Natural Language Search
B2B HR Platform
![]()
Project overview
As we built out People Search, a core workforce search experience, we were tasked with considering how to incorporate natural language processing (NLP), an emerging product strategy, into the experience.
Project goals
- Easily educate and onboard users to a new search paradigm
- Provide a personalized experience by surfacing tailored search recommendations
- Partner with the AI team to integrate people search patterns into the global AI and chatbot strategy
Role & team
In this phase, my role expanded beyond designing the core People Search experience to shaping how natural language capabilities would be introduced. I partnered closely with the Data Science team to understand model limitations and translate them into actionable design decisions. While still collaborating with my Product and Engineering peers, I also aligned our local search patterns with broader AI initiatives, including the global search experience and chatbot strategy. This required influencing not only implementation details, but also how the organization thought about scaling AI-driven interactions across the platform.Process
Discovery
To frame the problem, I asked:
- What level of complexity makes sense in this context?
- What level of complexity can we support?
- How does the platform AI chatbot fit in?
- How do we manage cost and scope?
- Are we keeping up with consumer trends and expectations?
I analyzed usage data and saw that users typically applied 1-2 filters at a time. That insight guided me to focus the MVP around the 5 most common filters: employment status, hire date, termination date, position, and cost centers related to department and location.
I defined some initial query examples.
From there, I developed visual concepts to effectively convey my ideas. With more open-ended directives, such as “integrate NLS”, having a visual representation of the overall vision was necessary to achieve quick alignment.
First visit:
![]()
- Conversational placeholder text
- Default suggestions
Subsequent visits:
![]()
- Recent searches
- Dynamic suggestions based on usage
Design & iteration
After aligning on direction, I partnered with Data Science to translate query ideas into a feasible MVP. Together, we narrowed the scope to the filters that could be reliably supported within the timeframe: employment status, hire date, termination date, and position. We excluded cost centers (often used for departments or locations) because their usage varies across clients. This decision ensured we could launch with a high-quality foundation and expand in the future.
I designed a system of dynamic suggestions that adapted to what a user typed, showing them how queries could be phrased and combined.
Typing “new hires” could trigger suggestions such as:
- people hired this week
- people hired last month
- people hired in 2025
To reinforce the new paradigm, I also defined contextual placeholder text in the search bar.
Default state: Broad and conversational
![]()
Focus state: Specific and instructional
![]()
MVP design
Outcomes
- The first iteration of natural language search was built in one quarter. It was rolled out internally and to a select group of clients in Beta. Early feedback confirmed the potential of NLS to make searching more intuitive.
- Because the MVP supported only a subset of available filters, stakeholders decided to hold back the functionality from the broader People Search launch. This allowed us to protect trust in the new search experience while positioning NLS for future integration with the upcoming universal search initiative.