busywork
Home / Blog / Method
Method

Where to start: choosing the first process to automate

Five questions to pick the right process to start with AI automation, and the most common mistakes to avoid.

The first automation project often decides whether there will be a second. If it goes well, the team believes in it. If it drags on for months, AI becomes "that thing we tried". That's why choosing the first process matters more than the technology.

Five questions

  • Does it happen every day or every week? Frequent processes show results quickly.
  • Does it have rules someone can explain? If the person who does it can describe how they decide, an agent can learn it.
  • Can you measure it? Hours, errors, response times. Without a starting number you can't say whether it worked.
  • Can a mistake be corrected? Start with processes where the AI suggests and a person approves, before touching critical ones.
  • Does someone want it? The right process is the one that weighs on the people who do it every day.

Typical candidates

Bank reconciliation, posting supplier invoices, entering orders from email, supplier confirmations and follow-ups, answers to recurring customer questions. Almost every company has these processes, with clear rules and high volumes.

Mistakes to avoid

  • Starting with the hardest process to prove AI can do anything.
  • Only involving management: the people who know the exceptions are the ones doing the work.
  • Buying a tool before understanding the process.
  • Making everything automatic from day one, without approvals.

Our method

A free 45-minute call with the people who know the processes, one week in which the team records a screen video of how they work today, then a fixed-price pilot on a single process. It's the fastest way we know to find out whether automation works on your data before you spend more.

Want to know if this works for your company?

Free assessment

← All articles

More articles