Skip to main content
An agent becomes more useful when it can do more than speak. In practice, that usually means two things:
  • it can reference the right business information
  • it can take the right business actions

Actions

Actions let an agent make something happen beyond the call itself. Examples:
  • create or update a customer record
  • check an opening or availability
  • send a follow-up message
  • hand the call to a person
The most important rule is scope. Give an agent only the actions it needs for its job.

Knowledge

Knowledge gives the agent context it can use during a conversation. This might include:
  • service information
  • business policies
  • office hours
  • location details
  • commonly asked questions
Knowledge should help the agent answer consistently without forcing your team to place every answer in the prompt itself.

Keep the two separate

It helps to think of these as two different roles:
  • knowledge helps the agent explain
  • actions help the agent complete work
An agent may need one, the other, or both.

Start small

For early rollouts:
  • begin with a small action set
  • include only the most useful knowledge
  • review calls to see what the agent actually needed
This usually leads to a better result than loading the agent with every possible integration on day one.

Common failure pattern

A weak setup often looks like this:
  • too much information
  • too many actions
  • unclear rules about when to use them
When that happens, the fix is usually not more complexity. It is tighter scope.