VeriComms
🔎 Upgrading an AI-powered tool that supports public health communicators with research and content creation
OVERVIEW
Connecting science & communication
Public health communicators have limited time and budgets to fight health misinformation. To address this, Science to People is building VeriComms, an AI-powered tool that helps them create trustworthy content using vetted scientific sources.
The Problem
Poor layouts make it hard for communicators to find key information like sources, and limited features make collaboration difficult.
The Solution
Our cross-functional team at Tech Fleet designed a research-informed prototype with clearer navigation, making it easier for communicators to find key information like sources. We also concepted sharing and commenting features to support collaboration.
MY ROLE
Leading design thinking & collaboration
I guided the UX Design team to ensure our work was grounded in usability best practices and research. I collaborated closely with researchers, strategists, and writers to rapidly improve the product in 10 weeks.
Contributions:
- Audited the base tool to identify usability issues and opportunity gaps
- Designed a streamlined prototype with enhanced navigation and useful features
- Moderated interviews to gain a deep understanding of users and their workflows
- Facilitated stakeholder workshops, client demos, and team meetings
DISCOVERY
Interviewing public health communicators
VeriComms was originally developed with limited research: 16 surveys and 329 unstructured interactions.
Due to confidentiality restrictions, we could not test the base tool with participants. Instead, I helped our researchers with interviewing 30 public health communicators to understand their workflows, pain points, and use of AI. Our researchers prioritized the questions with the client, recorded the interviews, and analyzed transcripts in Dovetail to find key insights.
Key insights:
- AI tools should be versatile – Communicators expect AI tools to handle varied requests from drafting campaigns to data analysis.
- AI should assist, not decide – They use AI to help draft content, but they rely on their own expertise to shape and finalize the message.
- Trusted sources are essential – They verify whether AI responses are backed by reliable sources, especially on sensitive topics such as vaccines.
- Content creation is collaborative – When creating content, they involve multiple reviewers to ensure accuracy, sensitivity, and compliance. Participant 23 describes the in-depth process:
“A subject matter expert will provide the content, which is reviewed by their division leader. I create a mockup and submit it to my manager, who does a literacy and design review. After that is approved, it’s sent to the communications director to ensure that it is compliant with our branding. We then send it back to the subject matter expert to review and they give us feedback. Then it will be approved for dissemination to the public.“
AUDIT
Identifying gaps & opportunities
Drawing from best practice guidelines, interview insights, and competitor designs, I identified usability gaps and opportunities to better support communicators. We compiled our recommendations in a slide deck, and I reviewed them with the client in a workshop. Due to confidentiality, these example slides show a de-branded version of the tool.
DESIGN
Prototyping with research
With the design and research teams, I constructed a low-fidelity prototype with two goals:
- To gather feedback from communicators on whether the prototype supports their key work tasks.
- To illustrate for the client an example of more useful and accessible experiences.
Feature prioritization
We brainstormed features based on user needs identified in our research, then prioritized which ones to include in the prototype based on value and effort.
Prioritized features:
- Example prompts that illustrate the tool’s capabilities
- Information sources that are easy to verify
- A text translation feature
- Sharing and commenting tools to involve collaborators
Task Flow
We mapped a task flow that served as an outline for our prototype. Each step matched the prioritized features and is designed as a task for participant testing.
Initial Prototype
With a solid task flow, we assembled a low-fidelity prototype, looking closely at competitor designs to craft the layouts. We incorporated the prioritized features and opportunities identified in the audit.
TESTING & ITERATIONS
Improving feature discoverability
We ran concept tests with five health communicators, then uploaded the transcripts to Dovetail for qualitative analysis. Participants appreciated the prototype’s ease of use, source visibility, and productivity potential. One participant remarked that the tool would be straightforward for stakeholders to collaborate with:
“I think people would learn to use [VeriComms] pretty quickly. It’s kind of like ChatGPT; you just jump in.”
Yet, participants overlooked the template and translation features, instead wanting to use the prompt field to complete the tasks. Looking closely at design patterns in ChatGPT, Google Translate, and Gemini, we addressed these issues by improving discoverability and aligning with participants’ preference.
Users need cues for new features
Problem
0 of the 5 participants noticed the Use style template feature.
Solution
A coach mark draws attention to and introduces the feature.
Users need features that align with their mental models
Problem
0 of the 5 participants noticed the translation feature, instead wanting to use the prompt field.
Solution
Users can translate text with a contextual prompt field.
FINAL DESIGNS
An intuitive tool for communicators
The prototype reflects the insights we identified in our user research, supporting communicators with verifying AI responses, brainstorming and writing, and collaborating with stakeholders.
Communicators can verify sources via hover citations and a dedicated tab
A contextual prompt field supports intuitive AI-assisted editing, such as translating text
REFLECTIONS
Collaborating to build
In just 10 weeks, our team gained insight into communicators’ workflows, explored AI design patterns, and adjusted to confidentiality restrictions. Our close collaboration enabled us to provide improved user experiences for VeriComms, earning positive feedback from the client:
“Your team’s research and UX work directly supported our ability to move forward with fundraising and development. The prototype gave our team a clearer understanding of user needs, leading to more intentional product scoping and feature prioritization for VeriComms. Several of your UX recommendations have been shared with [our developers] and are being incorporated into the current design and build phase.“
Brinleigh Murphy-Reuter
CEO & Founder, Science to People
Given more time, I would integrate our persona—finalized at the end of the phase—into the prototype. Incorporating the persona would further anchor our designs in user needs and illustrate the product’s user journey for the client.