Skip to content
Exclusive partner BME Symposium18–19 November · Berlin · Booth A16Book a meeting
Contact

When analysis is cheap, judgment is scarce.

Cost drivers, indices and alternatives can now be prepared in minutes. What decides a negotiation is knowing which lever matters. A person asks that question.

In short

Negotiation preparation in procurement builds the decision base before the meeting: cost drivers and indices, contract deadlines, your best alternative (BATNA), target range and possible concessions. AI makes this analysis almost free, so every relevant negotiation can be prepared, not only the biggest. What becomes scarce is judgment about which lever matters. That is why preparation needs method, and the negotiation itself stays with people.

Data: third-party research with source. Our view: the joint position of H&Z and Elevatry, indicative observations from project work. Model: calculations and examples, not customer results.

Analysis has become cheap. Judgment has not.

Market scans, bid comparisons and cost breakdowns now cost almost nothing. In a field experiment with 758 BCG consultants, participants using AI on tasks inside the models' capability frontier completed more work, faster and at clearly higher quality. For procurement, what used to take a week of preparation is ready within an hour.

That does not make preparation easier. When both sides can produce a cost breakdown in minutes, preparation no longer differs in the amount of analysis. It differs in the question put to it: is the lever the unit price, the index clause, volume bundled across four plants, or the notice window?

Experienced buyers usually answer that question from memory. It is the scarce part of the work, and it becomes more valuable the cheaper everything else gets. An AI that weights every lever equally produces a complete list and no answer. Pull the wrong lever and you lose the negotiation, however well prepared.

Figure 1DataInside the capability frontier, AI makes analytical work broader, faster and better.
+12.2%More tasks completed
+25.1%Completed faster
>40%Higher quality

Dell'Acqua et al., Harvard Business School working paper 2023. Field experiment with 758 BCG consultants, tasks inside the AI frontier, compared with participants without AI.

The most dangerous answer is fluent, confident and wrong.

The same study shows the other side. On a task just outside the frontier, consultants using AI were 19 percentage points less likely to be correct than colleagues without it. Quality does not decline smoothly with difficulty. It fails where a task only looks familiar.

In negotiation preparation, that is where the conclusions sit. An index analysis can be flawless while the read on leverage is wrong, for example when a re-tender is recommended although requalifying a drawing part takes nine months and the contract ends in five.

Without method, AI scales the average. With H&Z's method from over 5,000 projects, Companions work through levers, assessment criteria and cost logic step by step and show their reasoning. The buyer can see exactly where to push back. Visible reasoning is what makes judgment possible at scale.

−19 ppcorrect answers when consultants used AI on a task just outside its capability frontier.DataDell'Acqua et al., Harvard Business School working paper 2023
An index analysis can be flawless while the conclusion about leverage is wrong.

The negotiation stays human. That is a design choice.

A commitment to a supplier is hard to take back, and it often binds the company for years. Accountability, the relationship and the trust of works councils sit there. So the Negotiation Coach prepares: objectives, BATNA, concession ladder, arguments and award scenarios. Red lines, leading the meeting and the award stay with the team. With Training & Simulation, the team rehearses the negotiation beforehand, against the other side's arguments.

What changes is coverage. In our observation, only about one supplier in five is actively managed today. The remaining price talks are handled from memory, and that is where increases go through without a counter-calculation. When preparation costs almost nothing, even a tail supplier's price increase meets a documented counter-calculation.

That is where the real value lies. An increase that passes without a counter-calculation becomes part of the price base for every year that follows. Whoever knows which share the raw-material, energy and wage indices can justify negotiates the difference rather than the demand. That difference is the room to negotiate, and it is usually larger than the conversation suggests.

Figure 2ModelWhat the supplier asks for and what the indices justify: the difference is the room to negotiate.
7.0%
2.8%
4.2 pp
Requested by the supplierJustified by material, labour and energy indicesRoom to negotiate

Model. Values from the worked example below (turned parts, €4.8M a year); index movements illustrative, not market data.

Companions for this topic

What Elevatry does for you.

Companions work around the clock in your data, with the H&Z method from more than 5,000 procurement projects. They prepare decisions, work in your experts' feedback and, once you approve, carry out the next steps. Your team decides.

All Companions
Model

A price increase nobody would have prepared for

A supplier of turned parts announces 7% more from January, citing steel and wages. Spend is €4.8M a year across three plants. Nobody has ever negotiated with this supplier in a structured way.

  1. 01
    Decompose the cost

    Material is 45% of cost and the steel index rose 4%. Wages are 30%, with a 5% pay deal. Energy is 10%, down 5%. The indices justify 2.8%, not 7%.

  2. 02
    Read the contract

    The framework agreement has no price adjustment clause and ends in eight months, with three months' notice. The negotiation window is open and closes in five months.

  3. 03
    Test the alternative

    Two qualified suppliers are within delivery range. One plant already buys similar parts elsewhere. Requalification takes four months and fits the window. The alternative is credible.

  4. 04
    Recommend

    Reject 7%, counter at 2.8% and prepare a bundled RFQ across three plants as the alternative. The negotiation is about 4.2 percentage points.

  5. 05
    The person decides

    The category manager weighs the relationship, timing and red lines and decides whether to trigger the RFQ. She leads the meeting.

At stake are 4.2 percentage points on €4.8M, about €0.2M a year, with a supplier whose increase would otherwise have gone through unchallenged.

Illustrative, not a customer result. Cost structure and index movements are assumed. Price effect per driver = cost share × index movement.

Questions worth asking.

  1. 01

    How many price increases did we accept last year without a documented counter-calculation, and what did they add up to?

  2. 02

    What share of our negotiations is conducted with a documented alternative, and who checks whether it is credible?

  3. 03

    How do our buyers recognise that an AI analysis lies outside its capability frontier?

  4. 04

    Where is it written down which negotiation decisions are never taken without people?

Free download

Practical guides on this topic

The Elevatry guides take the topic into day-to-day work: as a PDF, free of charge.

What procurement leaders ask us.

Can AI negotiate with suppliers on its own?
No, and that is a design decision, not a technical limit. A commitment to a supplier is hard to take back and often binds the company for years. What wins a negotiation is created beforehand anyway: the right lever, a credible alternative, a concession ladder with a price on each step. The Negotiation Coach prepares this; the team leads the conversation. What is new is that this preparation is possible for every price round, not only the large ones.
How do I check whether a supplier's price increase is justified?
Break the demand down into cost drivers and check each against its index. In our worked example, a turned-parts supplier asks for 7%, citing steel and wages. Material is 45% of cost and the steel index rose 4%; wages are 30% with a 5% pay deal; energy is 10% and fell 5%. That justifies 2.8%. The negotiation is then about 4.2 percentage points, not the demand. For smaller suppliers there used to be no time; today this counter-calculation takes minutes.
How much time does AI save in negotiation preparation?
A week of preparation becomes about an hour, but the time saved is not the point. In a field experiment with 758 BCG consultants, participants using AI on suitable tasks worked 25.1% faster and at over 40% higher quality. When analysis costs almost nothing, judgment becomes scarce: which lever matters. The real gain is coverage. Even a tail supplier's price increase then meets a documented counter-calculation instead of becoming part of the price base for years.
How can I tell when an AI negotiation analysis is wrong?
Not from its language: the wrong answer sounds as confident as the right one. In a Harvard Business School field experiment, consultants using AI just outside its capability frontier were 19 percentage points less often correct. The error sits in the conclusion, not the arithmetic: a recommended re-tender although requalifying a drawing part takes nine months and the contract ends in five. Check switching costs, deadlines and leverage. Companions show their reasoning so that challenge lands where it matters.
Is structured negotiation preparation worth it for smaller suppliers?
Yes, because that is where price increases go through unchallenged today. In H&Z and Elevatry project work, only about one supplier in five is actively managed; the other price talks are run from memory. In our worked example, 4.2 percentage points on €4.8M are on the table with a supplier never negotiated with in a structured way, about €0.2M a year. When preparation costs almost nothing, no supplier is too small for an evidenced position.

Friday notes on the topic

Next step

See it on your own procurement.

We show you on one of your processes how Companions do the work from analysis to contract. With your data and your approval rules, if you like.

See the platform