LEVEL 01 · LESSON 06
Deterministic vs probabilistic systems
Let models interpret ambiguity and let explicit rules enforce hard boundaries.
What it is
A deterministic rule produces the same result from the same inputs under the same conditions. A probabilistic model works with likelihoods, and generation may also involve sampling from those likelihoods.
Why it matters
Business systems need both flexible interpretation and dependable enforcement. Mixing up these roles creates avoidable risk.
A business example
AI interprets a complaint and proposes a refund. Ordinary application logic checks whether the order exists, whether a refund was already issued, and whether the amount exceeds an approval threshold.
How it actually works
Use models where language or ambiguity makes fixed rules difficult. Use explicit code for calculations, permissions, limits, and transaction checks. A model’s generated proposal becomes an input to those checks, not a replacement for them. Even if generation is configured to be repeatable, an incorrect proposal remains incorrect. Evaluate the model’s quality and test the surrounding rules separately.
A common misconception
Probabilistic does not mean arbitrary, and deterministic does not mean correct. A repeatable rule can encode the wrong business policy.
ONE SENTENCE TO REMEMBER
Let models interpret ambiguity and let explicit rules enforce hard boundaries.
CHECK YOUR UNDERSTANDING
An AI assistant drafts customer refunds. Where should the rule “refunds above €500 need manager approval” be enforced?
Show the answer
Answer: B.
Let the model interpret the request and draft a proposal, and let explicit rules enforce hard limits such as amounts and permissions. Probabilistic does not mean arbitrary, but repeatable generation still does not guarantee that a proposal respects policy.
Take it into your next meeting
Which decisions in your process must be enforced regardless of what a model suggests?
One more concept, understood.
Mark this lesson complete to keep your place.