Email Automation: From Chaos to Control
The inbox takes more time than it should. Here's how an AI agent can sort, reply and move information into the right systems, with people still involved where they're needed.
The inbox became a workplace
In many smaller businesses, the inbox has become the place where most work starts. Customer enquiries and supplier invoices arrive there, and colleagues ask about things that really should live in a system. In between sit newsletters and other things that might be important.
It takes time. A 2012 McKinsey Global Institute report, The social economy, estimated that knowledge workers spend about 28 percent of their workweek on email. Based on a 40-hour week, that's a little over eleven hours. The report is old and was written before Slack and Teams became common, so the figure should be read with that in mind.
The inbox has simply been given work it was never designed for. That can be changed without switching email client.
Why the inbox gets messy
Anyone can send anything at any time. That makes email good for communication but poor as a work tool, because very different kinds of messages end up in the same place:
- Information that needs nothing, like receipts and newsletters
- Tasks that should go into a system, like orders and invoices
- Communication that needs judgment, like customer questions and complaints
Because everything arrives in the same inbox, someone has to open each email to work out what it is and what should happen to it. That's where the time goes.
What the inbox costs
The time around each email
The reply itself is often quick. What takes time is finding earlier emails and checking in another system whether something has already been done.
Interruptions
An email read in the middle of other work costs more than the time it takes to read it. Gloria Mark, Victor González and Justin Harris at the University of California, Irvine, followed knowledge workers in a 2005 study. When work was interrupted, it took on average just over 23 minutes before the person was back on the interrupted task, and in the meantime they usually worked on a couple of other things. Someone interrupted by the inbox again and again through the day struggles to settle into their work.
Lost threads
When important information only exists in email, it's easy for something to go astray. A customer sends a specification to the salesperson but not to whoever will do the work, or a change to a contract ends up in a thread nobody finds again.
Unclear status
In an inbox, it's rarely clear what's done. A read email might be finished or completely forgotten. If it's unclear whether anyone has replied to the customer, the result is double-checking and sometimes duplicate work.
What an AI agent can do with the inbox
An AI agent can now handle a large part of the work with incoming email. It reads each message and understands what it's about, even when the sender phrases things in a way nobody anticipated.
A supplier invoice is read, and the amount and reference are passed to the accounting system. A quote request is added to the CRM with the customer's requirements filled in.
The agent can answer routine questions itself, based on your own documents and the information in your systems. If a customer asks about delivery time, the agent can look up the order and reply with the right date.
The agent also keeps track of what hasn't been answered. A quote that has gone unanswered for ten days gets a friendly reminder, and a customer question nobody has picked up within a day is passed on to the person responsible.
Anything sensitive or unusual is handed over to the right person, ideally with a draft reply. You decide how much it may do on its own based on the risk. The agent only gets the permissions it needs, and everything it does is logged so it can be reviewed afterwards. Replies that commit the company can require a person's approval first.
An illustrative calculation
The numbers below are made up to show how the calculation works. They don't describe a real company.
Imagine a services company where five people handle customer contact and admin by email. Each spends an average of 8 hours a week reading, sorting, replying and following up.
- 5 people × 8 hours = 40 hours a week
- Over a year with 44 working weeks: 1,760 hours
- At an hourly cost of €45: about €79,200 a year
We assume a third of that time disappears with an agent. That's about 587 hours a year, or a little over €26,400. The agent also costs money to run. If we assume €150 a month, or €1,800 a year, the net saving is about €24,600 a year.
How much you can actually save depends on what your flows look like, and the only way to know is to measure.
Where do you start?
Measure for a week
Ask the team to note roughly what they do in the inbox over five working days. Simple categories are enough, such as whether the email was forwarded or got a standard reply. After a week you can see which kinds of email keep coming back.
Start with one flow
Don't try to automate the whole inbox at once. Pick the most common kind of email, such as quote requests, and build a flow for just those. Once it works, it's easier to add the next one.
What's left for people?
The things that need a person, like conversations with unhappy customers and quotes that need judgment. And some of the work that doesn't get done today because the inbox takes all the time.
Want to know how much time your email takes and where it's worth starting? Get in touch and we'll go through your flows together.