idea of the payback, and our cost guide explains what moves the price.