About

Organisations are pointing AI at their data and finding the answers harder to trust than expected. Often the cause has built up over years: reporting that grew team by team, with definitions that made sense locally but were never brought together.

People have always worked around this because they knew which figure to use. AI doesn’t have that context. It reads the semantic model as it is, so inconsistent definitions lead to unreliable answers, rising token costs and harder governance.

Henri and Mehdy explain why the semantic model and business definitions are now the most important assets for AI accuracy. They cover how to design semantic models that both business users and AI can interpret, what your definitions need so AI reads them as intended, and practical steps to simplify your semantic layer so everyone can rely on the answers.

Takeaways

• How to design semantic models that both business users and AI can interpret correctly
• Where semantic model complexity drives up token and compute costs
• What your definitions need so AI reads them the way your business intends
• Four practical steps to simplify your semantic layer