Five agents read customer order emails for Unilever
Unilever built NOVA, five agents on Microsoft Foundry: four read order emails and extract the order, a fifth checks their work. Across 11 pilot markets the first order-to-cash step fell from 15-20 minutes to about 90 seconds, and orders are 95% touchless.
- Agent or tool
- NOVA (Microsoft Foundry)
- Topic
- Business & Startups
- Reported on
- Case study (Microsoft)
- Added
- 2026-10-29
Why it matters
Small teams wear many hats. Agents that pick up research, admin and operations let a founder spend more time on the work only they can do. It is one of 24 sightings filed under Business & Startups on Agent Sightings (number 1 by interest), and one of 1 involving NOVA (Microsoft Foundry).
How you could try it
Here is a reusable prompt, written by us in our own words, that captures the task. Paste it into NOVA (Microsoft Foundry) or any similar agent and adjust the details to your situation.
I get customer orders by email in messy formats. Read each email, pull out the customer, products, quantities and delivery date, then have a second pass check the extraction against the original. Anything missing or odd goes to a person with a note on what is unclear.
The prompt breaks down into 3 steps you can check as the agent works:
- I get customer orders by email in messy formats.
- Read each email, pull out the customer, products, quantities and delivery date, then have a second pass check the extraction against the original.
- Anything missing or odd goes to a person with a note on what is unclear.
Tip: Start small: run it once on a low-stakes example, read what the agent did, and only then let it run on real work.
Source
Read the original first-hand account on Case study (Microsoft): https://ukstories.microsoft.com/features/unilevers-multi-agent-creating-custome…. Our summary is a paraphrase; details are in the source.
Featured in these guides
Built an AI agent? Feature your agent on Agent Sightings, free submission or a featured spot.