AI-enabled Automation Workflows

Overview

Sprout customers rely on Rules to categorize and route incoming social media messages to the right agents so they can be resolved. Rules were simple if/then conditions; for example, if a message contains the word “refund,” then route to the customer support team.

We have clearly outgrown the Rules we have today.

To support sophisticated workflows and AI-driven needs, we must move away from today’s fragmented experience, toward a new scalable automation engine where deterministic logic and agentic intelligence coexist, making endless workflows are possible.

The problem

Legacy rules are unable to support more complex logic, such as an “else” condition, so customers had to create a separate rule for every possible path. Some enterprise customers ended up with hundreds of rules, even as many as 800. Not only was that tedious to set up and maintain, but our internal support and services teams had to help troubleshoot all that complexity as well.

While powerful in functionality, legacy rules unfortunately:

• Only trigger on message ingest—no other trigger options
• Have no visibility into rule conflicts or overlaps
• Lack conditional logic (if-then-else structures)
• Have no time-based configurations or overall orchestration capabilities
• Have no AI capabilities to help customers do more

Approach

This was a long-term, multi-quarter project, but starting design work in advance will position us to move quickly when roadmap capacity opens. As a design-driven initiative, this project it to be done in the following phases:

• Discovery & Audit – Research customer challenges, audit current functionality, and gather insights from internal stakeholders across the org
• Synthesis – Analyze and create artifacts to help understand and internalize the research, come out with a well-informed POV 
• Cross-Org Collaboration – As Rules were used throughout the app in various other functions, leading a working group with members from other teams is crucial for transparency, awareness, and feedback.
• Design, Test, and Iterate – Leverage Claude Design to rapidly develop designs and conduct ongoing user testing to ensure that we remain aligned to our customer needs and expectations.

Solution

Here, less is more– in some cases, customers could consolidate what had been hundreds of individual rules into four or five workflows. It was a big simplification for the customers managing these automations and for the internal teams supporting them.

Branching: Allows users to create multiple paths in one view

Selection Indicator: Easily lets users know which nodes they’ve filled out and which have empty values

Draft Mode: Indicator at the top shows users if they have unpublished changes

Node States

Workflow attribution in the app

Takeaways

• Start from the pain customers already live with. Years on the Inbox team meant I knew how people actually built rules, so the redesign started from real behavior, not assumptions.
• This one took took a bit of convincing to get onto the roadmap. Enterprise customers who had hundreds of rules often also used Sprout Professional Services to set up. Meaning, internal Sprout teammates were spending countless hours painfully doing setup and maintenance, as well as diagnostics when something went wrong. However, this also meant that a lot of the pain was hidden to these customers– good and bad thing. Customers don’t feel 100% of the pain, but also it’s highly inefficient on the backend using human resources, not the way good software should be built.
• Simplify the fundamental structure, not just the UI. There were lots of small papercuts that could definitely improve the UX; things like autosave, draft mode, searchability within menus, etc. However, without a fundamental change in what users can create, we’re only saving seconds vs saving hours of time.
• Users often found it difficult to construct workflows from scratch, so templates or AI Assist are a huge help.