First, I want to recognize that Autonomous Agents (AA) are still early in the product’s lifecycle. We don’t have it all figured out, but we think we have some good ideas about both based on what we’ve learned from customers and partners in the past ~9 months. Below is an attempt at a concise summary. We continue to learn and evolve our point of view and welcome your feedback.
1. When a Flow is Enough
Straightforward Processes: Routine approval workflows or document routing with minimal edge cases.
Low Variability: Frequent, repeatable steps—like triggering an email when a file is added to SharePoint.
Limited External Inputs: You control all the systems, and real-time user interaction is minimal.
2. When an Autonomous Agent Excels
Complex stakeholder interactions inside the process E.g., support routing, where user requests can evolve in unexpected ways, requiring on-the-fly classification and rerouting. Another example is order fulfillment with multiple external partners: Real-time stock updates, shipping disruptions, or partner errors can trigger new actions that wouldn’t be spelled out in a standard flow.
High Uncertainty: Loan applications or compliance reviews that depend on multiple data sources or changing regulations, where you can’t feasibly map every scenario into a static flow.
3. When is a blended approach appropriate
In cases in which the overall process might be different every time and/or the maker prefers to provide instructions in natural language, but some parts of the process need to perform in prescribed way (the prescribed parts should be handled via a flow).
Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.
Share
Report
All responses (
Answers (
265
