Designing an intelligence layer for legacy B2B workflows, without disrupting the trust professionals placed in their tools.
Outcomes
- Launched across EverCommerce's core customer-insights products with strong adoption.
- Cut customer response times by integrating naturally into existing professional workflows.
- Drove measurable business impact across both new revenue and retained accounts.
- Set the design language, interaction patterns, and component foundation for every AI feature across the EverPro suite.
The Challenge
EverCommerce's customer-facing products store rich data (survey responses, customer sentiment, interaction history). However, professionals relied on basic keyword search to scour feedback, composed responses from scratch, and identified at-risk customers by subjective judgement alone.
The opportunity wasn't to bolt on a single AI feature. It was to introduce an intelligence layer that could permeate the entire platform, without disrupting existing client workflows. My first move was to audit every unique context across the product suite where AI assistance needed to live, finding interaction patterns flexible enough to span all of them.
The Intelligence Layer
Before a professional can act on an at-risk customer, they need to know who that customer is and what went wrong. AI Insights solves this upstream problem, monitoring survey feedback in real time, flagging dissatisfied customers and generating AI summaries of the sentiment behind each one. Recurring themes are ranked and surfaced across teams and locations, so patterns can be caught at scale rather than discovered case by case.
The Action Layer
Knowing a customer is at risk matters, however, users want to be able to address their at-risk customers in a timely fashion. From here, we set out to build the Assist AI customer response feature, designed responsively to fit within multiple interactions across the app in efforts to respond and resolve a customer issue promptly.
The Subtle-First Principle
The core design insight was placement. Early exploration (early 2024) surfaced a clear risk: AI features that interrupt existing workflows create friction and resistance, especially in products where users have deeply ingrained habits and initial skepticism to AI technology. The answer was to go subtle first.
Inspired by Superhuman and Apple Intelligence, Assist lives at the bottom of existing text inputs, with all shortcuts/editing tools provided. Never announcing itself, rather it is ready to help when/if a user needs it to serve an at-risk customer.
Scope Evolution
Responding to a survey is one moment in that recovery. But a contractor might also need to call the customer directly, loop in a team member via notes, send a follow-up survey to confirm the issue was resolved, and finally mark the relationship as recovered. Designing only the reply left the rest of that journey without a home.
That gap expanded the scope into a full resolution workflow.
Accessibility
The AI Insights dashboard originally used color alone to communicate customer risk levels — a pattern that creates real barriers for color-blind users making decisions that directly affect customer relationships and revenue.
Redesigning this meant replacing color-only indicators with universal iconography and WCAG AA color tokens, ensuring risk signals could be read by anyone regardless of how they perceive color. The result was cleaner for all users, not just those with accessibility needs.
Impact & Reflection
Assist launched across EverCommerce's core customer insights products to strong adoption. Response times dropped measurably as the toolbar integrated naturally into existing professional workflows.
The business impact was significant across both new revenue and retained accounts. But the longer-term outcome mattered more: Assist established the design language, interaction patterns, and component foundation for every AI feature that followed across the EverPro suite.