Shaping AI
Product Vision.

Creating the workflows of the future with an evolving AI product: turning unmet needs into a coherent product vision at speed.

  • PRODUCT DESIGN
  • AI
  • SERVICE DESIGN
  • STRATEGY
Challenge
Redesign how a voting rights research team captured, curated, and trusted what the AI driven system produces.
My Role
Lead product and service designer. I owned discovery, the product audit, the sprint roadmap, and the new designs, from framing the problem through prototype and built solution.
Team Structure
Lead Designer (me), 2 Engineers, 2 Product Managers, Client, Creative Director.
Impact
Increased adoption by 14x across 21 researchers and managers, securing an enterprise-wide contract.

Overview

I joined a team that had done discovery by the book: requirements co-created with users and a roadmap full of requested features. The product still was not working for them. I shifted the work from feature delivery to creating the workflows of the future with an evolving AI product: uncovering unspoken needs, making a coherent vision tangible, and moving from mapping to tested designs without losing delivery speed. The result connected prioritization, research, trust, and team visibility into one workflow.

The problem.

The initial scope focused too narrowly on a single search moment: while technical constraints made meaningful search improvements difficult, we had yet to consider where that moment fit within the wider workflow or where the product could deliver greater value.

The first view isolated the repeated search pattern: broad and specific queries both led to the same scan-and-flag loop.

Widening the scope to the full workflow revealed opportunities to connect prioritizing, monitoring, reviewing, and sharing research with AI, creating a seamless experience and delivering far greater product value without relying on costly improvements to search alone.

The resistance to adoption stemmed from technology that was creating more disruption in daily workflows than it was resolving.

Discovery.

Twenty-eight interviews across leadership, management, and research assistants revealed the unspoken needs the original requirements had missed.

Time, pain, and impact mapped across the team's day
Mapping time, pain, and impact across the day showed where the manual workflow consumed the team, and where AI could give time back. It revealed how to weave AI more seamlessly into their needs, balancing moments that called for human judgment and connection with support that reduced effort without interrupting their flow.

New Product Vision.

I turned the research into a coherent system for the team's future workflow, connecting the AI features across the experience.

A shared foundation

A shared foundation.

The new vision started by mapping the full ideal data workflow the product would support, from incoming information to analysis and action. A reusable summary became the shared foundation, connecting inputs, insights, and next steps across that workflow.

The product vision

The product vision.

This site map outlined the ideal product direction and where each feature should live, grounding the experience in the team's jobs to be done and aligning the product structure with the mapped data workflow. The bolded boxes are the new features, phased in for the roadmap.

Sketching the product.

Before and after.

Before

An untriaged feed.

The feed collected information without helping researchers decide what needed attention first.

After

One connected workspace.

The redesigned research workspace brings briefings, search, and coverage into one workflow.

Key design moments.

Daily Briefings

Rather than layering more relevance signals on the original feed, I designed Daily Briefings as a separate personalized morning view. The Feed stayed the firehose; Briefings became the context-setting starting point.

Team Coverage

A strip showing each researcher, their new item count, and reviewed fraction created a shortcut into focus-area filters while giving managers a clearer view of capacity.

Source Coverage

An internal dashboard made ingestion success visible through a summary strip and per-state status cards. Researchers could expand a jurisdiction, inspect its scraped sources, and preview content without losing context.

Key AI Search

I turned Explore from a blank search bar into a dual-mode surface: keywords for broad clip search, or a question for a synthesized answer with source clips attached. Suggested prompts reduced the cold-start problem for new users.

Outcomes.

A broader product direction emerged from looking past the feature list and into the rhythms of the work itself.

28
User interviews across leadership, managers, and researchers
14x
Adoption increase across 21 researchers and managers
3 tabs
Restructured into briefings, search, and coverage