Geest

Process management SaaS for installation companies

Some details of this project are under an NDA, so I keep the description high-level out of respect for the client.

Geest · mockup pending

Context

Geest is a B2B SaaS for installation companies that helps centralize all their information and build their work processes in one place: flows with nodes, automations, and decision points. Users receive pending tasks within those flows and resolve them by filling out lookup fields and evidence until they're complete. I was the product's Product Designer over a long stretch.

The problem

When I joined, the team was focused almost entirely on fixing bugs. The features that did ship helped users who were already on the platform, but none of them were designed to attract new users: the product wasn't growing.

My role

I took charge of prioritizing which features we shipped based on two criteria: real impact on current users and the potential to attract new ones. Under that criteria I designed Hey Geest, an AI to help new users build processes from scratch, along with the product's table structures, real-time chat, the token system for automations, and the offline app. I led both the desktop version and the mobile app.

Process

Research

Direct feedback from real users through support tickets, which I prioritized by urgency and impact. I also benchmarked the competition to identify best practices in the space.

Constraints

A good part of the product carries a lot of technical context (automations, data structures, node logic) that I initially struggled to fully decode. I solved this with direct meetings with the technical team and my managers to really understand what was needed before designing the solution.

Iterations

For confidentiality reasons I can't go into the specific detail of each version, but the process relied on A/B testing to validate design decisions before shipping them.

Validation

A/B testing.

Final solution

Hey Geest (AI for building processes from scratch), table structures, real-time chat, a token system for automations, and an offline app, on desktop and mobile.

Result

120+ new companies joined the platform, drawn by the features we prioritized with a focus on growth, not just retention.

Takeaway

The biggest friction wasn't design, it was translation: a lot of the context arrived in very technical language and it took me time to understand it before I could design. Next time I'd put together less technical PRDs for the design team from the start, so I wouldn't lose time decoding instead of designing.