Notizen

The Problem with “The Other Side”

In complex negotiations, we tend to treat organizations as coherent entities. The customer. The competitor. The government. The investor. The regulator. The consortium. On one side, us. On the other, them. It makes the world manageable, but usually not for very long.

Behind “the government” are ministries, agencies, political levels and individual decision-makers. Behind “the company” are functions, business units and people with different objectives, risks and mandates. Internally, things rarely look much different. The closer we look, the more layers appear.

The problem is not that these people behave irrationally. Quite the opposite. Each person can act perfectly rationally from their own perspective, while collectively creating a negotiation that no longer behaves like a conversation between two coherent parties.

The convenient fiction of “the other side”

“What does the other side actually want?” sounds like a straightforward question. Many negotiation methods and training cases necessarily make it one. On one side there is us, on the other side them. Both have interests, positions, alternatives and a mandate, often already defined as part of the case. That makes perfect sense for teaching. Otherwise, a two-hour exercise would quickly turn into a two-week project.

Reality is less accommodating. Parties are not always as internally consistent as we assume, and in that respect they are often not very different from ourselves. What Engineering considers an acceptable compromise may create a problem for Finance. What one person has already promised may still have to be fought through internally by someone else. A demand that appeared to have disappeared yesterday may return because another actor has entered the discussion whose interests had not previously been considered.

From the outside, this can look contradictory or unpredictable. It may, of course, be deliberate strategy. But there is also a good chance that everyone involved is acting rationally — just not according to the same rationale.

These differences are not only a problem. They can also create room for negotiation. Different interests, priorities and mandates create fault lines, but also potential lines of compromise. Something that appears impossible to resolve with “the other side” may look different once we understand which actors need, tolerate or can prevent which outcomes.

Negotiation architecture cannot eliminate this complexity. It can make it visible and workable. Instead of reducing “the other side” to one set of interests and positions, the team can make different perspectives explicit and develop a richer understanding of who is actually involved, how their interests differ and where dependencies exist.

What changes with AI

This becomes particularly interesting when AI enters the picture. AI is remarkably good at processing complexity. Humans are remarkably good at reducing it early. Unfortunately, AI tends to believe us.

We write in a prompt, “The other side is demanding …”, and AI produces an impressively articulated proposal for how we should respond and what we might offer next. AI has answered our question, and because the answer sounds convincing, we may even feel more confident about our original interpretation. But the question already contained two assumptions: there is one party, and it has one position.

The problem is therefore not necessarily the quality of AI reasoning. We may have simplified the object it is reasoning about before the reasoning even begins.

AI becomes more interesting when it can work with a richer representation of the situation: different actors, interests, mandates, relationships, statements and assumptions. It can then examine whether our observations and assumptions fit together. Why does Finance support a proposal if it creates a risk that Engineering rejects elsewhere? Who would have to change their position for a particular compromise to become possible? Which actors place different emphasis on elements of the position being presented? Where does our strategy depend on a single unverified assumption?

Humans and organizations have always had to reduce complexity in order to remain capable of acting. Abstraction is a prerequisite for decision-making, but abstraction also discards information. Until recently, retaining substantially more detail came at a cost: somebody had to collect it, structure it, keep it current and reason across it.

AI changes that trade-off. Its advantage may therefore not initially be that it gives us a better answer to the same question. It allows us to keep a more complex question open for longer.

We can retain more actors, more perspectives and more contradictory information explicitly while still reasoning about them systematically. Simplification will still be necessary eventually. The important difference is that it does not have to happen quite so early.

And that may be one of the more consequential changes AI brings to complex negotiations: not better answers after we have simplified the problem, but greater freedom to decide when and how much we simplify it in the first place.

Team Size Is Not the Issue. Team Design Is.

One of the easiest ways to weaken a negotiation is to attend it with a whole soccer team.

Everybody has a legitimate reason to be there. Sales understands the customer. Engineering understands the technology. Legal understands the contract. Finance understands the economics. Management owns the decision. More expertise should lead to better decisions.

Yet negotiations often become less effective as more people join the table. A legal clarification may be correct but derail the discussion. Engineering may answer a question that was better left open. Management may reopen an issue that appeared settled. Meanwhile, the counterpart can observe priorities, constraints, internal disagreements and pressure points.

The lead negotiator is suddenly negotiating on two fronts: with the counterpart and with the own team.

The problem is not team size

Sending the hero negotiator alone is not the solution. Complex negotiations genuinely require different expertise. The more important question is how that expertise is organized once people enter the room.

Franz Beckenbauer's famous instruction before the 1990 World Cup final was essentially: go out and play football. But he did not send eleven talented players onto the pitch and hope that they would work things out. The line-up was clear. Roles were clear. There was a game plan. That structure allowed the players to use their individual judgment once the match started.

Negotiation teams work in much the same way. A legal expert remains a legal expert and an engineer remains an engineer. But expertise explains why someone is in the team; the role defines how that expertise contributes to the negotiation.

Who leads? Who observes? Who answers which questions—and which questions should remain unanswered? How does the team communicate internally? When does it interrupt? Who can change a position? What happens when the negotiation develops differently from what was expected?

These questions do not require a script. Quite the opposite. Negotiation architecture should create enough clarity that people can use their expertise and judgment without several team members simultaneously trying to play the same ball.

Often, defining the roles also answers the question of team size. Once someone's contribution is clear, it may turn out that they do not need to spend six hours in the negotiation at all.

Preparation is about how the team will operate

Negotiation teams commonly spend substantial time discussing objectives, positions and talking points. Less attention is paid to how the team itself will operate once the negotiation starts.

That distinction matters because many of the relevant questions have no single correct answer. Responsibilities overlap. Boundaries are ambiguous. Different approaches may work. The team therefore has to explore situations, challenge assumptions and gradually develop a shared understanding of how it wants to operate.

This is where AI can become useful.

Not because AI knows the correct team design, but because it can accelerate the exploration. It can challenge role definitions, identify overlaps and gaps, stress-test responsibilities and rehearse situations in which the planned approach begins to break down.

What happens if the discussion suddenly becomes technical? What may Engineering explain without giving away unnecessary information? What happens if Legal identifies a new risk during the meeting? Which decisions can the team make itself? When does it need a timeout? What happens if the counterpart attacks precisely where internal interests diverge?

A team can ask all these questions without AI. The difficulty is asking enough of them, exploring the consequences and challenging its own emerging logic at the same time.

AI provides a sparring partner for that process.

Give AI something concrete to challenge

The quality of that sparring depends on how explicitly the team has prepared the negotiation. If objectives, interests, mandates, roles and assumptions remain largely implicit, AI has little more than general negotiation knowledge to work with.

Once those elements are explicit, the discussion becomes much more concrete. AI can identify contradictions between roles, test scenarios against mandates, expose gaps in responsibilities and challenge whether the planned game actually fits the situation.

The objective is still not to create a script. A negotiation cannot be planned move by move, and a capable team needs room to react to what actually happens.

The objective is a shared game plan.

The coach will not score the goals. AI will not win the negotiation. But both can help make sure that when the team enters the field, everybody knows how they intend to play.

The Inversion Principle: Move AI Upstream

A common pattern of AI-supported decision-making appears straightforward: provide AI with the available information, ask it to analyze the situation and generate a recommendation, then have a human check the result.

For complex organizational negotiations, this creates an awkward division of labor. AI produces the downstream answer, while humans have to verify whether it actually fits the organization's objectives, assumptions, constraints, mandates, prior commitments and organizational reality.

The Inversion Principle reverses this emphasis. Instead of using AI primarily to produce answers and humans to validate them afterwards, AI is used further upstream: to structure the problem, challenge assumptions, identify inconsistencies, explore alternatives and prepare the reasoning from which humans make decisions.

From decision preparation to reasoning preparation

Consider the conventional workflow. A team provides contracts, meeting notes, presentations and contextual information and asks AI what it should do. The system interprets the material and produces a strategy, recommendation or negotiation position.

The difficult organizational questions then follow. Has the system understood the actual mandate? Which assumptions did it treat as facts? Has it recognized that different functions hold different views? Is its recommendation consistent with previous decisions? A knowledgeable human has to reconstruct much of the underlying reasoning before deciding whether the answer can be trusted.

The inverted sequence starts earlier. AI can help organize the available information, surface assumptions, identify relationships and contradictions, expose missing information and explore alternative interpretations before the team commits to a particular course of action.

Humans interact with that process rather than merely reviewing its final output. They provide context, reject implausible interpretations, resolve ambiguity, establish priorities and decide which assumptions and alternatives deserve further consideration.

Only then does the discussion move towards a decision. AI may still generate strategies and recommendations. The inversion does not restrict its analytical role. It changes where the collaboration begins.

Different strengths, different responsibilities

Much of the work surrounding a complex negotiation is not the final decision itself. Information has to be structured, assumptions tested, dependencies understood, alternatives explored and contradictions identified. As circumstances change, parts of that reasoning have to be revisited.

These are areas in which AI can provide substantial leverage. It can explore more alternatives than a team could reasonably examine manually, challenge emerging logic, compare scenarios and repeatedly test whether conclusions remain consistent with the information provided.

Humans contribute something different: organizational judgment. They understand context that may never be fully captured in the available information. They establish priorities, evaluate trade-offs and decide which risks, interpretations and commitments the organization is prepared to accept.

In simplified form, AI does more of the exploration and integrity work; humans retain responsibility for judgment and commitment.

This is not a distinction between AI doing administrative work and humans doing the interesting work. AI may identify patterns people have missed, challenge existing assumptions and propose better interpretations. The distinction concerns authority. AI can expand and challenge the reasoning space; people decide what the organization does with it.

Move checking upstream

The inversion also changes where checking occurs. In the conventional workflow, humans check the integrity of an answer after it has been generated. In the inverted workflow, much of that scrutiny happens while the problem is still being developed.

Is a proposed position inconsistent with an existing mandate? Does a new assumption contradict available information? Are two objectives in tension? Has the team prematurely excluded an alternative? Is something being treated as a fact that is actually only a hypothesis?

These questions do not determine what the organization should decide. They improve the reasoning before the decision is made.

This matters because downstream validation becomes increasingly demanding as AI output becomes more sophisticated. A polished recommendation may integrate dozens of assumptions and relationships that a human must reconstruct before deciding whether to trust it. Moving scrutiny upstream makes those assumptions and relationships visible while they can still be challenged.

Explicit representation makes the inversion stronger

The Inversion Principle does not require a particular technical architecture. A team can use it in a structured preparation session, an AI-supported workshop or a sequence of exploratory conversations. The essential move is to use AI to improve the reasoning process before asking it for the answer.

An explicit representation provides a particularly robust implementation. Objectives, interests, assumptions, constraints, stakeholders, mandates and dependencies can be made visible and maintained rather than reconstructed for every interaction. AI then has a stable basis for checking consistency and exploring alternatives, while humans can inspect and deliberately change the organization's current understanding.

The Three-Layer Architecture described separately takes this one step further by separating the underlying evidence, the maintained organizational understanding and the analytical views generated from it. But that architecture supports the Inversion Principle; it does not define it.

Beyond the human in the loop

Keeping humans responsible for consequential decisions remains important, but “human in the loop” says little about where human attention is most valuable.

If people become involved primarily when a finished recommendation needs approval, oversight occurs relatively late. The Inversion Principle moves human judgment closer to the assumptions, interpretations and trade-offs from which the recommendation will eventually emerge.

The resulting collaboration is iterative rather than linear. AI structures, explores and challenges. Humans provide context, judge and redirect. AI explores again. A recommendation becomes the result of that process rather than its starting point.

The objective is therefore not to have AI make better decisions for people. It is to use AI to reduce and strengthen the analytical work surrounding complex decisions, while concentrating human attention where it creates the greatest value: judgment, priorities and organizational commitment.

Three Layers, Three Different Jobs

Once an organization decides to maintain an explicit representation of a negotiation, a practical problem appears: the source material, the organization's current understanding, and the analyses derived from it are not the same thing.

A contract may be evidence. An assumption derived from it is an interpretation. A stakeholder map based on that interpretation is an analysis. Putting all three into the same information pool makes it increasingly difficult to distinguish what the organization knows, what it currently believes, and what somebody — or an AI system — has proposed.

The Three-Layer Architecture separates these functions. The first layer preserves the evidence. The second maintains the organization's current understanding of the negotiation. The third provides analytical views for particular reasoning tasks.

Layer 1: The project repository

The project repository contains the evidential record of the negotiation: contracts, correspondence, meeting minutes, presentations, financial analyses, technical studies and other relevant material.

Its purpose is preservation rather than interpretation. It can legitimately contain old proposals, contradictory information and documents that no longer reflect the current situation. They remain part of the history of the negotiation and may later be needed to understand or challenge the organization's current view.

The repository therefore answers a relatively simple question: What information do we have?

Layer 2: The persistent organizational representation

The second layer answers a different question: What is our current understanding of the negotiation?

This representation is deliberately selective. It captures the stakeholders, issues, interests, assumptions, constraints, mandates, decisions and unresolved questions that currently matter for organizational reasoning.

The distinction from the repository is important. An early presentation may contain an assumption that has since been rejected. The presentation remains part of the evidence, but the assumption should no longer silently determine today's reasoning. Conversely, a management decision may establish a new constraint that should become part of the organization's current representation even though most of the underlying project material predates it.

The representation is therefore persistent, but not static. New evidence can challenge it and organizational decisions can change it. What persists is a shared organizational reference whose changes are deliberate and visible rather than a new reconstruction of the negotiation every time somebody asks a question.

Layer 3: Generated reasoning representations

The third layer contains the views created for particular reasoning tasks: stakeholder maps, issue structures, dependency analyses, timelines, scenarios, negotiation strategies, executive briefings or decision matrices.

Unlike the persistent representation, these views do not need to become permanent organizational objects. A team examining stakeholder influence needs a different view from one examining commercial dependencies. Management may need a concise decision view, while the negotiation team may need a detailed issue structure.

AI makes these representations comparatively inexpensive to generate. Multiple views can be produced from the same underlying organizational representation, explored, compared and discarded when they are no longer useful.

The important distinction is authority. An AI-generated scenario or stakeholder map remains an analysis. It does not become part of the organization's accepted understanding merely because it is plausible or useful. If an analysis reveals something that should change that understanding, the change is made explicitly in the persistent representation.

Why separate the layers?

A conventional AI conversation tends to collapse all three layers. Documents enter as context, AI interprets them and generates analyses, and those analyses remain in the conversation where they can influence what comes next. Evidence, interpretation and recommendation gradually become part of the same conversational history.

That is convenient for an individual interaction but problematic as an organizational basis for reasoning. A provisional inference can acquire apparent authority simply because it remains in the context. An obsolete document can influence a new analysis because its status is unclear. Different users may construct different versions of the same negotiation from the same underlying material.

The Three-Layer Architecture introduces a simple separation: preserve the evidence, maintain the current organizational understanding, and generate analysis as required.

Stability below, flexibility above

The separation allows each layer to serve a different purpose. The repository can be comprehensive because it does not need to be cognitively manageable. The persistent representation can be selective because the underlying evidence remains available. Generated representations can be exploratory because they cannot silently rewrite the organization's accepted understanding.

This also preserves one of AI's most useful capabilities: the ability to explore alternatives. The objective is not to make every AI analysis identical. Different representations can reveal different patterns and support different decisions.

What needs continuity is not every answer, but the organizational state from which those answers are produced. The repository provides evidential continuity, the persistent representation provides organizational continuity, and generated representations provide analytical flexibility.

Together, the three layers allow an organization to change its understanding without losing its history — and to explore alternatives without confusing exploration with commitment.

Why Represent a Negotiation at All?

Organizations rarely manage complex activities from memory alone. Finance has financial statements, engineering has drawings and models, and projects have plans and schedules. None of these contains everything there is to know about the underlying business, plant or project. They select and organize what matters so that people can reason about the same object.

They all are representations. A map, for example, is not the landscape itself. It deliberately leaves most things out while making roads, distances and connections visible because they matter for navigation.

Negotiations are a striking exception. A complex negotiation may involve numerous issues, internal and external stakeholders, different interests, dependencies, mandates, assumptions and unresolved questions. Yet organizations rarely maintain an explicit representation that brings these elements together. Instead, the negotiation exists partly in documents, partly in meetings and partly in the heads of the people involved.

Experienced negotiators construct such representations mentally. The problem is not that this does not work. The problem is that different people may be working with different representations without realizing it. Assumptions remain implicit, relationships obvious to one function may be invisible to another, and much of the reasoning behind a decision disappears once the people involved move on.

Information is not yet a representation

Most negotiations already generate plenty of information: contracts, presentations, spreadsheets, emails, meeting minutes and increasingly AI conversations. Collectively, these may contain almost everything the organization knows about the negotiation.

But having the information is not the same as representing the negotiation.

A representation makes selected elements and relationships explicit: who the relevant stakeholders are, which issues are open, what interests appear to drive the parties, which assumptions remain uncertain, where issues depend on each other, what has been decided and what remains unresolved.

The purpose is not to capture everything. Like a map, a useful representation is selective by design. Nor does making something explicit make it true. An assumption about another party remains an assumption. The advantage is that it can now be examined, challenged and revised rather than remaining hidden inside someone's reasoning.

Why don't organizations already do this?

The main obstacle has historically been cost. Constructing a representation requires information to be extracted, structured, interpreted and continuously updated. Negotiations make this particularly demanding because the underlying situation keeps changing.

For many organizations, relying on experienced people, meetings and documents was therefore a rational solution. The additional benefit of an explicit representation did not necessarily justify the effort required to maintain it.

AI changes that calculation.

Large language models can extract information from heterogeneous documents, connect related observations, identify inconsistencies and help maintain structured information as the negotiation evolves. Tasks that previously required substantial manual effort can increasingly be performed at comparatively low cost.

Why not just give AI all the documents?

That appears to offer an even simpler solution. If AI can read the documents directly, why construct a representation at all?

Because the AI still has to construct some working understanding of the negotiation before it can reason about it. Without an explicit representation, that understanding is reconstructed inside each interaction. Another user, another conversation or another model may construct a somewhat different one.

More information does not solve this problem. An old presentation and a new management decision may both be available. A meeting note may contain an assumption that was later rejected. Two functions may interpret the same issue differently. Access to all of this information does not by itself establish what the organization's current understanding is.

The problem has not disappeared. It has become less visible.

AI changes the economics of representation

This creates a choice about where to invest the productivity gain from AI. Organizations can use cheaper analytical capability to generate more summaries, scenarios and recommendations. Or they can invest part of that gain one step earlier and use AI to construct and maintain the shared representation from which those analyses are generated.

The second approach creates a persistent organizational asset. Participants can begin from an explicit current understanding rather than repeatedly reconstructing the negotiation from documents, meetings and memory. Different analytical views can then be generated from the same basis.

This does not mean building the richest representation technically possible. Every additional element still has a cost. The appropriate test is practical: does making something explicit improve organizational reasoning enough to justify maintaining it?

AI shifts that threshold because it substantially reduces the cost.

A shared basis, not a calculated answer

A negotiation representation will never function like a financial statement. Negotiations contain uncertainty, incomplete information and strategic behavior that cannot be eliminated through better representation.

That is not the objective.

The value lies in making explicit what the organization currently knows, what it assumes, where people disagree and what remains unresolved. A questionable assumption that everyone can see is different from one silently embedded in an individual's reasoning.

A representation therefore does not tell an organization what to do. It provides a better shared object to think with.

AI makes constructing and maintaining that object much less expensive than it used to be. The next question is what should be preserved in that object, what should remain as underlying evidence, and what can be generated only when needed.

When Organizations Negotiate Against Themselves

One of the most dangerous moments in a project is when everybody agrees too early.

A discussion starts with several suppliers, timelines, specifications and implementation approaches. A few meetings later, the preferred supplier has become the only supplier, the target date has become fixed, and a technical preference has become a requirement.

It feels like progress. Decisions have been made and the project has become more concrete.

But something else has happened at the same time: the solution space has become smaller. By the time the external negotiation begins, much of the organization's flexibility may already have disappeared—not because the counterpart constrained it, but because the organization did.

Decide the What before fixing the How

Organizations often find it easier to converge on concrete solutions than to align on foundational questions. Selecting a supplier or defining a specification creates visible progress. Agreeing what success actually means, which trade-offs are acceptable and where genuine flexibility exists is more difficult.

The result is that the How gets decided before the What has been sufficiently explored.

This matters in negotiations because optionality creates room to maneuver. Every unnecessary commitment removes a degree of freedom. A fixed supplier eliminates alternatives. A fixed specification removes possible technical solutions. A fixed timeline can turn time into leverage for the other party.

None of those choices is necessarily wrong. The problem is making them without recognizing that they are choices.

Not every constraint is a hard constraint

Most organizations can list the constraints surrounding a negotiation. Far fewer can say which of them are genuinely fixed.

Some constraints are hard. Legal requirements or safety conditions may leave no meaningful room for negotiation. Others look hard because they have not been questioned for some time: technical specifications, preferred partners, timelines, governance models or budgets.

That does not make them irrelevant. Changing them may have significant consequences. A different technical specification may require redesign. Changing a partner may delay the project. Extending a timeline may have substantial economic cost.

But an expensive choice is not the same thing as an impossible choice.

Once a constraint can be relaxed at a cost or under particular conditions, it becomes part of the solution space. It can be evaluated against other trade-offs rather than treated as a boundary outside the negotiation.

Good preparation therefore does not simply define constraints. It tests them.

AI can expand—or shrink—the solution space

AI creates an interesting tension here.

Give an AI system a fixed supplier, a fixed timeline, a fixed specification and a fixed implementation approach, and it will usually accept those conditions and optimize within them. The resulting analysis may be coherent, persuasive and sophisticated.

It may also be solving a problem that the organization unnecessarily created for itself.

This is not primarily an AI failure. The system is answering the question it was given. If the assumptions embedded in that question remain unchallenged, more capable analysis can simply produce increasingly sophisticated answers inside an increasingly narrow solution space.

Used differently, the same capability can help reopen that space.

What happens if the timeline moves? Which alternatives become possible if a specification changes? What would need to be true for another supplier to become viable? Is a particular requirement genuinely non-negotiable, or does changing it simply have a price? Which alternatives did the team stop considering several decisions ago?

AI is useful here not because it knows which constraint should be relaxed. It can help a team explore more combinations, trace consequences and challenge assumptions faster than the team could reasonably do manually.

Preserve choice until there is a reason to give it up

The purpose is not to keep every option open indefinitely. Decisions have to be made, and narrowing the solution space is eventually necessary.

The question is when and why.

If an organization deliberately chooses a supplier because the advantages outweigh the value of maintaining alternatives, optionality has been exchanged for something useful. If the supplier becomes effectively fixed because everyone gradually started treating a preference as a requirement, optionality has simply disappeared.

That difference becomes particularly important before a negotiation. Every option retained gives the organization another possible response when circumstances change. Every assumption tested reduces the chance that the negotiation is being conducted inside boundaries that were never actually necessary.

The best strategy may ultimately be exactly the one the team preferred at the beginning. But after the alternatives and constraints have been tested, it is a choice rather than an assumption.

Many difficult negotiations become difficult before the counterpart ever enters the room.

Diverging Definitions of Value in Projects

Project discussions where nobody challenges the case—and nobody endorses the project—are surprisingly common.

The market assumptions have been reviewed. The technical concept is sound and has been thoroughly vetted. The economics have been stress-tested repeatedly. Sensitivity analyses have been prepared for the relevant risks.

By most standards, it is a well-prepared investment case.

The project is rejected.

What appears surprising is often almost inevitable.

The problem is frequently not in the numbers. The problem is that the people around the table are applying different worldviews and different definitions of value.

Some view the project primarily through the lens of financial return. Others focus on strategic positioning. Some worry about secondary effects on portfolio or reputation. Others see an opportunity to build capabilities that may become valuable years later. All of them carry different expectations within the given uncertainty.

The discussion appears to be about assumptions, forecasts, and probabilities.

In reality, it is often a negotiation about value.

Nobody challenges the case directly. Instead, stakeholders request additional analyses, alternative scenarios, and further validation. The discussion remains analytical while the disagreement itself does not.

The question is not whether the analysis is correct. The question is what the analysis needs to prove.

Until that is understood, even the best analysis often generates heat rather than alignment.

What is often overlooked is that disagreements about value rarely emerge on their own. They are usually rooted in different assumptions about reality.

People disagree about what is attractive, risky, urgent, strategic, or worth pursuing because they are operating from different interpretations of the situation. They expect different market developments. They assess competitors differently. They assign different probabilities to technological, commercial, or regulatory outcomes.

The disagreement about value is therefore often downstream of a disagreement about assumptions.

And that disagreement is frequently hidden.

The Expensive Word: “Obvious”

One of the most expensive words in projects and negotiations is “obvious.”

Teams routinely describe market opportunities, customer needs, strategic rationales, and risks as obvious. Yet what is often presented as a fact is actually an assumption that has become accepted.

In preparing complex negotiations and strategic decisions, hard facts are often limited. Much of the discussion is built on interpretations, projections, judgments, and expectations about the future. The market may develop in a certain direction. A competitor may react in a certain way. A technology may mature. A regulator may intervene. A customer may value a particular feature.

None of these statements are facts.

They are assumptions.

Yet teams can spend weeks debating recommendations while never examining the assumptions that produced them.

The disagreement sits further upstream.

People are working from different mental models of the situation. Once that happens, they also begin to define value differently. What looks attractive, risky, urgent, strategic, or worth pursuing depends largely on the assumptions underneath.

The objective is not to achieve consensus on reality. Nor is it to turn every assumption into a hard fact. That is neither realistic nor necessary. Assumptions are practical approximations of reality.

What matters is achieving sufficient alignment on which parts of reality matter in a given context and what people are assuming about them. The point is not to eliminate uncertainty. The point is to make implicit assumptions explicit and part of the shared context.

Only then does it make sense to debate value, compare alternatives, or agree on the standards by which competing options should be judged.

There is a second reason assumptions matter.

So far, this is primarily an internal alignment problem. In negotiations, however, facts and assumptions should not be treated the same way. The distinction influences risk allocation, objective criteria, mechanisms for managing competing interests, and ultimately what parties are willing to agree to.

A party may accept a contractual commitment based on a verified fact while demanding safeguards, contingencies, or risk-sharing mechanisms when the same issue is based on an assumption. The distinction shapes both the substance of the agreement and the willingness to make it.

Why AI Amplifies the Problem

The challenge becomes even more relevant when AI enters the process.

One of the most dangerous AI outputs is not a hallucination.

It is a well-reasoned answer built on an incomplete picture.

When humans argue in a project or negotiation, uncertainty often becomes visible. People hesitate. Contradictions emerge. Questions get asked. They recognize that something important may be missing. We have also learned that inconsistency and incompleteness are often symptoms of incorrectness.

Large language models behave differently.

They generate the statistically most plausible continuation of the context they are given. Gaps are filled by inference. For a language model, careful inference and aggressive extrapolation are fundamentally the same operation.

Give an LLM an incomplete description of a negotiation, investment case, project, or strategic decision and it will still produce a coherent answer. The result will be logically structured, persuasive, and internally consistent because generating coherent language is precisely what the system is designed to do.

The problem is that humans instinctively associate coherence with correctness and understanding of context.

With LLMs, that inference is far less reliable.

LLMs answer your question, not your problem.

Humans and organizations operate with countless implicit assumptions, priorities, and value judgments. We often struggle to make them explicit, even to ourselves. As a result, important parts of the context frequently remain unstated.

LLMs have no reliable way of distinguishing between information that is missing because it is irrelevant and information that is missing because nobody thought to mention it.

In many cases, the risk is not faulty reasoning.

The risk is flawless reasoning applied to an incomplete representation of reality.

This is why effective AI use is often less about generating answers and more about exposing assumptions, identifying context gaps, and revealing what may be missing from the picture.

Some of the most valuable AI outputs are therefore not answers.

They are questions.

Context Precedes Clarity

Strategic decisions, negotiations, investment cases, and AI-supported analyses are often treated as separate disciplines. In practice, they frequently fail for the same reason.

The underlying assumptions remain hidden.

People debate conclusions while operating from different definitions of value. Analytical work becomes disconnected from the standards by which success is judged. AI then produces increasingly persuasive answers within the boundaries of an incomplete picture.

The result is not necessarily bad reasoning.

It is reasoning disconnected from the reality that ultimately determines success.

Before debating recommendations, organizations need sufficient alignment on the context in which those recommendations will be evaluated.

Because context precedes clarity.

And clarity precedes good decisions.

This article was originally published on LinkedIn . Slightly modified from the original version to fit the format.

Preparation for complex negotiations has never been a clean process. It was iterative, loosely structured, and often frustratingly slow. At times outright messy. Compared to production, logistics, or sales processes, negotiation preparation lacked linearity and discipline. It turns out that this was not simply inefficiency. It served a function.

The friction inside these processes forced implicit constraints to surface over time. Finance challenged assumptions. Legal raised concerns. Engineering hesitated to commit. Commercial teams questioned feasibility. Historical precedents reappeared late in discussions. Stakeholders introduced objections that initially seemed vague, but later proved relevant. None of this was efficient. But it ensured that a meaningful portion of what could later derail a negotiation became visible before teams entered the room.

Much of what mattered sat below the waterline. The process, however imperfect, gradually revealed it.

AI has fundamentally changed these dynamics. Today, comprehensive negotiation strategies can be generated within minutes. The outputs are structured, coherent, persuasive, and operationally useful, even when the underlying input is incomplete. This is where the problem begins.

Where traditional preparation processes slowed down when encountering uncertainty, missing information, or unresolved boundary conditions, AI quietly proceeds. It fills gaps with plausible assumptions, smooths inconsistencies, repairs logic, and produces results that appear complete. The distinction between what is known and what is inferred becomes increasingly difficult to detect. Invisible assumptions are significantly harder to manage than visible gaps.

Previously, missing information created friction. Today, it creates output. But fluency is not correctness. And it is certainly not completeness.

At the same time, the informal organizational mechanisms that once challenged assumptions are less likely to engage. There is less visible tension in the process. Fewer escalation signals. Fewer indications that something remains unresolved. The result appears internally consistent and therefore credible.

The danger is subtle because it rarely appears in simplified demonstrations or standard test cases. Those environments tend to be closed, controlled, and structurally convex. Real negotiations are not. Negotiations are shaped by incomplete information, shifting interests, internal politics, approval structures, relationship dynamics, ambiguous mandates, and evolving power relationships. Many of the relevant variables are neither explicit nor stable. This is where failure emerges. Not in the visible preparation process, but below the waterline.

A negotiation is not merely a data problem. It is an interpretation problem. What matters most is often not formally documented. Interests, fears, leverage, real influence power, internal constraints, walk-away points, concession structures, and political sensitivities frequently remain implicit until pressure exposes them.

AI does not automatically understand what matters in a specific relationship, mandate, organizational context, or negotiation moment. It does not inherently know which assumptions should be challenged, which constraints are negotiable, which risks require escalation, or which concessions fundamentally alter the balance of a negotiation.

The current AI discussion often becomes technical too quickly. Better access to knowledge, tools, workflows, and enterprise systems matters. But the technology itself is only the enabler. The central challenge remains organizational and cognitive.

This is also where AI becomes genuinely valuable. Not because it magically produces the right strategy, but because it can help expose the quality of preparation before teams enter the room.

AI can pressure-test assumptions. It can map stakeholder interests, identify missing information, simulate counterpart perspectives, generate scenarios, prepare concession structures, and expose inconsistencies in reasoning. Used correctly, it can make preparation more explicit, more complete, more repeatable, and faster. But this only works if AI supports the process rather than silently replacing unresolved human judgment with plausible-looking assumptions.

The distinction matters.

In human organizations, we would never expect a new colleague to solve a complex strategic problem simply by reading annual reports, strategy decks, and role descriptions. We onboard, align, challenge, brief, review, escalate, and clarify in order to surface reality. Negotiation preparation requires the same discipline.

AI can accelerate and strengthen that work. What it cannot yet do reliably is determine which unresolved assumptions actually matter before reality tests them.

We have significantly reduced the friction in negotiation preparation processes. What we have not yet replaced is the function that this friction fulfilled.

Because in negotiation, preparation is not the phase before the real work. Preparation is the real work.

This article was originally published on LinkedIn . Slightly modified from the original version to fit the format.

C-TO-BE: Hello Markus, you’ve led high-responsibility negotiations for more than 20 years. Looking back, what’s the core takeaway for you?

Markus: Negotiations can feel very different depending on context – company cultures, regions, people, and pressure. What I’ve learned is how important it is to treat negotiation as a discipline in its own right, not just as a means to an end. When you do that, you start recognizing patterns and reading dynamics more clearly. That structure frees up mental space when it matters most.

C-TO-BE: Then AI entered the picture. What changed with AI?

Markus: At first, it was simply striking to see what suddenly became possible through AI. Today, AI drafts emails, summarizes documents, runs scenarios, and supports preparation – on both sides of the table. At some point, Jutta and I began taking a closer look at the topic and revisiting it again and again.
AI is clearly transformative – but only if it’s embedded properly in negotiation contexts. That led to a coherent concept: C-TO-BE Negotiation Intelligence, connecting capabilities, process, and AI. That’s what we outline in our 12-part series.
Ignoring AI isn’t an option, and the gap between AI-savvy and hesitant negotiators is continuing to widen.

C-TO-BE: We often hear negotiators ask: “Why not just ask ChatGPT?”

Markus: You absolutely can – and we do it ourselves all the time. AI is often intuitively convincing, and sometimes that’s sufficient. But “convincing” doesn’t mean correct, complete, or responsible. In negotiations, later corrections are often impossible. Once something is said, it’s out there: a proposal is made, information is shared, or an ego is bruised. That’s why intuition alone isn’t enough.
C-TO-BE Negotiation Intelligence adds a frame: it integrates AI’s strengths while keeping decisions reviewable and accountable.

C-TO-BE: A question many negotiators worry about: Will AI eventually replace us at the negotiation table?

Markus: That’s a fascinating – and naturally speculative – question. Technically, AI will negotiate more autonomously in certain settings, especially repetitive or distributive ones. But responsibility remains human – ethically and legally. AI will change how we negotiate, not who is accountable. And because AI is non-deterministic, clear processes and rules become even more important.
That will only change if and when we move from today’s generative AI to a truly general form of artificial intelligence – a shift that would be far more fundamental.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

If your team’s negotiations still feel like heroic firefighting, you’re paying a premium - in time, value leakage, and stress.

2026 changes the game. Negotiations have become faster and more complex. Multiple stakeholders join every deal; compliance expectations tighten; and your counterpart likely shows up with a structured playbook to his own benefit. Reactivity no longer wins - it drains resources and exposes risk. The solution is not “more effort,” but a maturity shift: from reactive tactics to strategic execution. Smarter, not harder.

That shift requires three steps:

1. Capability at the individual level. Negotiators who know how to prepare systematically, interpret patterns, and stay calm under pressure.

2. Structure at the organizational level. The joint use of shared frameworks and common language that make best practices in negotiations repeatable. Without structure, effectiveness drops in the short term - and development disappears in the long term.

3. Focused AI-support in negotiations. The targeted use of technology that enhances clarity and speed without pretending to replace human judgment. AI becomes an advantage only when it is embedded in structured processes and guardrails.

Together, these three elements form Negotiation Intelligence – enabling your organization to deliver quality, speed, and cost advantages consistently. Negotiation becomes a scalable strength, not an individual talent or reliance on instinct.

Research from the Program on Negotiation at Harvard Law School* reinforces this: thorough preparation is the highest impact action negotiators can take. Underprepared teams make unnecessary concessions, miss sources of value, and walk away from beneficial agreements. High-performing teams invest time upfront to define interests, alternatives, and scenarios - turning uncertainty into strategy. The result? More confident teams, more transparent leadership, and risks addressed early instead of discovered too late.

*) Program on Negotiation Staff. (2025). 5 Tips for Improving Your Negotiation Skills. Program on Negotiation at Harvard Law School. Negotiations are won before you enter the room. This text appears on a clean white background.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Winning negotiations isn’t about luck or charisma. It’s about how effectively your organization builds, activates, and scales its negotiation system.

The Negotiation Intelligence model strengthens that system through three progressive tiers that reinforce one another.

That shift requires three steps:

Tier 1. Capabilities: The capability to negotiate is the foundation that drives performance. Negotiators learn to prepare deeply, analyze dynamics of negotiations, apply tactics, navigate complexity, and lead discussions with clarity. They replace guesswork with confidence and turn skills into repeatable strengths.

Tier 2. Structure: Where capabilities turn into consistent results. With the WIN (Workflow-Intelligence Negotiation) framework, we anchor behaviors, standardize preparation, and align teams around a shared process - from mapping the negotiation field to executing with precision in the meeting. Our six-phase structure guides and accelerates the entire cycle: Probe - Plan - Plot - Prepare - Perform - Pursue. The outcomes are reduced variance, fewer detours and recursions, faster decision cycles, higher first-time-right quality and clearer collaboration. Structure doesn’t restrict talent - it multiplies it.

Tier 3. Focused AI Support: Finally, we add AI where it truly helps synthesize information, generate scenario drafts, and support reflection. It’s a negotiation accelerator leveraging AI to create efficiency. A trainer, a coach, a sparring partner at your fingertips. Not a replacement for responsibility.

The result is visible across metrics: faster prep, fewer errors, and stronger outcomes. Complex negotiations will never become predictable. But they do follow a well-established logic. When negotiation becomes structured and facts-driven, success scales across borders, hierarchies, and roles.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Before templates or AI, one constant drives every negotiation: the human being in the room. Personal capabilities build the core that turn reactive negotiators into strategic operators - the foundation every high-stakes team relies on.

Negotiation capabilities enable:
- Preparation mastery: Turning scattered information into strategic clarity.
- Pattern recognition: Spotting behaviors, leverage points, and recurring traps early.
- Tool fluency: Applying BATNA, ZOPA and concession planning with intent.
- Complex leadership: Orchestrating multiple stakeholders without losing control.
- Handling difficulty: Staying composed with tough, emotional, or political counterparts.
- Decisiveness: Making aligned trade-offs when pressure peaks.

This is what C-TO-BE conveys in Negotiation Skill Trainings. The programs blend deliberate practice, realistic simulations, and business actor training. Participants don’t just learn - they perform under pressure, then debrief how they acted and why.

A comprehensive meta-analysis* by ElShenawy (2004) reviewing 57 controlled studies, shows that negotiation training significantly improves economic and relational outcomes. Training doesn’t just transfer knowledge - it shapes cognitive routines and emotional control. We observe the same strong impact in over 20 years of supporting our clients in developing their negotiation skills.

Why it matters: Confidence under pressure is contagious. Skilled teams execute faster, follow structure more consistently, and make smarter decisions - across sales, procurement, legal, project management, and beyond.

Capabilities build the people muscle. Structure builds the process backbone.

*) ElShenawy, E. (2004). Does Negotiation Training Improve Negotiators’ Performance? International Journal of Conflict Management, 15(4), 305–335.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Even excellent negotiators fail when time pressure or handovers destroy preparation.

Aligned structure fixes that (WIN Workflow-Intelligence Negotiation), a six-phase loop from strategy to follow-through:

1. Probe – Understand market, stakeholders, and context.
2. Plan – Define goals, limits, and success metrics.
3. Plot – Craft tactics, arguments, and fallback paths.
4. Prepare – Finalize materials, team roles, and logistics.
5. Perform – Execute with discipline and presence.
6. Pursue – Follow up and secure implementation.

Each step has structure, templates, checklists, and therefore embedded compliance. Goals, strategy and tactics are developed and aligned. Offers, concession plans and walk away points become clear. Escalation paths are defined; roles are clear before anyone joins a call.

Statistical evidence and own experience demonstrate how structured processes cut critical-task errors and improved team communication. The same principle applies to negotiations: a disciplined, stepwise process prevents oversights and supports better decision quality.

The business logic is simple: reduce chaos to increase value. Teams have an aligned game plan before they enter the room. Not instincts, not multiple variants and interpretations of vague negotiation targets. WIN removes variance, cuts prep time, and ensures institutional knowledge survives turnover.

Once the backbone is in place, AI can safely amplify and speed up what humans already do well. This is the essence of the 3-tier approach. Structured processes provide guard rails to AI. AI creates efficiency in the structure.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

What does structured negotiation look like in action?

• Consistency: Everyone follows the same WIN rhythm. All teams involved speak the same language. Different players within the organization align their interests before negotiations commence.
• Speed: Structure and validated steps replace intuition and perceived priorities.
• Risk reduction: Legal, compliance, and financial checks happen early and proactive, not in panic mode and reactive.

A process defines roles and agenda control upfront. That keeps meetings focused and decisions crisp. No lost emails, no “who owns this?”- moments.

Research on Toyota’s standardized work (Harvard Business Review, Spear & Bowen, 1999) found that clearly defined processes paradoxically increase flexibility and innovation. By standardizing the routine parts, team’s free cognitive bandwidth for creative problem-solving. The same effect structured negotiation systems achieve.

Results:
1. Shorter preparation cycles are necessary.
2. Fewer recursions and escalations are noticed.
3. Clearer follow-through after deals take shape is possible.

The system doesn’t restrict creativity, it protects it. Negotiators can think strategically because the basics are handled. This stability is what makes AI integration (Tier 3) not just possible, but safe.

AI plays to its strengths, crunching endless data and distilling proposals from that, to be chosen and decided by humans. See how this works in our next post.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

AI is already inside our negotiations, whether we acknowledge it or not. People draft emails with AI, summarize contracts with AI, and even ask AI how to respond to tough stakeholders. Large language models (LLMs) are powerful pattern machines: they generate fluent text, structure messy information, and help us think faster. Used well, they can dramatically reduce preparation time and make complex situations easier to grasp.

But there’s a catch. LLMs don’t “understand” your business, your context, or your risks in the same way you do. They predict plausible sentences based on training data. That’s their strength and their weakness.

Three AI challenges matter most in negotiation contexts:

1. Hallucinations or confabulations (confident nonsense). LLMs can produce output that sounds right but is simply wrong or fabricated. In low-risk scenarios this is annoying. In negotiations, legal, or compliance-heavy environments, it’s dangerous.

2. Hidden bias and blind spots. These systems learn from existing data and inherit the distortions, biases, and gaps of that data. Without guardrails, they can normalize one-sided perspectives or reinforce unhelpful patterns in how we speak and decide.

3. Illusion of competence. Because AI answers so fluently, it’s easy to confuse good form with good judgment. We start trusting wording as if it were wisdom. We confuse apparent plausibility with truth. That’s where over-reliance creeps in.

For more complex tasks, attention decay and lack of metacognitive abilities in current models can be more than a nuisance. Researchers like Emily Bender and colleagues* describe LLMs as “stochastic parrots”: systems that stitch together likely word sequences without genuine understanding. There’s a lot that can be discussed and disputed about that notion including whether “understanding” could be an emergent property of complex statistical correlation or something qualitatively different. But their point isn’t that AI is useless, it’s that we must be honest about what it can and cannot do. And let’s be fair: getting the facts wrong, showing bias, and pretending competence is nothing we haven’t seen from humans before :-)

For negotiation leaders, the core design question for AI implementation is very similar to any good governance: How do we leverage AI’s strengths (speed, structure, patterning) while containing its weaknesses?

*) Bender, E. M., Gebru, T., McMillan-Major, A., & Mitchell, M. (2021). On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of FAccT’21.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Once we understand both the strengths and limits of AI, the real question becomes: How do we use it responsibly and effectively in negotiations?

The answer sits in something we already built earlier: Structure. Defined processes don’t just improve human preparation; they create the governance system that enables AI use. Clear workflows, defined decision points, and shared templates become the rails that keep AI aligned with your intentions rather than its own statistical guesswork.

Here’s what that looks like in implementation:

• Structured inputs: AI modules operate on curated information – the same templates, scenario plans, and stakeholder data your team already uses. This reduces the risk of hallucinations because the AI works within a controlled context rather than pulling from the entire internet.
• Role boundaries: Each negotiation phase defines what AI may support (structuring, summarizing, drafting) and what remains strictly human (judgment, trade-offs, mandate decisions). Because responsibilities are explicit, over-reliance becomes far less likely.
• Human review gates: Every AI output passes through validation points built directly into the workflow. No dark corners. No invisible automation. But transparent, reviewable steps.
• Traceability: Because the process is documented from Probe to Pursue, AI-generated elements are anchored in the process. That’s governance by design – not after-the-fact control.

This balance isn’t just philosophical – it’s practical. The U.S. National Institute of Standards and Technology (NIST)* stresses that AI must remain human-centered, with clear oversight and accountable decision-making. This is exactly how AI implementation must be designed: AI augments human negotiation decisions, never replaces them.

When AI operates inside a structured human process, it becomes a disciplined accelerator – reducing prep time, improving clarity, and strengthening consistency without ever taking the wheel.

*) NIST (2023). AI Risk Management Framework (SP 1270) – emphasizing human responsibility, governance, and controlled integration of AI in decision-critical environments

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Let’s make AI tangible. Our Negotiation Intelligence Modules aren’t chatbots – they’re role-specific assistants built around how real negotiators work. Each is designed to slot seamlessly into your workflow:

• Checklist Coach: Walks you through preparation, prompting the right questions at each step. It ensures that context, strategy, and counterpart analysis are never rushed.
• Negotiation Advisor: Drafts strategic options based on your real negotiation data – not random internet text. It acts like a senior sparring partner, not a script generator.
• Training Modules: Simulate counterpart personalities and intercultural nuances. Teams practice difficult situations before facing them for real.

Outputs are structured, traceable, and editable – designed for auditability and learning.

Why does this matter? Because data* support own experience that AI assistant improve productivity. Findings validate the measurable business value of AI augmentation when focused on task-specific guidance.

- Preparation time drops
- Negotiation success rates rise through clearer argumentation.
- Value leakage decreases as unforced concessions fall.
- Onboarding cycles shorten dramatically – new joiners become effective faster.

The key isn’t technology; it’s relevance. AI earns its keep by turning preparation into precision – freeing professionals to focus on what they do best: deciding, influencing, leading.

*) Source: Brynjolfsson, E., Li, D., & Raymond, L. (2023). Generative AI at Work. National Bureau of Economic Research, Working Paper No. 31161.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Imagine you’re about to enter a tough negotiation – a supplier with leverage, a stakeholder under pressure, or a client demanding the impossible. Now imagine you’ve already practiced that conversation in a safe, realistic environment.

That’s what our conversational AI training delivers. It doesn’t “teach” theory – it creates live simulations.

Here’s how a session unfolds:
1. Context setup: You define your role, objectives, and boundaries.
2. Scenario generation: The system crafts a counterpart with specific motives and style – sometimes logical, sometimes emotional.
3. Active dialogue: You negotiate. The AI adapts its tone, pace, and tactics dynamically.
4. Immediate feedback: You receive live hints during the talk and a detailed summary afterward – including argument strength, emotional handling, and structural balance.

It’s experiential learning at digital speed. It's “stress rehearsal” – the kind that builds calmness, agility, and conviction before the real meeting.

Research* shows that role-play and scenario-based practice substantially enhance negotiation performance and transfer of learning. Deliberate simulation allows professionals to test strategies safely and internalize effective behavioral patterns before real-world application.

The business benefit is measurable: teams become faster, more composed, and far more consistent in tone and approach. They no longer “hope” they’ll perform well – they know because they’ve practiced, reflected, and improved.

Negotiation confidence is not a personality trait; it’s a trained skill. And this is how we train it.

*) Source: Movius, H. (2008). The Effectiveness of Negotiation Training. Negotiation Journal, 24(4), 509–53

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Great negotiators don’t rely on memory – they rely on structure. And in complex environments, structure is what protects value.

That’s where our AI-supported WIN checklists modules come in. These modules transform the six phases of the WIN process – Probe, Plan, Plot, Prepare, Perform, Pursue – into a living, interactive guide for every negotiation.

They don’t replace thinking; they make sure the right thinking happens.

Here’s how it works:
- Before the negotiation: The checklist helps teams clarify goals, map stakeholders, and test fallback options. It prompts what matters most: What’s our real objective? What assumptions are untested? Who owns which decision? The result – cleaner preparation, fewer surprises, and shared clarity before the first call starts.
- During execution: The structure keeps the team aligned under pressure. No matter who speaks, everyone follows the same rhythm – strategy stays intact even when dynamics shift.
- Afterward: The same module supports reflection and follow-up.

Outcomes, lessons, and next steps are captured in minutes – feeding back into templates, playbooks, and future prep.

Over time, these digital checklists become a knowledge backbone. They capture experience, standardize quality, and let AI highlight recurring risks or blind spots.

The system learns as you do – step by step, conversation by conversation. That’s the bridge from Process to AI: structure enhanced by intelligent guidance.

Not automation for its own sake, but a framework that ensures every negotiation follows best practice without slowing people down.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

AI is a tool, not a strategy. This is why strategy still starts traditional - and where AI truly changes the game, especially in negotiations.

Unless you plan to build an LLM platform to compete with OpenAI or Google, AI is probably not your strategic endgame. AI is a strategic enabler.

Strategy itself remains old school: What really moves the needle on cost or market value? Where are your must-win battles? What will matter in 12–36 months, not just this quarter? And where is competition heading next? Skate to where the puck is going to be, not where it has been.

Nothing new there. Except that you may use AI to answer some of those questions (then see "AI in negotiations: the promise and the challenge").

What you do in the field of AI only comes into play in this context. It does so through two very concrete questions:
1. Internal: Where can AI help you execute your strategy faster, more consistently, or with less friction? In short: where is AI part of the solution?
2. External: Where is AI eroding your existing edge and where could it sharpen value creation instead of commoditizing it? In short: where will AI likely become a high risk factor, and how do you respond?

Be careful: even if AI isn’t the strategic target, it will be transformative. Resistance isn’t an option - it will hit you anyway.

How and where do you start your AI integration? You will already have a good idea from the above. But there’s more.

• The organizational strategist will say: AI is so transformative that you have to touch everything in parallel: technology base, legacy systems, data, capabilities, governance. Build something future-proof fundamentally new.
• The practical one: Better to start somewhere than not at all. Don’t boil the ocean. It’s a moving target anyway.
• And you will find all kinds of implementation approaches: horizontal through the organization, vertical along your production chain, management-centric, product-centric. You name it.

And all have a point. The only thing you should avoid is buying a bunch of software products without a clear game plan. The important transformation is in your organization, not on your servers.

This is where negotiations re-enter the picture. Negotiations sit exactly at the intersection of strategy and execution. They turn priorities into trade-offs, uncertainty into decisions, and strategy into outcomes. That makes them one of the first places where unclear strategy, weak governance, or poorly integrated AI becomes visible - and costly.

If AI changes how strategies are developed, executed, and competed, negotiations must change with it - in preparation, governance, and decision discipline.

*) recommended further reading on the topic of AI strategy: Rebecca Homkes, Oct. 2025, “Want to lead with AI? Drop your AI Strategy and focus on these four planks instead”

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.

Once strategy is clear, execution has to start somewhere. AI is not your strategy but a tool. Here we outline a practical entry point — and why negotiations are often a smart place to begin.

Why? Because negotiations sit at the intersection of strategy, execution, risk, and human judgment. They surface priorities, expose trade-offs, and make weak assumptions visible - fast. They also usually have a defined scope, lifetime and involve a broad cross section of stakeholders across the organization. A good place to get started.

A pragmatic starting point in negotiations looks like this:
1. Make preparation explicit. Clarify what “good preparation” actually means. What decisions must be made before entering the room? What information really matters? Where do mandates, limits, and escalation paths come from?
2. Define the process first. Use a shared structure so preparation, roles, and follow-up don’t depend on individual style. Consistency creates the guardrails AI needs.
3. Add AI where it reduces friction - not responsibility. Let AI help structure information, compare options, draft scenarios, or support reflection. Keep judgment, trade-offs, and accountability firmly human.
4. Measure what changes. Shorter preparation time? Less rework? Fewer escalations? Better decision clarity? Simple signals are enough at the start.

AI won’t replace negotiators. But it will raise the bar for preparation, alignment, and decision discipline. And that’s not a technology question - it’s an organizational one.

This article was originally co-published with C-TO-BE. Slightly modified from the original version to fit the format.