Designing an identity for LinkedIn Hiring Assistant

LinkedIn launched Hiring Assistant to general availability in 2025 as its first enterprise AI agent. This was an evolution of Recruiter’s existing AI-assisted features into a truly agentic partner that works alongside users’ complex hiring workflows.

We needed to align on and establish an identity for Hiring Assistant — one that would accurately set expectations for the product in a world where agentic tech was still novel. And since the first release wouldn’t include a chat UI, we’d need to establish an agentic affordance through static identity, response frameworks, and architectural patterns.

I led design work for Hiring Assistant’s identity system: its personification, point of view, and interaction patterns within conversational modalities and the broader Recruiter experience. This work evolved alongside the team and product from early 2024 through 2026.

Hiring Assistant was a natural evolution of Recruiter’s suite of existing AI-assisted features. These operated across search, content, and outreach, and they were each designed with deliberately impersonal identities: no names, no first-person references, just system-based components.

And while this decision made sense for those features, Hiring Assistant was different. It was designed to be an autonomous partner, one that would work across workflows, take direction from users, and report back on its work. This autonomy meant that it would need to have a clear, consistent identity for users to engage with. And one that would allow it to accurately report on its work.

This is where my work began. During the discovery sprint for Hiring Assistant, I led the team through a lightning talk session that explored the ways Hiring Assistant would need to diverge — and move beyond — our existing identity practices. It needed to become someone that users would work with across their hiring workflows.

The main shift revolved around personification: a feature’s ability to demonstrate human-like characteristics, behaviors, or communication styles. Given the ways Hiring Assistant would need to work, we needed to build in a higher level of personification.

This didn’t mean creating a character for Hiring Assistant though. We’d still leverage the core voice and tone built into LinkedIn’s brand guidelines — the difference would be in how it interacts with users.

The direction landed with the sprint team and set intentions early for how we’d handle this new paradigm of agentic identity across Recruiter. A few things in particular resonated with leadership:

1. Personification meant a clear graduation path for when we added chat to the product
2. Rooting the identity in LinkedIn’s brand — just personified — allowed for flexibility during a time when the company was working through parallel brand-based decisions.

Hiring Assistant’s charter release was anchored in a full-screen intake experience — the surface where recruiters gave the agent its first direction on the role, the candidates, and how outreach should feel. It was the first moment where users would meet and interact with Hiring Assistant. And without a chat experience baked into the charter release, the page structure itself needed to carry the weight of making the experience feel “agentic.”

We learned pretty quickly that this wasn’t a simple task. Agents were still novel tech at the time, and our team lacked a shared, solid reference for what made something truly feel agentic in a non-conversational UI. As a result, we spent the early months shifting between two approaches:

1. Having Hiring Assistant speak directly to users through the page headers + content in a full-screen experience
2. Using Recruiter’s system-level content to introduce Hiring Assistant + task context — within the Recruiter shell

Both approaches were defensible, and each one had real implications for how we’d build out the rest of the charter experience.

I took the lead on stress-testing these directions with the goal of aligning our group on a single, preferred direction. This involved mapping each approach across the entire intake flow — and exploring each with varying levels of Hiring Assistant’s visual indicators: gradients, colors, and iconography.

These explorations resulted in what became a sort of pseudo-conversation between Hiring Assistant and the user — one where Hiring Assistant spoke in the header and the user responded through inputs, CTAs, and the canvas at large. A few things in particular worked here.

1. The full-screen approach allowed us to lean fully into Hiring Assistant as a new paradigm and tier of experience, rather than something that had fit into the standard Recruiter system.
2. This established a scalable architectural framework for how users would engage with Hiring Assistant before the addition of a chat UI.
3. The visual aspects of Hiring Assistant’s design directly reinforced this architecture for response and engagement.

This pseudo-conversation became a repeatable framework that helped anchor Hiring Assistant’s design and mental model throughout its early days. And it scaled effectively across the full intake flow — as well as the interaction patterns that emerged in the lead-up to GA.

As the product grew post-charter, so did this pseudo-conversational pattern. The full-screen intake experience persisted for projects, but Hiring Assistant also began appearing more contextually in other corners of the Recruiter product — in notifications, dashboards, contextual callouts and highlights.

The framework evolved from a screen-based pattern into a system with a sharper point-of-view model:

• Full-screen surfaces: Hiring Assistant speaks in the first person, in the header.
• Shared Recruiter surfaces: the UI refers to Hiring Assistant in system language, but where the agent speaks for itself — its own cards, notifications, and callouts — it stays in the first person.

The principles remained the same, just with more nuance. First person is the agent’s voice, and it appears only when the agent is actually talking — and within contained, contextual moments.

In spring 2025, conversational capacity became viable for Hiring Assistant, and the product direction moved with it. This meant consolidating everything we’d built previously across intake, setup, and contextual experiences into a dedicated chat modality. It was the biggest test yet of whether these principles, built for a non-conversational UI, would hold up in actual conversation with the agent.

I worked with a cohort of a few other designers to build, test, and lock in product support for this shift — with my work specifically looking at how we carried Hiring Assistant’s identity forward. In practice, this involved adapting the existing personification and POV frameworks into rules for when and how Hiring Assistant talks to users. How it structures responses, how it surfaces suggestions, how it owns and reports back on async work, and how those patterns are encoded into prompting for the agent itself.

The pseudo-conversation we’d resolved the year prior mapped almost exactly onto this new chat affordance. Chat took over from headers as the home for Hiring Assistant’s first-person voice, while the surrounding UI and canvas carried traditional product content. By Talent Connect 2025, the conversational pivot resolved years of accumulated edge cases — and it allowed Hiring Assistant to land as a truly agentic layer woven throughout Recruiter.

Three releases moved the same set of layers around. What changed was where each one sat and how much of Hiring Assistant’s identity it carried.

That last move was the most significant shift in the framework. First-person voice became reserved for in-chat content, and third-person attribution handled nearly every surface outside the chat window. By GA, users understood the product well enough that constant self-reference had diminishing returns — the point-of-view rule written into the 2024 guidelines, applied in practice and validated by a product pivot a year later.

Hiring Assistant went from a managed charter cohort in 2024 to global GA in 2025, and it’s continued to accelerate and grow through 2026. And it’s landed by both financial performance and industry sentiment. The agentic Talent Solutions products it anchors crossed a $450M annualized revenue run-rate, and the product itself was received as a true partner for the hiring industry.

• ~36% weekly customer growth after launch — “outpacing every product LinkedIn’s launched,” per chief business officer Mark Lobosco.
• 21 charter customers to global GA — from a managed cohort in 2024 to general availability in 2025.
• 20+ designers leveraging and building on the identity framework — plus product, design leadership, and adjacent agent teams.

The industry read Hiring Assistant as a partner rather than a feature. Aurecon’s talent lead described it in exactly those terms, and Josh Bersin — the most closely watched HR-tech analyst, and the first major voice to weigh in — called it the first highly-integrated agent he’d seen at launch, then named it among the leading AI-powered talent acquisition products a year later. Satya Nadella cited the agentic Talent Solutions run-rate directly on Microsoft’s FY26 Q3 earnings call, placing a LinkedIn product inside Microsoft’s enterprise AI story.

What ties the results together is how the product felt at every stage of its development: agentic, with or without a chat window. The identity framework is what carried that through. It supported design work across my team and became a foundational part of how LinkedIn builds agents across its enterprise products.