Using rule-based AI agents
When a rule-based AI agent fits
Using and combining smart and rule-based AI agents correctly
Tip
A rule-based AI agent follows a fixed, programmed flow: input parameters, business logic and output handling. It is the right choice when a process must run exactly and verifiably. For everything else, a smart AI agent is simpler.
Smart or rule-based?
| Question | Rather smart | Rather rule-based |
|---|---|---|
| Can the flow be described in natural language? | Yes | Only with many exceptions |
| Must values be calculated or checked according to fixed rules, e.g. a minimum advance payment? | No | Yes |
| Must every decision be reproducible? | Not necessarily | Yes |
| Who sets it up? | Specialist department, without programming | With technical knowledge, via code |
| Should it conduct the conversation in chat and phone? | Yes | No – it is used as a tool by a smart AI agent |
| Should it create a resolution plan? | Yes | No – its action becomes a step in the plan |
Combining both
Rule-based and smart AI agents are not mutually exclusive. A smart AI agent leads the request – reads the message, talks to the customer, writes the reply – and calls a rule-based AI agent with a fixed flow as a tool. How this works in the resolution plan, in chat and on the phone: Rule-based AI agents as tools.
Read more
- Custom rule-based logic – set up a rule-based AI agent
- Examples – bank data, advance payment and meter reading agents