Quick question for the Pega community
While working with Pega Platform GenAI features - Agents, Tools, Coach, and Connectors
Can someone help clearly differentiate what to use when?
Quick question for the Pega community
While working with Pega Platform GenAI features - Agents, Tools, Coach, and Connectors
Can someone help clearly differentiate what to use when?
This is a great question, and it comes up frequently as teams move beyond experimentation.
A reliable way to differentiate is to focus on scope, control, and lifecycle, not just capability:
GenAI Connect is best for single, bounded tasks where you want one prompt, one response, and predictable mapping to case data
GenAI Agents are appropriate when reasoning spans multiple steps, requires context over time, or needs to invoke tools conditionally
GenAI Coach is designed for in‑context user assistance, not autonomous action
Tools define what actions are allowed, not when or why they happen
The most important design principle is that agents should not replace workflows. Agents reason and propose; workflows decide, enforce policy, and remain the system of record.
If others are willing to share examples, it would be helpful to hear which signals caused them to move from Connect to an Agent, or from Agent to Coach, in real projects.
I think a key consideration for Coach is to think of it as a way to embed best-practices guidance into the workflow in a way that’s “detail-aware.” Coach can use knowledge sources and the workflow itself to provide specific guidance on the case that helps drive the right user behavior or guide the user to the best-case actions to use. Most of the builders I work with that have adopted Coach look at it as way to support multi-variable user assignments, where there may be many pieces of information or context to consider when determining how to complete an assignment. A real-world example: one of our customers saw a user assignment where the users had a large “checklist” of various items to consider against their knowledge before they selected one of several paths to send a piece of work down. Some of that we built into workflow logic, but Coach also helped by providing contextual guidance, helping newer users become more proficient in evaluating what is important or critical to completing their assignment when the workflow is relying on their knowledge instead of business rules.