Grove
Grove rebuilt IXACT's sea of overdue CRM tasks into a goal-based cadence system, grounded in 10 customer interviews, a heuristic CRM evaluation, behavioral data, and two independent AI synthesis passes.
- Project Type
- Platform CRM
- Year
- 2026
- Role
- Sr UX/UI Designer
- Deliverables
- Research, Product Strategy, Product Design
IXACT created a reminder for every touch it expected. At real scale, those reminders piled up faster than anyone could clear them.
Grove couldn't just inherit the CRM. The task list wasn't failing because it was badly built — it was failing because it kept an honest count of everything an agent hadn't done.
Every overdue row is a promise the system made on the agent's behalf — and then kept score of.
Open tasks on a single migrated account, most of them past due on arrival.
Ten agents, one hour each, one question.
We ran a heuristic evaluation against modern CRM patterns, then interviewed customers for an hour each — probing for where the decision to contact someone actually comes from.
The tech-averse late-adopter segment needed fewer alarms and a clearer way back on track.
“Walk me through the last time your CRM helped you decide who to contact next.”
“I want to untry them.”
Renee, customer interview
“I open it and immediately feel like I'm behind.”
Participant 04, session excerpt
“The red makes it feel like I already failed before I start.”
Participant 07, session excerpt
It wasn't a task list problem. It was a guilt problem.
The complaint was never the interface — it was opening the app already behind. Agents weren't avoiding the work. They were avoiding the scoreboard.
The math behind the sea of red.
Using Claude to query the schema and cross-reference PostHog data, we modeled exactly where task creation outpaces what any agent can complete.
Four inputs were enough to show the backlog wasn't a discipline problem — it was arithmetic.
Their coach already set the cadence. We just weren't building for it.
Most agents don't invent their outreach schedule — their coaching program already tells them who to contact and how often.
A, B, C categories are how agents already think about cadence — the CRM just never asked.
“A, B, C categories are how agents already think about cadence.”
Marty, customer interview
From “overdue” to “suggested”: changing what the system asks for.
Even behind pace, the goal view shows a way back on track instead of a bigger pile of red.
Before: overdue framing. After: suggested-outreach framing.
We replaced individual tasks with goals — tracked against a cadence, not a checklist.
Built with the room, not just in Figma.
Engineering and product brought the real constraints — migration scripts and coaching-cadence rules — and modeled that logic before any UI existed, so it had to work the moment a migrated customer's already-overdue data landed.
Redacted excerpts from the internal handoff that set the constraints.
[redacted] migrated IXACT records arrive already behind cadence.
[redacted] cadence engine must resolve coaching rules before UI render.
[redacted] goal view cannot depend on manual cleanup after import.
Two AI passes, one set of transcripts.
We ran a second, independent AI synthesis over the same transcripts to pressure-test our own findings — and kept the places the two passes disagreed.
Task overload was the symptom; cadence confidence was the product problem.
Task overload was the symptom; cadence confidence was the product problem.
Overdue framing suppressed action.
How much automation users would trust at first launch.
One goal, four doors in.
Migrated data, plain-language templates, full advanced settings, or a chat with AI — whichever way an agent already thinks about their pipeline.
The goal view.
What we're watching
- Pre-launch
- Grove hasn't shipped. Rather than claim outcomes, here's what we modeled, what we tested, and the leading indicators we're tracking once migrated customer data meets the live goal view.
- 10Customer interviews, one hour each
- ~100Users in the behavioral segmentation
- 2Independent AI synthesis passes
- 4Ways into a single goal
Usability test planned pre-launch.
The claim stays deliberately narrow until migrated customer data meets the live goal view.