Turning rigid approvals into a flexible visual workflow builder
Problem
Some enterprise clients left
The approval builder only handled linear workflows, so clients couldn’t model how their procurement really works.
My role
Only designer, with 1 PM and 8 engineers
Led the Workflow Setup redesign end to end and made the case to add Initiatives to the scope. Extended the design system along the way.
Outcome
+20%
Initiative creation among early adopters in the first month. The new builder rolled out to 50+ enterprise customers.
Sastrify helps companies control their software spending. Its approval builder couldn’t model how enterprise procurement actually works.
Affinity map from 10+ stakeholder interviews
The insight
The brief was to redesign Workflow Setup. Talking to our procurement team, I realised that was only half of it.
No one in the company could hold all the paths from creating a workflow through to Initiatives in their head. Not even the procurement team, who lived in it daily. And the workflows themselves were too rigid to model real approvals. Neither piece worked on its own, so I pushed to put both in scope.
Making the case
Two features instead of one needed a yes from leadership, so I brought evidence.
First, I ran a usability test of Initiatives with our CEO and CTO. They struggled with basic tasks and couldn’t find basic information in their own product, and they saw it happen.
Then I asked our procurement team how many of our clients ran this exact step on a competitor’s platform while also paying for our procurement team. That showed where the gap was: the team was strong, and the product was weak right where clients work with them.
We got the yes. The cost was a longer launch, and Workflows shipped in smaller chunks.
Before: a linear form
One step stacked on the next, in a flat form. No branches, no conditions, no way to say “if it’s over 10k, add finance.” One assignee per step and nothing in parallel, so one person on holiday froze every approval behind them. Teams that needed more rebuilt the same workflow by hand every quarter.
After: a node-based builder
A workflow is now a canvas. Drop in steps, branch them, set conditions, see the whole path at once. Departments approve in parallel, so one absence no longer stalls the queue.
The drawer
I wanted a pop-up next to the node, like our competitors. Deadlines pushed me to the drawer, already in our design system. It aged better: room to grow, so each release could add what customers asked for.
10 incremental releases
Each small enough to test with real customers before the next. Release 2 tried drag and drop; users found it too complicated, so release 3 swapped it for a plus sign. I designed all 10. I left after release 5, and the rest shipped with only small tweaks from testing.
Decisions
Four calls, what each replaced, and what it cost.
Two features in scope
cost · A longer launch, with Workflows shipped in smaller chunks
instead of
The one in the briefThe drawer from the design system
cost · A less spatial feel on the canvas
instead of
A pop-up next to the nodeA plus sign to add a node
cost · Dropping a feature one release old
instead of
Drag and drop (release 2)Ten small releases
cost · A half-built canvas in production for a while
instead of
One big launch
The builder rolled out in the first release cycle. Teams that used to rebuild workflows by hand every quarter were spinning them up in minutes, and first-wave customers came back for a second workflow within a week.
Mixpanel, first month after launch vs the month before. 10+ is against the launch adoption goal.
From a blank page to a working workflow in under two minutes.
Reflection
The brief was one screen. Stepping back before committing is now the first thing I do on anything new, because on Sastrify that’s what caught the real problem. Shipping in 10 small releases meant customers got value from release 1 and we could test between each.
The fun part was the scope fight: one feature in the brief, two in the real problem. If you work on enterprise tools where the hard part is the logic underneath, I’d love to talk.
More work