Workato · 2024 — 2025
Recipe-driven data pipelines
Workato moved data between systems in real time but couldn't replicate it at volume. I led design on the features that closed the gap — inside a product someone else had already shipped.
Workato's integration platform let disparate systems talk to each other in real time, but it could not move large volumes of data for analysis or storage. Enterprises kept a second vendor — Fivetran, Informatica — for that half of the job.
A v1 of data pipelines had already shipped and the designer on it had moved to other work. I joined to take the team through a roadmap of more sources and destinations and deeper replication capabilities, on a core UX that was already in users' hands. The constraint shaped everything: propose patterns that scale to features not yet specified, without introducing changes users didn't ask for.
Every feature on the roadmap landed on one of these four. Where it landed decided which existing pattern it should extend.
Put PII control on the field, not in a settings page
Sensitive fields have to be protected before they are replicated, and the decision is per field. The PM and I wrote the copy for a field-level control that lets a user hash a value or replicate it as is, and placed it in the right margin of the schema row — next to the thing it applies to.
I explored a range of hash icons so the mark would sit systemically with the surrounding copy and read correctly on its own, without a label to carry it.
Reuse the margin affordance one level up
Users needed to trigger a one-time sync on an individual object, so data movement wasn't locked to the schedule. That meant object-level actions and object-level settings had to live somewhere the schema view already went.
Rather than open a new surface, I extended the right-margin affordance introduced for hashing to the object row. One learned gesture, two scopes — and a pattern the next feature could take without a redesign.
Three kinds of sync now had to be legible in one chart
Adding object re-syncs meant the dashboard was suddenly reporting three things at once: initial syncs, incremental syncs, and re-syncs. Reading the chart no longer told you what kind of run you were looking at.
We introduced indicator icons in the chart views and carried them into tooltips, so type is visible at a glance and explained on hover — a small addition that kept the existing dashboard intact rather than replacing it.
Each release added components and patterns back to the shared library, so the next feature started from something rather than nothing. A later file-connector pipeline needed more metadata surfaced per row; the icon-and-tooltip pattern already existed and absorbed it without a new layout.
Inheriting a product is a design constraint with a payoff. Respecting prior work — extending its affordances instead of replacing them — is what let three features land in a row without users having to relearn the screen.