Open Gov
During the summer of '23 as a product design intern at Opengov, I lead the designs for their first AI feature. A writing assistance for government documentation helps users write faster and better.
Scope of work documentation is complex and time consuming to write. Previous automation attempts did not resolve user problems. This AI-feature helps users create a quick contextual first draft - saving hours of time.
Outcomes
Shipped, 70.4% increase in efficiency,
Presented at
Role
Owned 0-1 Designs & Research
Methods & Tools
Interviews, Hierarchical Task Analysis, Usability Testing User Flow, Hi-Fi Prototypes, Design System (Figma)
Context
Writing SOWs has traditionally been a tedious and time-consuming process, leading to significant time and cost impacts for governments. While past solutions such as libraries, templates, and references offered some relief, users still faced high effort requirements and lengthy drafting cycles.
Image : HTA digram of different steps a user can take to write a scope of work section
Users struggled most at the start of the writing process. An empty document created anxiety about saying the “right” thing, with fear of mistakes and pressure to sound polished immediately. Without guidance, users had to organize ideas, define structure, and choose language all at once, making it hard to translate raw thoughts into the expected format and often leading to slow starts or repeated restarts.
4/5 users explicitly described starting the document as the hardest part
0
5
*Data is based on on user study with 5 users
Users found it time-consuming & high effort to locate a starting point that resembled their use case. They had to review multiple templates or examples before committing, which slowed momentum and increased uncertainty early in the process.
3/5 users spent 30+ minutes finding a suitable starting template
0
5
*Data is based on on user study with 5 users
After selecting a template, editing required multiple manual steps. Users had to import relevant sections, decide what applied to their use case, and then rewrite content to fit their context. This fragmented workflow increased effort and reduced the efficiency templates were intended to provide.
5/5 users rewrote more than 90% of the template
0
5
*Data is based on on user study with 5 users
Framing Opportunity
How might we reduce the effort it takes to start, shape, and edit content
Blank Canvas Paralysis → Guided Start
Manual Content Mapping → Contextual Content
Multi-Step Editing Workflow→ Streamlined Editing
Design Iterations
I went through multiple rounds of iteration and guerrilla testing with proxy users and stakeholders, alongside technical reviews to ensure feasibility. I also participated in an org-wide design critique, incorporating feedback and insights from across the organization.
Designing for Today and Tomorrow
OpenGov had an existing live editor and a tight four-week timeline to ship its first AI feature ahead of NIGP ’23, while a broader platform redesign was underway in parallel.
To balance speed and scalability, I designed two versions of the feature: an MVP optimized for feasibility within the current editor, and a longer-term pattern designed to scale across future editors and products.
MVP Designs
The MVP builds on OpenGov’s existing editor while directly addressing key user pain points. A contextual guided start reduces blank canvas paralysis by customizing the prompt based on the user’s project title, with backend prompt engineering generating a strong first draft. Contextual examples help users find a relevant starting point quickly, and a streamlined editing flow replaces multi-step editing—users simply select the sections they want and import them in one action. Content is inserted in the editor’s native structure (title and body blocks), making edits faster and more intuitive.
Blank Canvas Paralysis → AI-Guided Starting Point
Manual Content Mapping → Context-Aware Content Generation
Multi-Step Editing Workflow→ Native, One-Step Content Insertion
Final Designs : Scalable, Block-Level AI Pattern
The long-term design introduces AI at the block level rather than the document level, informed by research showing better outcomes when users work on smaller, focused tasks.
This pattern aligns with how users naturally write—engaging AI at the moment they begin editing a block—while remaining flexible enough to scale across different OpenGov products and editors. To support trust and review, AI-generated blocks are visually distinguished and clearly labeled, enabling both authors and reviewers to easily identify and verify AI-assisted content.
Blank Canvas Paralysis → In-Context AI Entry & Starting Point
Manual Content Mapping → Block-Level Context Awareness
Multi-Step Editing Workflow→ Inline, Traceable AI Blocks
Outcomes
In a within-subject study with five participants, average time to complete the task dropped from 25 minutes to 7.4 minutes—a 70.4% reduction. Given that this task often takes several hours in real-world workflows, the actual efficiency gains delivered by the AI assistant are likely significantly higher.
The feature shipped in September 2023 as OpenGov’s first AI capability and debuted at the NIGP ’23 national government conference.
Team
Design & Dev
Jen Elhers (Design Lead & Mentor), Matt H
Product Owners
Toah Hill, Matt K