Automation · Procurement Compare quotes and check confirmations with AI: uses, limits and how to start

All tasks of the department and the connection to your programs are on the page AI in procurement.

Where procurement loses time

Most purchasing departments struggle less with big decisions than with many small checks. One quote arrives as a PDF, the next as an Excel list, the third in the body of an email. Order confirmations have to be compared line by line with the purchase order. Delivery dates shift and have to be chased. Framework agreements have notice periods and price adjustment clauses that someone has to keep an eye on.

This is exactly the kind of work AI is well suited to: reading documents, extracting details, comparing them with existing data and reporting deviations. According to Eurostat, analysing written language is the most common use of AI in European enterprises. Unlike rigid templates that have to be set up again for every supplier form, a language model copes with changing layouts and also understands details that only appear in the running text of an email.

Six uses in procurement

1. Make quotes comparable: the AI reads quotes in any format and transfers prices, price breaks, delivery times, payment terms and delivery terms into a single comparison table.
2. Check order confirmations: every confirmation is matched against the purchase order. Deviations in price, quantity, item or date are reported with their location in the document.
3. Track delivery dates: open orders are monitored. If a confirmation is missing or a date is approaching, the AI drafts a query to the supplier.
4. Contracts and deadlines: terms, notice periods, minimum quantities and price clauses are extracted from framework agreements and set up as reminders.
5. Prepare requests for quotation: a requirement turns into a request to several suppliers, with the right technical details and your standard text.
6. Keep certificates in view: certificates, test reports and supplier declarations are recorded with their expiry date. The AI reports what will soon be missing.

A typical process by example

A machinery manufacturer orders from around a hundred suppliers. An order confirmation arrives as a PDF by email. The AI recognises the order number, fetches the purchase order from the ERP system and compares every line. For two lines the price is higher than ordered, for one the delivery date is two weeks later than requested.

The responsible buyer sees a short summary: three deviations, each with the line in the document and the value from the purchase order. He decides whether to accept the price or renegotiate. Confirmations without deviations are automatically marked as checked. This example describes a typical process, not a client project.

Where the limits are

• Negotiating and deciding: the AI provides comparisons and pointers, the choice of supplier and the negotiation stay with purchasing.
• Complex tables: nested price breaks or surcharges in footnotes are not always read correctly. That is why every detail comes with its location in the document.
• Ordering on its own: technically possible, but only sensible within tight limits on amount and supplier. OWASP counts excessive permissions among the ten biggest risks of AI applications. To start with, we recommend proposals with approval.
• Data quality: if item numbers, supplier records or prices in the ERP system are wrong, the AI compares against wrong values too.
• Tone with suppliers: queries and reminders drafted by an AI should suit your company and be read before sending. Good supplier relationships do not come from automatic reminders.

What you need to start

• Read access to the ERP system, meaning purchase orders, items and suppliers.
• Access to the inbox or folder where quotes and confirmations arrive.
• Clear tolerances, for example from which price deviation a confirmation goes for review.
• A person who handles deviations and gives feedback when the AI gets something wrong.
• Data protection and security: data processing agreement, servers in the EU where possible, only the permissions needed.

What it brings, honestly considered

The gain rarely lies in a single big saving. It comes from every confirmation being checked, even on busy days, from deviations being spotted early and from delivery dates no longer being forgotten. Your purchasing team has more time for suppliers, negotiations and alternatives. How we implement such processes is described on the page about process automation.