Reducing churn by 50% by redesigning an AI platform's onboarding flow.
Voyc uses AI to monitor every call-centre conversation for compliance breaches, a product sold to regulated financial services and insurance firms that pays for itself the moment it catches a single mis-selling event. But new clients were stuck in onboarding for up to six months before they could run a single script check, and Customer Success was drowning in repetitive setup calls.
As the first product designer at Voyc, I led the end-to-end redesign of onboarding, from user research and journey mapping through to a shipped in-app self-service flow.
"We had customers on the verge of cancellation because they hadn't seen value yet. Every new client was burning Customer Success hours we didn't have."
Internal, Voyc Customer SuccessFrom hours to minutes: transforming onboarding for rapid value.
Voyc's users faced a cumbersome onboarding process that required up to five 45-minute calls and an additional five hours of manual support, delaying their ability to see the product's value for six months.
This inefficient process led to user frustration, increased churn, and overwhelmed our customer success team. We had key clients on the brink of cancellation. The risk was real and immediate.
The project aimed to make onboarding faster and lighter by introducing a self-service flow that would reduce support interactions by at least 50%, shorten time-to-value, and boost user retention.
Three problems were holding onboarding up
Manual onboarding bottlenecks
Setup was call-dependent and slow. Customers could not move forward without Customer Success, and we could not scale.
Insufficient in-app guidance
No help content in the product. Hitting confusion left one option: open a support ticket.
Limited onboarding visibility
With no overview of pending tasks, users could not tell if they were doing it right, or nearly done.
Six research moves to pinpoint the friction.
Six research activities: onboarding sessions, user interviews, support ticket analysis, stakeholder workshops, affinity mapping and a prioritisation session. What surfaced was a fragmented manual process that delayed value and burnt out the support team.
What I learned, from whom
"I waited two months for my first useful insight from the platform."
"Every time I got stuck I had to email support. It killed momentum."
"My week is half setup calls. There's no time for strategic work."
"We almost cancelled. We weren't getting any value out of what we'd bought."
Visual mapping of user pain points
I clustered the findings into five themes, which made the patterns visible to the rest of the team.
A snapshot from our FigJam workshop. Sticky notes clustered user feedback to pinpoint the key friction points in onboarding.
Validating the new flow with iterative service-level testing.
I drafted a self-service flow around how customers actually thought about getting started, then tested it in live onboarding sessions with the Learning Experience team before any of it was built.
Flow chart validation: three iterations of the onboarding flow, refined live during sessions with the Learning Experience team.
Tested as a service first
Real customers walked the flow in guided sessions while it was still a flowchart.
Three iterations, live
Each round was reshaped by customer feedback in the session rather than after it.
Build effort saved
Gaps in the service surfaced before engineering wrote any of it.
How might we embed help content where users actually need it?
Guidance was not where users needed it, so getting stuck meant contacting support. We embedded tips and setup instructions directly into the relevant screens: the help now comes to the user rather than the other way around.
Final implementation: an in-app onboarding checklist.
Users could not see what was done, what was next, or how close they were to finishing. I designed a progress dashboard with clear next steps, which iterative testing turned into an in-app checklist.
Onboarding checklist implementation, shown at 0% and 30% completion. Four design moves work together: (1) the checklist itself, (2) a completion score, (3) effort indicators per task, and (4) a floating mini-checklist that persists across screens.
Four design moves, working together
01Onboarding checklist for easy access
On the homepage, so key tasks are visible from first login. Less confusion, less reliance on support.
02Completion score for progress tracking
A sense of momentum. People finish a setup more often when they can see how close they are.
03Task & effort indicators for clarity
Expected effort shown upfront, so users could plan instead of guessing. It removed a whole category of support ticket.
04Floating checklist, always in reach
Visible across screens, so users never had to navigate back to remember where they were. Fewer dropped sessions.
From six months to 64 days.
We cut setup time by 70%, onboarding calls by 40%, and brought time-to-value down to 64 days.
Behind the numbers: customers reaching value faster, and a Customer Success team with time for strategic work instead of setup calls. The change also held. The self-service flow scaled cleanly as customer numbers grew.
What this project taught me.
Looking back, the lesson I take from this is simple: deep research plus iterative design beats clever guessing every time. Along the way I sharpened my prototyping speed, my cross-functional facilitation, and my willingness to validate before building.
01Process impact
A reusable stakeholder workshop template came out of this project, and the team used it on later work. Naming the artefact turned a one-off facilitation move into a repeatable practice.
02Collaboration impact
Working across Customer Success, Learning Experience, Engineering and Product left behind shared templates and clearer channels. Better documentation, faster decisions.
03Personal growth
Working close to real numbers sharpened how I weigh trade-offs. I now ship a small validated step sooner, rather than a polished but unproven leap.
04Onboarding is a service problem
The 70% came from treating onboarding as a cross-functional service, with the UI as one layer of it. Approached as "make the in-app flow better", it would have shipped prettier screens and saved a call or two.