Products are getting more complex, tariffs and supply chain disruptions make prices unpredictable, and new regulations tie up additional capacity. As a result, procurement teams mostly take care of the most important A items, while the long tail of B and C parts runs without active management.

In the webinar, Jakob Hafner (Product Manager at Tacto) and Daniel Owens (Engineering Lead at Tacto) show how AI agents in the new Sourcing Intelligence module manage price requests and framework agreements for the entire spend, and why procurement keeps full control throughout.

Capacity pressure: Why procurement teams actively manage only a fraction of their items

‍

Jakob Hafner describes the starting point with three examples. First, framework agreements: they are often renewed with the same supplier at the same or even a higher price without a new tender. Where a tender is run or the supplier is changed before renewal, savings of 13 percent can be achieved on average. Only about half of all framework agreements are renewed on time, 15 percent barely 30 days beforehand and 34 percent only after expiry or depletion. Purchase orders then go out against contracts that are no longer valid, order confirmations come back with deviations and the follow-up contract is negotiated under time pressure.

Second, capacity: With around 220 working days and one to two days of effort per item for tendering, renegotiation and negotiation preparation, there is time for about 110 items. Anyone wanting to cover the entire portfolio this way would need five times the available time. In practice, procurement organizations are actively working on about 20 percent of their item numbers, while the rest runs via email and Excel. According to the webinar, companies overpay by around ten percent on average on the part of the spend that is not actively managed.

Third, project business: According to a study from RWTH Aachen cited in the webinar, 85 percent of companies calculate quotes using rough estimates. 67 percent of the items fall within a corridor of plus or minus five percent, 12 percent deviate by five to ten percent and 21 percent by more than ten percent. The risk is negative margins on won projects or lost quotes due to buffers that are too high.

‍

Sourcing Intelligence: How AI agents manage price requests and framework agreements

‍

The foundation is Tacto's Procurement Brain. It combines internal ERP data (item master, order and invoice history, contracts, supplier master, info records, demand), supplier and process data, external market data with more than 20,000 price indices, the world knowledge of the models and the strategy of the procurement organization.

Sourcing Intelligence works as a multi-agent system with different triggers: a new demand without supplier and price, a requisition that has not seen a new price for a long time, a new project, an expiring framework agreement, an item without a valid info record or an AI alert from the procurement radar. Depending on the trigger and the stored strategy, the system chooses the suitable route: RFQ, negotiation preparation, search for new suppliers or item search in the catalog. For an RFQ, the agent creates and sends the request, conducts negotiation rounds or suggests messages, reads incoming quotes, compares them and creates an award recommendation. The buyer approves, after which the ERP is updated.

What the agent may do on its own is defined by procurement teams themselves: value limits, whether it may request quotes from alternative suppliers or negotiate itself, and how the award works, also per commodity group, for example with dual sourcing for plastic granulates.

‍

Live demo: Three cases from everyday procurement

‍

Case 1, new drawing part: A requisition for a stainless steel flange housing, 250 pieces for the start of series production, arrives from the ERP without supplier and price. The agent reads material, dimensions and tolerances from the drawing and the 3D model, models the manufacturing steps and calculates a target price of around 380 euros. It selects suitable suppliers, sends the request and negotiates further. The best bidder lowered its price after being confronted with the calculation and ends up at a twelve-month framework agreement at 389 euros, almost 24 euros below the first quote. Only now does the buyer come into play, sees the quotes side by side including delivery time against required date and awards with one click.

‍Case 2, expiring framework agreement: A quantity contract for a corrugated beam has reached 80 percent consumption, and the agent forecasts depletion for 12 November. Instead of taking the average of the last twelve months, it queries the forecast at the plants, also by email and in English, and reads free-text replies. Result: Over twelve months, a good 2.2 times the previous contract quantity is needed, among other things because of a new assembly line in China. The agent suggests around 70,000 pieces and recommends negotiating with the incumbent supplier first. Because it is an important supplier, it only drafts the email. The buyer adjusts it and sends it.

Case 3, price jump on a welded assembly: An AI alert from the procurement radar shows that the price has risen by 72 percent since the beginning of the year, while raw steel, electricity and labor rose by only 46 percent over the same period. Because the supplier is a strategic A supplier with high switching costs and the annual review is coming up, the agent chooses a different route than in the first case. It obtains comparison quotes from two qualified welding shops (around 590 and 620 euros) and deliberately does not ask the incumbent supplier. The results end up as a note in the negotiation preparation for the annual review.

‍

Conclusion

‍

Sourcing Intelligence is meant to extend procurement capacity from the actively managed fifth of items to the entire spend. The demo shows that agents can calculate target prices, run requests and negotiations and renew framework agreements on time, while buyers decide at the defined approval points.

Jakob Hafner and Daniel Owens (Tacto) show how AI agents in the Sourcing Intelligence module manage price requests, framework agreements and B and C parts: from target price calculation through RFQs and negotiations to the award recommendation. The webinar also covers three live cases from the demo, first results from practice and answers to questions on prerequisites and ERP integration.