Most negotiators I work with today feel the same tension: AI is suddenly everywhere, but their preparation process still looks exactly the same. They are under pressure on margins and timelines, and they know they should prepare better—yet they’re not sure how to turn AI into something that truly improves the way they negotiate.
Last year, in a study I ran with senior executives, AI-assisted preparation already led to clearly better negotiation outcomes. Since then, working with AI agents and structured workflows, the results I’m seeing in practice are even stronger—especially in complex, low-trust negotiations.
In the last two newsletters, we looked first at how to prompt AI like an expert and then at how to assess the quality of its negotiation analysis. This time, I want to show how those pieces fit into a simple, 25-minute system you can actually use.
Below are three practical ideas I’ll also expand on in an upcoming 30-minute Lightning Session on Maven.
1. The real value of AI in negotiation prep (beyond the hype)
For most professionals, “using AI” in negotiation preparation means typing a vague prompt into ChatGPT or Copilot and hoping for insights.
That usually produces generic output and a fair amount of frustration.
What works much better in practice is a structured approach where AI is used to:
Make your thinking more complete AI is very good at forcing you to articulate what is often left fuzzy:
your interests and alternatives
the other side’s likely interests and constraints
possible trade-offs across issues
When you use a structured template and a well-designed prompt, AI becomes a thinking partner that helps you see the deal from multiple angles, not a quote machine.
2. Explore multiple “what-if” scenarios
In complex B2B or cross-border negotiations, we rarely have time to seriously explore more than a handful of options before we walk into the room.
AI can quickly generate different packages—from more distributive options to highly integrative structures—and stress-test them under different assumptions.
This doesn’t replace your judgment, but it widens the space of possibilities you can evaluate before making a choice. In practice, that often translates into better trades, fewer blind spots, and more creative agreements.
3. Reduce prep time without lowering quality
Many leaders tell me:
“I know I should prepare better—I just don’t have the time.”
With the right system in place, you can move from days of unstructured preparation to about 25 minutes of focused, expert-level prep.
The real benefit is not only speed; it is consistency. A well-designed AI assistant can help ensure that different people on your team prepare to a similar standard, using the same logic, instead of everyone reinventing the wheel before each negotiation.
But what about reliability? (The three big LLM problems)
In a recent newsletter, I highlighted three structural limitations of LLMs that are especially dangerous in negotiation:
Non-determinism – the same prompt can produce different answers, which is a problem when you need alignment and repeatability.
Hallucinations – the model can sound confident while making up facts, which can distort interest analysis, BATNA estimates, stakeholder mapping, or risk evaluation.
Opacity – LLMs don’t show how they reached a conclusion, so you don’t see the assumptions or trade-offs behind the output.
The answer is not to ignore these limitations, but to design around them.
LLMs like ChatGPT or Copilot can already be very helpful in preparing for a negotiation, especially if you use well-designed prompts. But if you want reliable and consistent outputs across people, teams, and deals, you need more than a single chat window. That’s where no-code, agent-based workflows in Cassidy come in: they turn your best preparation logic into a small system of AI agents that work together the same way, every time.
Structured preparation templates, clear prompting logic, simple quality checks, and these no-code workflows make AI much more reliable as an analytical partner. Part of what I’ll show in the Lightning Session is how to use these elements in a way that supports real negotiations, not theory.
A practical resource: my negotiation preparation sheet
To make this more concrete, I’ve created a Negotiation Preparation Sheet that I now use as the starting point for AI-assisted prep with executives.
It captures the key elements we’ve discussed—interests, alternatives, issues, scenarios—and is designed to plug directly into AI workflows.
You can access it here: <prep sheet link>
In the Lightning Session, I’ll show how this sheet can be turned into a 25-minute AI-supported workflow using tools like Copilot and a simple no-code, agent-based assistant.
A free 30-minute Lightning Session (if this resonates)
If you’re curious to see what this looks like in practice, I’m hosting a free 30-minute Lightning Session on Maven:
Win High-Stakes Negotiations with AI 📅 Date & time: November 20, 18:30–19:00 CET
In this short session, I’ll walk through:
A 25-minute negotiation prep workflow you can start applying immediately
A simple system blueprint for building your own no-code, agent-based AI negotiation assistant
The key guardrails for using AI in real negotiations (confidentiality, reliability, ethics)
Most people stop at “using ChatGPT with better prompts.” That’s useful, but it still produces one-off, hard-to-repeat outputs. In this session, I’ll show how to go one step further: from advanced prompting to no-code, agent-based workflows that make your negotiation preparation more reliable, consistent, and reusable.
If you’d like to join, you can register here: <Maven registration link>
My goal is that you leave with at least one concrete way to improve how you prepare your next high-stakes negotiation, and a clear sense of what becomes feasible when you plug that preparation into a no-code AI platform.