The lost art of the proposition

You know the meeting. The slides are tidy. The strategy makes sense in the room. Everyone leaves nodding. A week later, three teams are heading in three different directions, each one convinced they're delivering on the plan.

Here's what actually happened. Someone said "we need to be more customer centric." Heads nodded. The marketing lead heard "better personalisation." The product lead heard "cut features people don't use." The COO heard "faster support." All three left aligned on a phrase. None of them aligned on a claim.

"More customer centric" is a topic. A proposition would be: "we keep more customers when we resolve issues on first contact than when we ship new features." Now there's a stake in the ground. Something to agree with, push back on or sharpen.

What was missing in that meeting wasn't alignment. It was a proposition.

What is a proposition?

A proposition is a load-bearing claim. It's a specific, testable statement about how the world works, attached to a goal it serves, that your decisions can actually rest on.

Think of it as a stake in the ground: this is where we are, and this is what we're heading toward. The goal is built into the claim. The stake might be in the wrong spot, and that's the point. You can only know it's wrong because you committed to a position.

Three things make a proposition a proposition:

1. A claim about what's true.
2. A goal the claim is in the service of.
3. The signals that would confirm or refute it.

Watch the escalation: "AI" is a thing. "AI in customer service" is a topic. "We should use AI in customer service" is an opinion. "AI will lift our net promoter score (NPS)" is a directional claim. None of those tell you what to do on Monday morning.

Now compare an actual proposition: customers rate us higher when AI handles routine questions and humans handle complex ones, because well-routed conversations cut wait time without sacrificing care quality.

There's a claim, the routing argument. There's a goal, NPS. There are signals to look for, like wait time, classification accuracy and post-resolution satisfaction.

You can build a flow around it. You can score new ideas against it. You can be wrong about it, and learn something specific when you are.

That last part matters most. You have to commit to a position before you can find out if it's wrong. That's exactly what topics and themes refuse to do. They dodge accountability by staying vague.

Propositions stay accountable by being precise. Philosophers built modern logic on that distinction. Researchers use it to bind theory to evidence. Most businesses, working in shorthand, lost it somewhere along the way.

In Decidr, every workflow starts with one. A proposition is the bridge between intent and decision: this is what we believe, this is what it serves, this is what would change our mind. From there, questions, scores, weights and decisions follow.

Why AI makes this urgent

AI didn't create the problem of vague organisational thinking. It made it expensive.

Models don't absorb ambiguity the way people do. A senior leader running on pattern recognition knows when the situation doesn't fit the template, when to override, when to slow down, when something feels off.

That judgement is the error-correction mechanism. AI has no equivalent, unless you build one in.

What AI needs to reason well is a proposition: a specific claim, a goal it serves, signals that would confirm or refute it. Without that structure, a model has nothing to check itself against. It fills the gap with confidence.

Enterprise hallucination rates still sit between 15 and 52%, not because the models are broken, but because vague inputs produce outputs that have no way of being wrong.

Hand a model a proposition and the dynamic changes. It has a stake in the ground. It can reason toward a specific goal, flag when something doesn't fit and produce output you can actually evaluate.

The proposition is the parameter. Without it, you're not directing intelligence, you're just prompting it and hoping.

What changes when propositions become the unit of work

Three shifts happen, and they're the kind you feel in your week.

Tacit judgement becomes explicit flows. The thing your best operator just "knows" is, in fact, a proposition they've never written down. Once it's written, it's a workflow. A new hire can run it. An agentic app can run it. The instinct stops dying when the operator goes on holiday.

Vibes become scores. "This idea feels right" becomes "this idea scores 7.4 against the proposition that customers prefer proactive service over speed, with three of four supporting signals present." It reads clinical until you realise you've replaced four meetings with one number you can argue about productively.

"I did my best" becomes an inspectable method. Every decision leaves a trail. Which proposition was it testing? What scored highest? Why? You can audit it, improve it, hand it to a successor or hand it to an AI. A method beats a memory every time.

What it actually feels like to work this way

The first proposition you write feels almost embarrassingly obvious. You think, surely we don't need to spell this out. Then you watch your team disagree about it in a meeting and realise nobody had actually agreed on it. That's the first jolt.

Other moments stack up. You put a number against a thing you used to vibe-check, and feel slightly silly.

A clean proposition kills a clever idea by scoring it badly, and you mourn the idea for an hour before noticing the team didn't have to argue about it. A discussion that usually takes a meeting becomes a five-minute review.

Then it flips. The next decision compounds on the last one instead of starting from scratch. Nobody re-litigates the same instinct in three different rooms. When something doesn't work, you have a claim to revise rather than a vibe to defend.

A new hire reads last quarter's propositions and gets up to speed in days, not months. The AI you bought stops being a smart-sounding pen pal and starts being something with a job. The institutional knowledge you've been worrying about lives in the system, not in three people's heads.

How Decidr puts propositions at the centre

This is the methodology we encoded in DecidrOS. Propositions, weights, scores, decisions and workflows, all linked, all visible, all upgradable. A live system that lets your business reason out loud, where humans and agentic apps work from the same explicit claims.

Your best decisions aren't accidents. They're propositions you haven't written down yet. Write them, and your business can finally run them.