Designed a prototype for Pulse Labs' advanced web-based "Audience" builder which meaningfully merges an AI assistant with a manual logic editor to intuitively and precisely filter a large pool of verified participants for naturalistic research studies. The prototype introduced user-first improvements like...

Constant input feedback

Flexible, precise user control

Visibility of system status

Skip to my process

Skip to solution

Background

Context

Pulse Labs’ main SaaS product connects any product with real users to draw out naturalistic testing insights.

One of our main value propositions was offering a large pool of eager, verified participants to help our clients draw out valuable, user-centric product insights. While this pool of participants was high quality, it was also diverse, so we needed a way to filter them down for studies—even the best testing data is useless if the wrong participants are involved, so it was vital that our filters are precise and intuitive.

Problem

Note: This feature takes place some time in the initial creation phase of the study. Before this prototype, Pulse Labs employees would create studies on behalf of users because the study creation tools were too complicated.

The CEO had a vision: “Make filtering for participants as easy as writing an AI prompt”. That’s fine; when generative AI works well, it’s magical. But AI is not always reliable, especially on complex, grammatically-ambiguous prompts.

And the requirements for study participants can get pretty specific, like:

We needed a way to make filtering participants as simple as possible without sacrificing precision or control; if the participants are wrong, the data is useless.

Approach

Championing the User

My team leader and I both agreed that having manual control over the filtering was mandatory to ensure filtering precision, so relying only on AI would not be best for the UX. After a quick mockup showing how AI tools would benefit from manual controls, the CEO agreed but preferred I initially lead users into an AI-assisted creation flow instead of manual creation. This way, everyone wins.

My team leader also added a few more requirements for the feature:

• Handle abstract user prompts that do not fit into the pre-determined demographic parameters.

• Alert users if their filters fetch less participants than required, probably because they are too specific.

I planned out the feature and it made sense to separate the UI into two main sections: (1) a section to prompt the AI assistant and (2) a section to view/edit filters manually, all in one screen. Users can bounce around both sections (or stay in one) depending on their needs or preferences.

Only showing one section at a time could feel prohibitive and inflexible.

Quick wireframe concept shows filters as independent options and the AI assistant making one-way changes to filters. Spoiler: That would change.

Filter Methodology

Considering all requirements, I felt strongly that a visual logic builder was ideal for these filters because its logical operators allowed for more precision and flexibility than typical dropdown filters. It would also help users see the status of their filters at all times.

I designed menus to reflect our demographic parameters, quantitative equalities/inequalities, and mathematical operators.

These dropdown components combine to create logical statements such as:

I designed the logical parameters to read like a casual, conversational sentence to help users learn how to program the logic, as opposed to a formal programming language.

Solution

Layout

The final layout is split into two distinct, vertical sections (vertical layout accommodates chat messages and stacked filters better than horizontally). The two sections are dependent and reflect changes made to each other in real time for optimal flexibility and efficiency.

Dynamic Design

The UI leads users with the primary option to enter a prompt describing the kind of participants they need with the AI Assistant. This makes building complex filtering very easy. Secondary options to build manually or upload a document (like a business brief) are also available.

Note: The "Build Manually" option disappears because manual edits become available elsewhere in the UI. The "Upload Document" button minimizes because it's no longer the user's main option for AI communication.

After a prompt is entered, the UI splits into two distinct sections separating the AI Assistant (left) from the Logic Builder (right).

After a valid prompt, the AI assistant builds the filter logic in real-time, allowing the user to see exactly how participants are being selected.

The AI Assistant will give users feedback about their prompts and filter logic, such as if enough participants are found or if there's an invalid prompt request.

Users may manually change the logic at any time to suit their needs or correct issues from the AI Assistant due to ambiguous prompts. This grants the ease of prompting with the flexibility of manual control.

Even after a manual change, the AI Assistant will confirm the user's input and give feedback so the user is never confused about what to do next.

These reciprocal interactions give the user visibility of the system status and continual feedback no matter how they choose to build their filter logic.

Edge cases

If the prompt cannot be described by our predetermined parameters (too abstract), the AI assistant will suggest an entry point to a screening survey.

Next steps...

Here's what I'd do differently if I had the opportunity, time, and resources:

• Implement the ability to build very complex requirements with many logical operators for extreme edge cases

• Conduct usability testing with our clients

See more work