I've been trying to work out for a long time why negotiation is the leadership skill people describe best and practise least.

Not because they don't know how. Most senior people I meet could teach a decent session on it.

Because the training is usually fine. The frameworks are sound - interests over positions, know your walk-away, all of that - and the people teaching them generally know what they're doing.

It's just that the real negotiations in a leader's week don't announce themselves. They happen in corridors, in budget meetings, in a quiet "can you help me with something", and nobody calls them a negotiation at all.

So people walk in with a framework and walk straight past the moment to use it. Quietly. Most weeks.

Welcome to Zombiepuram is the game at the heart of AI-augmented Negotiation and Influence, my training on negotiation in leadership. It's one of the 7 games.

How it works: a six-hour day, play first, then the EPPA loop

The programme runs on one loop. I call it EPPA - Experience, Patterns, Principles, Application.

The Experience is the game: three to four hours of play inside a six-hour day. The arc is the product, so I don't cut it.

Then we unpack it. Patterns first, which is what kept repeating across the factions. Then Principles, which are built to come from the players, not from me at a flipchart.

Then Application, where each leader takes it back to their own week.

That is the only place AI turns up. After the play, never inside it.

Nobody at the table has a laptop open. The game is people, a map and a lot of talking, and it stays that way.

The game at the heart of it: Welcome to Zombiepuram

This is my flagship, and the biggest thing I make. A whole room, all at once, split into factions trying to survive in a Bangalore that has gone very badly wrong.

Every faction has PUBLIC goals its members can see. Every single member also carries a PRIVATE goal nobody else knows about.

There is a big hex map with resources worth controlling, trading and occasionally fighting over. My team runs the zombies and the world.

And that is it, mechanically. It is built to produce three to four hours of dense political work - reading what somebody actually wants against what they said they wanted, deciding what to reveal, deciding who has earned trust.

Which is your Tuesday, with better lighting.

How the Play Is Built

Early on, factions huddle, pick a strategy and go make first contact. The game is built so the opening moves feel polite and slightly tentative, because everybody is working out who to trust and how much to give away.

By the second hour, alliances are meant to be forming. Some announced. Some very much not.

Every faction knows something the others need. Deciding when to share it, with whom, and for what in return is where the leadership actually lives.

The fourth hour is the one I designed most carefully. It is built so alliances strain, and hour-one decisions come to collect.

A faction that burned a bridge to hit its own target will find it needs that bridge. And the quiet person who kept everybody talking, nowhere near the top on points, can end up the most powerful player in the game.

Nobody wins the last phase alone. The design rewards groups who built real trust, and it is built to make the groups who played every conversation as a transaction spend the final stretch scrambling to catch up.

Not one person can win by being senior, because in Zombiepuram nobody is.

What the play is built to reveal

Leadership training has a measurement problem, mine very much included. Most of it measures how well people describe good behaviour. This game gives a group three to four hours of play, incomplete information and competing objectives. It is designed so people stop describing and start behaving.

• Who makes first contact across factions - and whether they open with an offer or a demand.
• Who gives information away to build trust, and who sits on it waiting for a better price.
• Who becomes the broker, holding good relationships with three factions at once without belonging to any of them.
• Whose promises hold when keeping them gets expensive, and who switches sides the moment the wind does.
• How open the politics get. A group may play everything in the open, or run a whole second conversation on the side.
• Whether the group can still act together once everyone's private goal is safe. This one is built to be the bruise.

Where AI comes in: the Application step

By Application, the debrief is built so the players have already named what they saw. The coalitions, the broken promises, the broker nobody expected. AI has nothing to do with any of that, which I think is the right way round.

Then each leader brings one real negotiation that is coming up for them. A budget, a headcount ask, a peer they need something from. Here is what happens next:

• AI plays the other side. Their public goal, their likely private goal, and their next three moves. It is Zombiepuram's public-and-private structure, pointed at a real person this time.
• The leader tests what to show and what to hold. Out loud, a few times, before it costs anything.
• They write the opening offer. An actual first line they can use, for an actual meeting, with a date.

What stays human is the part the whole game is about. What to reveal, and to whom.

AI can guess at a private goal. It can't decide how much trust a colleague has earned. That one belongs to the leader, and honestly, I think it should stay that way.

Come and have a look at the calendar

What your leaders walk out with

• One real negotiation, rehearsed, against an opponent who pushed back with three moves they hadn't planned for.
• An opening offer, already written, so the first sentence in the room is a decision and not an improvisation.
• A clear line between what they will show and what they will hold, worked out calmly rather than at the table.
• A shared language for the politics - factions, private goals, the broker - that lets a team talk about it without accusing anybody of anything.

The receipts

I've been training founders and leaders since 2019, with founders in the US, Canada, Australia, Singapore and India.

One programme at a Fortune 100 company scored 8.82/10 as a satisfaction score, across 113 respondents and six session-days. That is one programme, and I'd rather not stretch it.

Across every scored session-day I have, the satisfaction score is 8.40/10 across 244 respondents. A bit lower, and probably the fairer number to judge me on.

To be straight with you: both of those scores come from other programmes. Welcome to Zombiepuram has been played with a client team at an HR software company, but it has no survey score of its own.

Come and have a look at the calendar

Who it fits

• Mid-to-senior leaders already living in the soup - competing priorities, cross-functional dependencies, and a political weather system nobody put in writing.
• Senior teams where the politics are quietly hurting decision quality. The game is built to say the thing out loud, so you don't have to.
• Post-merger groups still politely operating as separate camps. The faction structure is built to mirror it.
• Organisations that keep having the alignment conversation. Here they get to feel what misalignment costs, in one day, with nobody's actual budget on the line.
• A diagnostic before you design something bigger. Three to four hours of real behaviour may tell you more than most surveys.

Who would get more from something else

• Groups under 15. You need enough people for the interests to genuinely compete.
• Frontline or junior teams. This is built for people who will recognise the dynamics, not meet them for the first time. Bloom is probably the kinder starting point.
• Anyone who needs it in under six hours. The day is six hours, and three to four of that is play. Cut the day to three and the play gets cut just as it builds. I'd rather suggest a different game than run half of this one.
• Settling one specific fight between two people. A game is the wrong tool for that, and I'll say so on the call.

Fair questions

"Why six hours? Can it be shorter?"
That's a fair ask, and honestly it is the one I get most. The six hours is the whole day: the teach, three to four hours of play, breaks and the debrief. The arc of the play is the product. Alliances forming and breaking, trust getting built and spent, private goals colliding with the shared one - that needs time. And the AI-augmented Application step sits inside that day, in the debrief, after the play. We plan it around your group, and I'll tell you plainly what I think it needs.

"What if real political tensions come up?"
They may, a little. The debrief is built for that. I separate what the game surfaced - patterns, tendencies - from what your office holds, which is history and specific people. Making a pattern discussable without making it personal is delicate work, and it needs the person who designed the thing to run it. That's me.

"Is it right for a team already in conflict?"
Depends what you want. If you want shared language for the friction, it's lovely for that, because the dynamic moves into the game world where nobody is being accused of anything.

"Can AI really help with a negotiation?"
With the preparation, I think it can. A sparring partner that will argue the other side for as long as you like is useful, and most of us never get one. It doesn't make the call on what to reveal. Which tools your leaders can use for a live negotiation, and where that information goes, is a fair question too - I'd rather answer that properly for your setup on a call than guess at it here.

Worst case? We have a conversation and you come out of it with a clearer brief, whether or not Zombiepuram is the answer.