What is business process automation, and what has AI added to it? 

Business process automation is the use of software to run repeatable business work from start to finish. IBM describes it as a strategy that uses software to automate complex and repetitive business processes. AI now extends it to steps that once needed a person to read, listen or decide.

What does it look like in practice?

Take a new customer enquiry. Without automation, someone reads the email, types the details into a system, tells a colleague and tries to remember to follow up. With automation, the enquiry is recorded, assigned and acknowledged at once, and a reminder appears if nobody has responded. It's the same work, except it no longer depends on somebody remembering on a busy Tuesday.

What has AI made possible?

Traditional automation breaks when the input is unpredictable. For years, this set a hard limit. A process could be automated up to the point where someone had to read a letter, understand a caller or judge a free-text request. Then it stopped and waited for a person.

AI removes many of those stopping points. We have built AI front desk systems that take phone conversations and turn them into bookings and records. We have also built AI companions for elderly people, which hold ongoing conversations and can alert family when needed. Neither fits the old model of forms and fixed rules.

Long, multi-step processes gain the most. Once the difficult step is handled, the whole chain can run: a voice message becomes a job, the job becomes a quote, and the quote gets followed up.

Rules or AI: which should do each step?

Use the simplest tool that works, because putting AI on a job that a three-line rule can do is like hiring a chef to boil water. Rules give the same answer every time, so they should do anything exact, such as totals, tax and permissions. AI is flexible, but it can be wrong, so give it the steps that need understanding and check its output. Most good systems use both.

How do you choose what to automate?

1. List the processes that use the most staff time or cause the most complaints.
2. Follow one process from start to finish with the people who do it.
3. Remove the steps that have no purpose. Automating a poor process only makes it fail faster.
4. Mark each remaining step as rules, AI or a person.
5. Build the smallest version that works, and measure it against the old way.

What usually goes wrong?

• Nobody owns the process after the system goes live.
• Exceptions were not considered, so staff work around the system.
• AI was given authority over actions that needed a person's approval.
• Results were never measured, so nobody can say whether it helped.

None of these are technology problems, which is the annoying part. Tell us what is giving you headaches. If we are not the right people, we will try our best to find who is, and connect you to them.

Frequently asked questions

These are the questions business owners ask most often about automating their processes.