A normal email rule in Gmail or Outlook matches words: a sender, a subject line, a keyword. Spark, Google's personal AI agent, still starts from a Gmail filter like that to choose which emails to look at. Then it reads each one and acts on what it says.
For example, a small distributor could give it one instruction: when a customer reports goods that arrived damaged, add the order number, the customer and the damage to a claims sheet, and create a Drive folder named after the order. A keyword rule built on "damaged" misses the customer who writes that two cartons were crushed, and catches the one who writes that nothing was damaged. Spark reads both emails the way the claims clerk would. At 20 reports a week and five minutes of copying for each, that is about a working day a month handed off.
Spark works on personal Google accounts, and on company accounts only for selected business users. It also needs Keep Activity switched on, a setting under which Google may train on your tasks and people may review them, so a team's customer emails do not belong in it yet.
Work like this stayed manual because no two customers describe a problem in the same words, and twenty cases a week never justified an IT project. Once a clear sentence is enough to set up a rule like this, the best person to write the rule is the clerk who does the job today.