A supplier announces a price increase, citing higher raw material costs. The negotiation is in three days. What follows is familiar to many buyers:

Pulling ERP extracts, merging Excel lists, researching indices, comparing price trends.

The result is days of preparation and often still only part of the facts on the table.

AI can fundamentally change negotiation preparation in procurement. When purchasing data is well structured, a robust market assessment can be produced in seconds instead of days. This article looks at what that means in practice, where the limits are, and why the data foundation is decisive. It is also the topic we will be focusing on at the 61st BME Symposium in Berlin.

Why negotiation preparation takes so long today

Anyone who wants to negotiate well needs answers to a few tough questions:

  • How have our purchase prices for this material or category developed?
  • How has the market developed over the same period? Is the supplier's demand justified at all?
  • Where have raw material prices fallen without our purchase prices following suit?
  • How much has our volume with this supplier grown, and have we received anything in return?

The data for this exists in almost every company. But it is scattered across the ERP system, Excel files and external sources. Bringing it together takes time which is why thorough preparation is often reserved for the most important A suppliers. With all the others, savings potential is left untapped.

On top of that, there is an imbalance: the supplier knows its cost structure in detail. Procurement, by contrast, often argues on gut feeling rather than solid figures.

What AI can do in negotiation preparation

Generative AI helps with writing and research. In a negotiation, however, numbers are what count: price trends, volumes, market indices. This is where mathematical and statistical AI models come in, recognising patterns in large volumes of data. Four applications show what this looks like in practice.

Comparing your own prices with the market:
Public price indices, for example from Eurostat or the German Federal Statistical Office, reflect market developments. The challenge lies in the mapping: which index fits which category? AI can suggest this mapping. That makes it possible to compare all categories with the market at once, rather than checking individual cases.

Tracing raw material shares:
If the raw material share of a purchased part is known, for instance through a shadow calculation, the price trend of the part can be compared with that of its raw materials. If raw material prices fall while the purchase price stays the same, that is a clear reason to talk.

Assessing volume:
In practice, rising order volumes often do not lead to lower prices. An analysis across all suppliers shows where volume growth has so far gone unrewarded, a starting point for bonus agreements.

Checking tiered prices:
Not every price tier follows a logical structure of fixed and variable costs. AI models can uncover implausible tiers and point to alternatives. Taking storage costs and tied-up capital into account, savings of 5 to 10 percent are realistic.

What matters most is not the AI model but the data foundation. Only when orders, invoices, master data and contracts are structured and available at document level are the results traceable and reliable. This connection between ERP data and external market information is exactly where we start with WebCIS AI.

What changes for procurement and procurement controlling

When data collection no longer takes days, thorough preparation becomes possible not just for A suppliers but for a much larger part of the portfolio. This also narrows the information gap at the negotiating table: suppliers know their cost structure precisely, but with market indices and its own price history, procurement can now assess price demands objectively. Conversely, it can also see when price reductions would be appropriate.

For procurement controlling, there is another point. What ultimately counts here is whether negotiated savings actually reach the P&L. A consistent data foundation makes this tracking possible and makes procurement's contribution demonstrable to senior management.

The limits: AI prepares, people negotiate

AI results are not flawless. An unsuitable index mapping or poorly maintained categories can quickly lead to the wrong conclusions. Plausibility checks and expert knowledge therefore remain an integral part of the process.

The work at the negotiating table also remains human: supplier relationships, conducting the conversation and a sense for the right moment cannot be automated. The real gain lies elsewhere: AI delivers, in a short time, a factual basis that used to require considerable effort. What to make of it is still procurement's decision.

WebCIS AI at the 61st BME Symposium in Berlin

From 18 to 19 November 2026, procurement and logistics leaders will meet at STATION Berlin for the 61st BME Symposium. We will be there and look forward to discussing how AI is changing negotiation preparation in procurement.

Visit us at our stand B7 and bring your questions: Together, we can look at what a data-driven market assessment for it might look like.

Would you like to book a meeting in advance? Request an appointment or find out today where your procurement controlling stands with our free maturity check.

We look forward to seeing you in Berlin!