Start with one week’s time log

If entering one order takes six minutes, 40 orders a week add up to four hours. Add up the time for other recurring tasks in the same way.

A plumbing-supplies wholesaler with eight employees might have the following weekly time log:

  • Answering customer questions about delivery and product availability: 60 × 4 min = 4 h.
  • Entering orders received by e-mail: 40 × 6 min = 4 h.
  • Preparing quotes: 8 × 30 min = 4 h.
  • Entering supplier invoices: 45 × 4 min = 3 h.
  • The weekly report for the manager: 1 × 2 h = 2 h.
  • Reviewing complaints: 3 × 20 min = 1 h.

A total of 18 hours a week. With a list like this, you can start deciding which tasks to automate and which to keep doing yourselves.

Which task to automate first

This business could start with entering orders. Each time, an employee finds the product name, quantity and delivery address in an e-mail and copies them into the stock system. AI can read those details and prepare them for entry. Invoices can be processed in a similar way, including scanned documents and photographed delivery notes.

Answering customer questions needs some preparation. You need to collect current delivery information and provide access to stock levels. Otherwise, AI cannot say whether a product is in stock or when it can be delivered.

For quotes, a price list alone may not be enough. If only one employee knows each customer’s discounts, write them down with that person first. If the manager asks for something different in the report every week, agree on its contents. AI can then prepare quotes and reports following those rules.

For complaints, AI can summarise what the customer wrote and find the order. The employee still decides whether the complaint is justified and what to offer the customer.

A person has the final say

At first, an employee checks every order processed by AI: do the products, quantities and delivery address match? This helps reveal recurring mistakes and what needs adjusting in the solution.

Before starting, agree on when AI should ask an employee for help. If the customer has left out the quantity, the employee asks them to clarify it. If the invoice total does not match the order, the employee finds out why. Initially, an employee also reviews replies to customers and sends them only when everything is correct.

If AI answers a customer in your website chat, tell the customer. The EU’s Artificial Intelligence Act has required this since 2 August 2026. The European Commission’s explanation

Try it with real examples

Once you have chosen a task, collect 20 examples from everyday work. For orders, these would be customer e-mails — both straightforward orders and ones with missing or unclear details.

During the trial, check whether AI reads the data correctly and spots missing information. Compare how long orders took before with how long they take now. Include checks and corrections to see how much time you actually save.

If you would like to talk your list through with us, book a free call. In half an hour we will choose together where to start.