UnifyApps Automations
Designing the iPaaS and workflow builder for an enterprise AI platform: making powerful automation infrastructure feel approachable for business users without taking control away from engineers.
Make enterprise automation feel approachable, without taking away control
UnifyApps is an enterprise AI platform: agents, a no-code app builder, data pipelines, and the part I worked on, Automations, the iPaaS where teams wire real business processes across 500+ connected systems. Through ICD, the agency I design with, I owned the design of the iPaaS and workflows module from early stage through maturation.
The brief was never "design screens." As the team framed it, the job was to make a powerful automation infrastructure feel approachable, reliable, and scalable for business users and technical teams alike. That one sentence hides the entire problem.
One builder, two users who want opposite things
A workflow builder in the enterprise serves two people at the same canvas. A business user wants to connect Salesforce to Slack and get on with their day; to them, complexity is a tax. An engineer wants retries, error handling, caching, and circuit breakers; to them, control is the entire point. Hide the power and you lose the engineers. Expose all of it and you scare off everyone else.
Most of the real work lived on that seam. The decisions below are each a way of serving both audiences without averaging them into something neither one wants.
Designing for trust, not just for tasks
Role-aware diffs and version control
Editing a live workflow is dangerous, so the first job was to make change visible, reviewable, and safe. A diff view with split panes and line-level highlights, but framed by role: engineers see the raw configuration, PMs see the business impact of the same change. One source of truth, read two ways.
Progressive disclosure of power
Retry logic, error handling, caching, and circuit breakers all live in a right-side controls panel, grouped and revealed progressively. The depth is always there for the engineer who needs it, and never the first thing a business user is forced to confront.
Assistance at the point of friction
Rather than a generic AI chatbot, contextual autofill surfaces exactly where people get stuck: configuring an API, filling inputs, debugging a failed run. Help appears where the friction is, not in a panel nobody opens.
A connector grammar that scales
Hundreds of connectors can't each be a bespoke screen. I shaped one consistent pattern, input, action, output, so learning one connector teaches you all of them. The mental model holds even as the integration count keeps moving.
The decisions, on screen
The hard part of enterprise tools is the people, not the pipes
The infrastructure underneath was already powerful. The design problem was closer to psychology: making something genuinely complex feel trustworthy to a beginner and respectful to an expert, inside the same interface. The win was never a prettier screen. It was a builder that lets a business user automate a real process and lets an engineer trust what shipped, which is the only way enterprise software actually gets adopted.