Your experts know it. Now encode it.

• Your models can't explain their own recommendations. "The model ranked it highest" is not a defensible answer to a regulator, a board, or a claimant. If you can't trace the recommendation back to a causal structure, you don't control the decision — you're ratifying it.
• Your analytics stop at description. Dashboards tell you what happened. If that's where your decision support ends, you're paying for historical records. The question that actually matters is what happens if you act — and that requires a different kind of model.
• The knowledge that makes your models right is sitting in someone's head. When that person leaves, the model doesn't know what they knew. That's not a talent risk. It's a structural one — and it compounds every year you don't fix it.
• You already know a wrong decision is expensive. Wrong pricing, wrong capital allocation, wrong intervention — each measurable in millions. The question isn't whether your current model can be wrong. It's whether it can be wrong in ways you'd never detect until after the loss.

Thirty minutes. No pitch, no slides. Bring what's keeping you up at night — the decision where you suspect the current approach is wrong but can't prove it. I'll sketch the causal structure on the spot: which variables drive which outcomes, where the confounders are, what the model needs to answer. By the end you'll know whether causal reasoning changes the answer. If your existing tools are sufficient, I'll tell you — and you'll have saved the engagement.

A causal model built on the wrong question is worse than no model. The engagement doesn't start until we've both agreed the problem is real and the approach is right.