People in purchasing spend much of their day chasing. Order confirmations that don't arrive, delivery dates that change, suppliers who answer half the question. The real work, choosing suppliers and negotiating, gets squeezed by emails and spreadsheets.
Order confirmations
Every order sent should come back with a confirmation. Someone has to open it, compare it with the order and update the system with the dates. An AI agent reads the purchasing inbox, recognises confirmations, compares them line by line with the order and only flags what doesn't match: a different price, a reduced quantity, a moved date.
Follow-ups
Every morning the agent checks open orders. If a confirmation is overdue or a delivery is late, it drafts the reminder to the supplier, in their language and with the right references. The buyer reads it and sends it, or, for suppliers they've already cleared, it goes out on its own.
The morning overview
Instead of opening ten email threads, the buyer gets a summary: confirmed orders, orders with differences, late deliveries, reminders sent, replies received. In a few minutes they know where to step in.
What it doesn't do
- It doesn't choose suppliers or negotiate prices.
- It doesn't accept price or quantity changes without a person.
- It doesn't send anything to a new supplier without approval.
What you need to start
Access to the purchasing inbox, read access to orders in your system (with or without an API) and a few examples of how the team handles exceptions today. The first pilot usually covers confirmations and follow-ups, because that's where most time comes back in the first weeks.