Dear Negotiation Explorer,

Welcome to issue 29 of our NegoAI series.

Last week, I shared the Competence Matrix and made a promise: this week would cover the single factor that made the biggest difference in my own AI workflow.

Context.

But not just "provide more background." Something bigger — a shift in how we think about working with AI entirely.

For two years, the AI world has been obsessed with "prompt engineering" — the art of writing the perfect question. Courses, certifications, even job titles have been built around it.

It matters. We covered it in issues 25 and 26. But here's what I've learned from teaching AI and negotiation courses: the prompt is maybe 20% of what determines output quality.

The other 80%? Everything you feed around the prompt.

Andrej Karpathy — former head of AI at Tesla and co-founder of OpenAI — put it directly: the real skill isn't prompt engineering. It's context engineering — the art of providing the right information, at the right time, in the right structure, so that AI can actually do useful work.

The difference:

Prompt engineering: "How do I ask this question better?"

Context engineering: "What does AI need to know before I ask?"

Think about it. You could write the perfect negotiation prompt — specific outcome, clear purpose, quality criteria, persona. Everything from issue 26. Feed it into a blank chat with no context? You get generic advice that could apply to any deal.

The same prompt, fed alongside your negotiation knowledge base, the counterparty's communication history, and lessons from your last three deals? You get something you can actually walk into a meeting with.

The prompt didn't change. The context did. And the output is unrecognizable.

Context engineering isn't one thing. It's four:

Most negotiators using AI work like this:

No knowledge base. No counterparty history. No memory. No instructions. Just a prompt into a void.

And every single session starts from zero.

It doesn't matter that you explained your negotiation style last Tuesday. It doesn't matter that you spent 20 minutes briefing AI on the counterparty yesterday. New chat, blank slate. All that work — gone. You do it again. And again. And again.

It's like hiring a brilliant consultant, giving them no briefing materials, no background on the counterparty, no context on the relationship — and then firing them at the end of every meeting and hiring a new one who knows nothing.

The consultant is smart. But without context, they'll give you generic advice. Every time.

Now imagine handing that same consultant:

Different conversation entirely.

That's the shift from prompt engineering to context engineering. Build the context once, and you never start from zero again.

Here's the practical part — and what I teach in my Maven courses. You don't need Claude Projects. You don't need agents. You don't need code.

You need a folder.

A document that captures your negotiation expertise. Not generic advice from the internet. Yours.

What goes in:

This is what I call the "Deepak Knowledge Base" — the one you received when you subscribed to this newsletter. It's the structured context that makes my AI negotiation agents work. Without it, the AI guesses. With it, the AI prepares like a senior advisor who thinks the way I do.

You need your own version. It doesn't have to be 50 pages. Start with one page: your top 10 negotiation principles and the frameworks you actually use.

This is where most people have nothing — and where the biggest gains are.

For each major counterparty (client, supplier, partner), you build a folder:

Meeting Notes & Transcripts — notes from past meetings, or AI transcripts if you use tools like Granola. The raw record.

Key Communications — important emails, messages, formal correspondence. Not everything — the ones that reveal positions, priorities, or pressure points.

Previous Deals & Outcomes — what was negotiated, what was agreed, how it went. The factual record.

This folder is your counterparty dossier. But raw data alone isn't enough — you need a way to distill it into intelligence. That's what we'll build next week: memory files that capture what you've learned, not just what happened.

For each live negotiation, a folder with two things:

Deal Brief.md — the structured summary of the current situation:

AI Outputs/ — what AI has generated for this deal. Analysis, scenarios, creative options. Save them — you'll iterate and want to reference earlier versions.

Your prompt library. The prompts we've built in this series:

You don't rewrite them each time — you pull them from the folder and adapt.

When you start an AI session for negotiation prep:

Three minutes of copy-paste / upload. The difference in output quality is enormous. And unlike starting from scratch, you're building on everything you've already captured — not repeating it.

Is it manual? Yes. That's the starting point. As you advance, you can move to Claude Projects, ChatGPT Projects, or Gemini Gems where files persist automatically, or even build AI agents that read from structured repositories. But the content is the same. The structure is the same. You just get more automation.

Start with the folder. The automation comes later.

Build your AI Negotiation System folder.

Step 1 (5 minutes): Create the folder structure. Google Drive, OneDrive, Dropbox, your desktop — wherever works.

Step 2 (10 minutes): Write your Knowledge Base. Start with one page: your top 10 negotiation principles and the 2-3 frameworks you actually use. It doesn't have to be perfect — you'll improve it over time.

Step 3 (5 minutes): Create a Deal Brief for your most pressing current negotiation. Fill in what you know.

Then try it: Run the metaprompt from issue 25 — but this time, paste in your Knowledge Base alongside the Deal Brief. Compare the output to what you got without context.

The difference is the lesson.

Next week: Teaching AI to remember.

You've built the system. Now we'll make it learn. I'll share the memory templates that capture what you've learned about counterparties and about yourself — and the deal debrief prompt that feeds every negotiation back into the system, so the next one starts smarter.

This newsletter gives you the concepts, one piece at a time. The course gives you everything — live practice, direct feedback, and the complete system built with you step by step.

If you've been following this series and want to go deeper, this is the next step.

Register here →