Skip to content

Agentic systems

Agentic AI vs automation

The short answer

Use deterministic automation when the workflow and correct behaviour can be specified explicitly. Use an agent when the system genuinely needs to interpret changing context, choose among tools and decide what to do next.

What deterministic automation does well

Deterministic systems are predictable, testable and inexpensive to operate. They are the right choice for validation, routing, calculations, permissions, policy rules and workflows where the possible states are known in advance.

What an agent adds

An agent adds discretionary planning. It can inspect context, choose a tool, observe the result and decide on the next step. That is useful when a rigid workflow would require an impractical number of branches or when language and ambiguous state are intrinsic to the task.

The cost of agentic behaviour

Flexibility introduces uncertainty. A production agent needs constrained tool access, explicit state, stopping conditions, recovery behaviour, task-level evaluation and observability. Without those components, autonomy becomes an uncontrolled source of product risk.

The usual answer is hybrid

Most useful systems combine both. Deterministic software controls permissions, state transitions, calculations and safety boundaries. The model interprets language, synthesises information or chooses among allowed actions. This separation makes the system easier to reason about and evaluate.

Build an AI system with explicit boundaries

Build what matters.

Tell us what you are trying to build, change or understand. A first conversation is about the problem, not the feature list.

Start a conversation