AI for procurement and supplier management

The purchasing department deals with two very different categories of documents: poorly written estimates to be compared quickly, and agreed conditions that then no one checks anymore. They are two problems with two distinct automation paths, and one of the most common mistakes is building a single project thinking that it solves both in the same way.

The first is a format problem: each supplier structures its offer in its own way, with differently aggregated entries and discounts expressed in three ways. The second is a problem of time: the conditions agreed at the beginning of the year are lost in an inbox and no one compares them anymore with what really comes on the invoice.

There is then a third theme that the purchasing office almost always handles by hand, when it manages it: knowing if a supplier, especially if he works in contract or subcontract, is in compliance with his own contribution obligations. It is a check that requires you to remember it supplier by supplier, which is why in practice you only do it at the beginning of the relationship and then you forget.

Keep the price list up to date before the comparison

The basis of comparisons may change when the supplier sends a new catalog. For update supplier price lists with AI need version, codes, currency and quantity per pack. The guide deals with previews of partial variations and imports: an article absent from a file must not automatically become an article to be deleted.

The comparison of estimates: normalize before deciding

1. Offers come via email in different formats, often as unstructured PDF attachments.
2. The system extracts items, quantities, unit prices, discounts, ancillary conditions.
3. Normalize the entries: acknowledges that "transport included" and "freak fate" describe the same condition in different words.
4. It produces a comparative table on a homogeneous basis, with reference to the exact point of the original document from which each data comes.
5. It highlights relevant differences and abnormal conditions, without deciding them for you.

The normalization step is what it really is: comparing two prices is trivial, understanding that one includes a service that the other bill aside is the work that gets hurt when there is a rush, and that's where you lose margins without noticing it.

Respect (and enforce) payment terms

The buyer is also the one who pays, and the payment times to suppliers are not left to custom: the legal term, in the absence of a different written agreement, is thirty days from receipt of the invoice, as established by the Article 4 of Legislative Decree 231/2002. The parties may agree on longer terms, but beyond sixty days the clause must be expressly agreed and proven in writing, precisely to prevent a too long term from becoming an indirect way to discharge the cost of the credit on the supplier. Once the deadline has passed, interest on late payment automatically runs from the next day, without needing a formal reminder from those who must cash in.

A system that monitors these deadlines not only serves to avoid an additional cost: it also helps to maintain a relationship where the supplier does not have to chase the company to get paid on the agreed schedule, which weighs when negotiating the next supply.

It is worth noting that the rule was born to protect those who sell from systematic delays of those who buy, not the other way around. Knowing it on the side of those who buy still serves, because a company that always pays on time, and can prove it with data and not with the memory of those who manage the relationship, negotiates better conditions on the next supply. It is a concrete commercial advantage that arises from a fulfillment designed for quite another reason.

The DURC of suppliers: a check that is rarely done

Anyone working with contracts, subcontractors or suppliers that employ staff should check that their social-security contributions are in order. Almost nobody does this systematically, because it means remembering to check each supplier individually. The Documento Unico di Regolarità Contributiva (DURC) confirms a business’s compliance with contribution obligations towards INPS, INAIL and, for construction companies, the Casse Edili construction-sector funds. It can be checked in real time through the online service provided by INPS and INAIL, as described by INAIL on its dedicated page. The generated document has a limited validity in time, after which it must be requested again, and those who have an interest in verifying it, not only the provider itself, can do so freely.

This check lends itself to automation because it does not require interpretation: it is a query with a binary outcome, compliant or requiring regularisation, which someone currently has to remember to repeat periodically for every relevant supplier. A system that maintains a schedule of expiry dates for each supplier and flags documents nearing expiry turns an intermittent check, often performed only when someone remembers after a problem, into a routine verification that does not depend on one person’s memory.

• Identify the suppliers for which DURC is relevant: procurement, subcontracting, companies with employees
• Schedule a periodic verification instead of checking only when the report starts
• Report the expiry of the document in advance, not when it has already passed
• Keep track of who received the invitation to regularize and at what time

The control of conditions: the use that almost no one does

You agree on a price, a discount, a minimum order, and then the invoices arrive without anyone systematically verifying that the conditions are met. Not out of negligence: because it would require comparing each line with an agreement that stands in an email months earlier.

The natural connection is with the passive cycle described in AI and Electronic Billing: the same invoices that go into accounting are the source to check if a supplier is complying with the agreement, and building control twice, once for purchases and once for administration, is a waste that happens more often than it seems when the two offices do not speak to each other during the design.

Construction and manufacturing: same tool, opposite priorities

In one example: a construction company that collects multiple quotes for each supply finds the most immediate benefit in comparison, because the documents arrive in always different formats and the risk of choosing looking at only the total is concrete. Here the verification of the DURC of subcontractors is combined with the comparison of estimates as a second control to be automated.

A manufacturing company with framework contracts, prices set at the beginning of the year on hundreds of lines, finds the value above all in the control of conditions: the differences that emerged in the first quarter of monitoring are often worth more than the time saved in the initial comparison, because they concern amounts that are repeated order after order.

A small business that always buys from the same two or three suppliers at list prices, without negotiating specific terms, is the case in which neither project makes sense: there is no normalisation to do because the documents are always the same, and there are no negotiated terms to monitor over time. It is worth saying this clearly rather than building something anyway that nobody will use: the variety of suppliers and the presence of negotiated terms matter more than the company’s turnover when deciding whether the project makes sense.

Reconciling what has been ordered and received

The purchase proposal and the document received by the supplier do not close operational control. Delivered quantities, units of measurement and prices may require separate verifications. The guide to DDT control, orders and invoices with AI shows how to link the evidence and prepare an exception queue without treating textual similarity as approval.

Compare quotes on the same basis

Two offers may have seemingly different prices because they include different quantities, services or conditions. The comparison of supplier quotes with AI enters the normalization of lines, units and ancillary costs. A comparable matrix helps the buyer to formulate targeted questions; if data is missing, the choice remains suspended instead of producing an artificial ranking.

Prepare the request before the purchase decision

The comparison of offers becomes useful when it arrives at the manager with verifiable amount, motivation and attachments.approval of purchases with AI distinguishes preparation, company thresholds and confirmation of the person. It also faces absence of the approver and variations after the decision, so a complete practice is not mistaken for an already authorized order.

What's not up to the system

AI does not choose the supplier, and this is not merely formal caution: the choice depends on factors that do not appear in the documents. How reliable the supplier is with deliveries, how it behaves when there is a problem, whether it is better to keep two suppliers rather than one to avoid depending on a single channel. The system makes the numbers clear; the buyer adds what they know and makes the decision. A project that tries to automate the final decision as well as the comparison is usually bypassed by the people meant to use it, because they do not trust a recommendation that cannot see everything they can.

Before you leave it is worth looking at a month of real purchases and counting two things: how many times have you compared more offers for the same need, and how many invoices have you really verified compared to the agreed conditions. The method of estimating what is best to automate first is in how much it costs to automate a process, while the choice between doing it with your team or relying on someone is discussed in DIY AI or agency. The connection with the warehouse, for those who also manage the stocks of the purchased materials, is in AI for stock and inventory management, and the general framework of the administration in AI for Accounting.

If you work in production and want to understand which control to start from between suppliers, conditions and deadlines, Giallo Studio follows manufacturing companies in setting up the first project on the documents that generate the most work today.

A final practical criterion for choosing where to start: look at which of the three problems, the comparison of quotes, the check of conditions or the verification of suppliers, has already caused you a concrete problem in recent months, not which one seems technically more interesting to build. The project that solves an already felt pain is adopted immediately by those who use it; the one chosen only because it seemed the easiest to program often remains on but ignored, and after a few months no one remembers why it had been built like this.