Designing a scalable workflow for admins to feature and organize agents across large organizations
May 2025 - July 2025
I led the design of the Agent Featuring workflow at Glean, enabling admins and moderators to better curate and manage an expanding library of AI agents within large enterprises.
Customers like Dell and Zillow flagged that cluttered admin views and limited discoverability were blockers to rolling out agents at scale. Our goal was to introduce a “Featured” category and moderator tools that improved efficiency, control, and discoverability for both admins and end users.
I partnered closely with another designer, a PM, a user researcher, and four engineers to define the end-to-end experience. My work spanned from framing the problem and facilitating design jams to iterating on flows, prototyping list and card views, and aligning designs with engineering constraints.
Throughout the process, I ensured the solution was both scalable for organizations managing thousands of agents and simple enough to ship within two weeks, unblocking critical customer rollouts.
The Problem Space
In Glean, admins and moderators are responsible for configuring the platform for their organizations. They set up integrations, manage permissions, and curate the experiences that employees rely on to search, discover, and act on knowledge. When Glean introduced agents, admins also decided which agents were visible, how they were organized, and ultimately shaped how employees would adopt them.
However, the way the system worked created significant friction. By default, admins and moderators could see every agent in the system, including private and unshared ones. While this gave them maximum visibility and control, it also led to two major issues:
Cluttered admin views – The agent library was flooded with internal and drafted agents, making it impossible to preview the experience from a typical employee’s perspective. For customers like Dell, this meant they couldn’t run training sessions or demos without confusion.
Limited visibility of key agents – As organizations created hundreds or even thousands of agents, there was no way to highlight which ones were most important. Without tools to feature or reorder agents, discoverability of important agents became an issue.
Customers expressed that they wouldn’t roll out agents at all until these problems were solved. It became clear we needed to quickly close this gap by providing admins with reliable curation and control tools.
I was brought onto the project on a tight timeline and had to onboard quickly. Using the PRD as a starting point, I familiarized myself with the requirements and context so I could immediately begin exploring solutions.
To jumpstart the process, I organized a design jam with a few teammates. In this session, we outlined three key questions to guide our thinking:
Why do we need this feature? – Customers like Dell and Zillow couldn’t roll out agents without it, and needed a way to organize their agent libraries and display important agents to users.
What is the core problem? – Admins had no way to feature or curate agents, and cluttered views prevented them from previewing the employee experience.
How might we solve it? – We sketched different approaches, from toggle-based moderator modes to inline actions and new ways of organizing agents.
Sketching and Riffing
With the problem space defined, I moved quickly into early explorations. I sketched out multiple views for how admins might feature agents and manage categories. I tried out a moderator mode toggle to switch between roles and offer a distinct state change, and also explored embedded moderator actions directly in the agent library.
I saw moderator mode as a compelling option because it not only gave admins full visibility into all agents, but also provided a clear space for admin management tasks. By surfacing a “Featured” category at the top of the library and Assistant home, it offered a straightforward way to curate and guide discoverability for their employees.
I shared a few early prototypes with my team for critique, and my team shared that these early prototypes worked for a small number of agents, but enterprise customers would likely need to feature and reorder dozens at once. I the decided to explore more compact, list-based layouts that could handle higher volume, while still keeping the MVP simple and aligned with familiar Glean patterns,
Implementing Feedback
Using the feedback from my team, I then explored three directions that would be more lightweight and increase visibility:
Dropdown Selection
Dropdown with Chips
Modal Selection
A key requirement for these new directions was enabling bulk selection and giving admins the ability to view and select multiple agents at once, while also exploring a compact list view to replace the earlier grid layout for better scalability.
Dropdown with Chips
Compact List View
I presented the modal selection workflow in an engineering and design sync as it felt the most practical, but a concern was raised that it introduced a new pattern that would take longer to build given our timelines. To keep the MVP feasible, we rerouted and looked to existing patterns already in Glean. The Collections page, which supported a similar workflow for managing documents, became a key reference point—allowing us to align with established UI, reduce implementation risk, and still deliver the functionality admins needed.
However, this discussion also raised a bigger question: were we thinking about this workflow in the right way?
Rethinking the Mental Model
The explorations I had done so far had framed the experience as a separate state: admins could toggle into it, see all private agents, and gain access to moderator-only actions through a dedicated entry point.
Through design and engineering discussions, it became clear that this framing introduced unnecessary complexity, bundled small functionalities into an overly rigid state change, and was not feasible within Glean’s current permission system. Thus, we decided to reframe moderator mode as a set of gated actions, available only to admins, embedded directly in the agent library rather than hidden behind a state change.
This shift not only simplified the workflow but also allowed us to give moderators more specific filtering controls, such as viewing agents created by the entire company or drilling down by department. We called this the “Shared With” filter—by showing only agents shared with specific departments, it gave admins a clear way to preview exactly what a normal employee would see.
With that being said, the initial designs didn't all go to waste. It laid a good foundation for the workflow and gave us a strong starting point to refine into something both feasible and scalable.
A "featured" category only seen by admins
Finalizing the Designs
After multiple rounds of exploration and feedback, we converged on a direction that balanced usability, scalability, and feasibility. The final workflow used the Collections UI modal and a compact list view, aligning with patterns already being developed in Core Product. This gave admins the ability to bulk select, reorder, and manage agents at scale, while reducing engineering complexity and keeping the experience consistent across Glean.
Here are some of the screens from the final designs:
Tooltips for Initial Rollout
Adding and Removing Agents from List
Displaying Featured Agents for End Users
The Core Experience
With the interaction states defined, I moved on to building the core prototype that demonstrated the real-time back-and-forth of the voice assistant. This prototype focused on the foundational experience—a simple conversation loop where the user speaks, the system listens and processes, and then responds. This is a demo of a conversation that I presented in my final presentation, walking the audience through how the assistant could feel natural, responsive, and engaging in practice, while showcasing how the states come together in practice.
With the basics in place, I wanted to explore a larger question: How can we make this a uniquely Glean experience? The experience I explored so far resembles other voice assistants in the market, but Glean has a unique advantage: access to a company’s enterprise knowledge, a rich knowledge graph, and deep integrations with tools like Jira and Slack.
This opens an opportunity to differentiate Glean by building on these strengths and creating a voice assistant that doesn’t just answer general queries, but understands your projects, your workflows, and the tasks you need to get done. In doing so, Glean can move closer to its vision of being a true work assistant, one that feels indispensable in the flow of everyday work.
By building on this foundation and leveraging features already in development, I designed assistive layers that expand the core experience and define a unique value proposition for Glean's voice feature. Together, they position Glean’s voice assistant as more than a generic interaction, and highlight how Glean can leverage its enterprise knowledge graph and integrations to deliver a voice experience that’s deeply embedded in work, setting it apart from competitors.
3 Additional Features
Agentic Looping
Agentic looping, a feature planned for Glean’s assistant, offers an opportunity to bring advanced reasoning into the voice experience. It operates in two modes—default and advanced—giving users the option to loop for more in-depth reasoning when handling complex queries. This flexibility makes conversations feel more fluid while providing transparency into how the system thinks.
The Bigger Picture
The value of this experience is that it positions Glean's assistant as a trusted collaborator—one that understands your working style, anticipates your needs, and is deeply connected to your projects and tools. With a personalized voice and visual experience, Glean can move beyond a search tool to becoming a true work companion—helping people stay productive and engaged whenever and wherever they need it.