House Rules

There’s a rule in our house version of Monopoly called AI Abuse. We play the PlayStation edition most weekends, two humans and two AI players set to hard, and the rule usually gets called out or debated at least once per game. If an AI completes a colour set on its own, it keeps the set. Neither of us humans is allowed to trade it away from them. We never sat down and drafted anything official. It just emerged, the way rules do, after one of us noticed that a robot player who can’t tell what a completed monopoly is worth will hand it over for next to nothing, and a game where you can do that isn’t much fun in the end.

The rest of the rulebook developed organically like that. The rolling order puts an AI between us in both directions, so neither of us gets first crack at whatever the machines land on. Neither of us can start buying into a colour set from an AI until two of its properties are already out on the board, which keeps the early game honest. And we don’t trade with each other at all. We’re too cutthroat for that. Each of us works with what we land on and whatever we can prise out of our particular AI Buddy.

It took me a while to see what we’d actually built. Nobody usually writes unique house rules about dice rolls. The dice can’t be exploited, so they don’t need governing. Every rule we’ve invented is about the AIs, and each one exists because the AIs can be exploited. Without the rules, the weekend would turn into a race to see who could strip the machines fastest, and the contest that makes the game worth playing, the one between the two humans, would be over before it started. We unintentionally built a regulatory framework for our AI players, because it became obvious the game needed one.

Now, after all of the rule crafting and debates, it’s very clear that the “robots” still earn their place on the couch. They keep the game moving. They buy what they land on and hold it without much of a plan, and a lot of our sets get finished with something bought out from under an AI at a key moment. The railroads mostly live with them as trading stock. There’s a shared understanding that the oranges are the properties most worth fighting for, and more than one game has swung on which of us reached an AI holding the third one first. The machines create the liquidity, and deciding what to do with it is still wholly our job.

The funny thing is that this concept also comes up in conversations I have with “real world” organisations deploying AI agents. The agents always add capacity, the same way our AI Monopoly players keep property flowing around the board. What most organisations haven’t done yet is write their house rules. Very few have named their version of AI abuse, or worked out what stays off the table when the machine side of the business offers a tempting price. The contest is still between humans (your organisation and the ones it competes with), and the agents matter to the extent that they free your people to play that game well.

My regular weekend human opponent likes to experiment with new strategies. I run more or less the same game every week, and most weeks that settles it, not to brag. The AIs, set to hard, make no real mistakes, but most games they finish holding just the utilities. Playing without mistakes, playing boldly, and playing well turn out to be very different things, and only one of them comes with a difficulty setting.