Under the hood: the tech behind the first agent from LinkedIn, Hiring Assistant

Since last October, we have reimagined our engineering strategy to expand our generative AI-powered experience for LinkedIn members, recruiters, learners, marketers, and sellers.?

Agents are the next evolution of our AI story, and today, we are launching Hiring Assistant, the first agent from LinkedIn. You can read more about how we’ll be helping recruiters by taking on their most repetitive tasks with LinkedIn Hiring Assistant in a post from Hari Srinivasan, Vice President of Product for LinkedIn Talent Solutions.?

In this post, I’ll highlight our engineering team's technological approach to building our first agent - specifically, the technology advancements required to bring this agent to market and our commitment to building a human-centric agent.??

The technology behind the LinkedIn Hiring Assistant

Hiring Assistant is the latest development in our generative AI journey and another example of what's possible when working with natural language as an input. To power this transformative tool, our engineers developed new technology and innovative approaches:

• LLMs for large scale automation: While we’ve released many AI-powered products over the past year, this is the first time we are using LLMs for deeply personalized and sophisticated workflow automation at scale. Some examples of this include:?
• Personalized assistance: One of the most novel features in our Hiring Assistant is what we call experiential memory, which is its ability to learn from its activity or interactions with each individual recruiter. This means that over time, Hiring Assistant will also become more personalized to each recruiter, learning from their feedback and preferences, so they can focus on more strategic, interpersonal aspects of work. For example, when a recruiter tells a hiring agent they prefer candidates with leadership skills, it will seek to understand the decisions via conversation with the recruiter and incorporate that into all future sourcing tasks.?
• Enabling agent-user interaction: Work doesn’t get done in silos and tasks aren’t often as straightforward as one would hope. Recruiters have to navigate multiple tools and tasks to accomplish their work, like sourcing, messaging candidates, conversations with the hiring manager, and more. To effectively delegate various workflows to Hiring Assistant, we’ve created a new agent orchestration layer that uses the reasoning abilities of LLMs to organize and act through interactions with recruiters and support from tools that enable things like search and messaging. This layer helps us take a real-world approach to tasks - iterative, asynchronous and collaborative.

We're also continuing to build upon existing technology used in Recruiter 2024, incorporating learnings and investing in new capabilities with Hiring Assistant:

• Our AI-assisted messages help with writing personalized, high-quality InMails. This tech will be leveraged by Hiring Assistant in the coming year to support automated candidate follow-ups that can respond to basic candidate questions about the role.
• Incorporating our semantic search capabilities allowed us to improve Hiring Assistant’s ability to answer complex questions, rank quality candidates, and more.?
• Powered by insights from the LinkedIn Economic Graph, Hiring Assistant helps create a candidate pipeline and infer a fit based on various attributes, such as background, skills, and more. These insights create transparency for recruiters wanting to understand how candidates can match and support their needs.?

Building responsibly with recruiters at the center?

Our commitment to building member trust and ensuring safe and responsible AI is at the forefront of our agent technology. Our agents are purpose-built to help assist our members and customers efficiently and securely. As we developed Hiring Assistant, we conducted rigorous evaluations to identify potential gaps and risks, such as hallucinations and low-quality content. Actions are audited, and reported in the same manner as human users. This ensures that activities maintain the same level of transparency and accountability.

This, coupled with trust defenses to prevent generative AI from creating content that doesn’t meet our standards, helps to power a valuable, trustworthy and reliable experience. We also grounded our approach in our Responsible AI Principles so our use of AI remains trustworthy, fair, transparent and consistent.

We ensured recruiters are always in control with Hiring Assistant – workflow and task management allows recruiters to start, stop, confirm or edit actions in every step of the sourcing process. It also maintains a complete audit log of its work, so the recruiter can thoroughly assess recommendations and provide feedback. This feedback is then used to continuously improve performance and better align with the unique working traits of the recruiter.?

Helping our members and customers stay ahead?

Just like the role of the modern-day recruiter, many of our members and customers are experiencing changes in working with generative AI. This is why we are always looking for new ways to refine our products to enhance our members’ and customers’ experiences and help them navigate work changes.?

We are excited to introduce Hiring Assistant as the next step in this journey, empowering recruiters so they can focus on the tasks that inspired them to take on the role in the first place.?

For more details about Hiring Assistant, check out Hari Srinivasan ’s ?post.