Ask Hubert:
Search, simplified
Ask Hubert helps employees get answers from ServiceNow and SharePoint without knowing where the source material lives. The product goal was straightforward: reduce search friction, improve self-service, and make company knowledge easier to act on.
I built a React + TypeScript prototype to explore the layout, interaction model, source display, and response patterns before the product moved deeper into delivery.
The prototype gave stakeholders and engineers a shared reference for how the assistant should behave. It also made design decisions easier to test: how answers should cite sources, where supporting links should appear, and how much control users should have when AI confidence varies.
That hands-on build helped bridge strategy, UX, and engineering detail so the team could discuss the product as an experience, not just a concept.
- Responsive layout built with Tailwind CSS.
- Dark and light mode support through
prefers-color-scheme. - Mocked GPT-4o response flow using the OpenAI SDK.
- Sidebar patterns for source attribution and supporting links.
- Project configuration and setup notes for handoff.
- Trust and governance: AI-powered search was new for the enterprise, so the experience needed visible source attribution, clear boundaries, and a cautious rollout model.
- Content quality: answer quality depended on source material that was structured enough for people to read and AI to parse. I proposed content standards to support both.
- Security and compliance: the assistant needed to return useful answers without bypassing corporate access rules or creating false confidence.
- Delivery model: the team was adapting to SAFe Agile, so UX work had to fit incremental planning while still protecting the product vision.
- Defined the assistant's UI direction, brand cues, and conversational tone so AI responses felt clear and trustworthy.
- Designed workflows that balanced automation with user control, especially around source review and next-step links.
- Led structured usability testing with qualitative feedback, then used the findings to refine flows and response behavior.
- Partnered with solution architects, engineers, and product owners to embed UX standards into planning and delivery.
- Helped prioritize incremental improvements so product decisions stayed tied to business goals and user needs.
Prototype flow mapping the ideal path from question entry to cited answer and supporting links.
Brand and interface exploration for a trusted assistant experience, including tone, visual language, and response patterns.- Established content guidelines that improved readability for employees and parsing quality for AI responses.
- Improved answer clarity through source attribution and direct links to relevant policies, guides, and documentation.
- Used structured testing feedback to guide iteration before broader rollout.
- Strengthened UX maturity by giving product, engineering, and stakeholder teams a clearer model for AI-assisted employee tools.