I led design for Spoke, an ML-driven enterprise product. I managed a small team, designed a drag-and-drop workflow editor, and learned a lot about designing for AI.
Request in Spoke
Any question or service request, whether submitted via the web platform or a chat app like Slack, gets pulled into the ticketing UI.

Spoke was a magical B2B product that used machine learning to automatically answer employee questions in Slack, MS Teams, or our own bespoke web platform. It would find existing answers in the company’s knowledgebase, or triage requests to the most relevant employees based on their availability and area of expertise.

I led design prior to Spoke’s acquisition by Okta in 2021, managing a tiny team of one other product designer and one marketing designer, while also doing my own product design work. I designed our Workflow Automations tool, which enabled admins to construct rich human-in-the-loop automations using a drag-and-drop interface. (Think Zapier or IFTTT, but for enterprise workflows that could be triggered via natural language commands in Slack).

Workflow builder in Spoke
Admins can construct advanced automations using the workflow builder.
Workflow logic in Spoke
Long story, but there's a good chance you know Spoke as either askSpoke or atSpoke. We ended our run as atSpoke, but this is my website and I still like the name Spoke best, so I'm going to call it that.

Design Principles

One of the three Spoke founders was a product designer, and the company had design baked into its DNA from the start. During my tenure as design lead, I kept us aligned to the following principles:

Audience settings in Spoke
Audience segments provide guardrails on who can and cannot access certain resources or workflows.
Service Catalog in Spoke
If they don't want to use natural langauge, users can trigger service requests directly in Spoke.