LinkedIn Fired 875 People While Its AI Hiring Products Made $450 Million: What That Means for Recruiters

LinkedIn cut 875 employees in May 2026, approximately 5% of its global workforce. The cuts affected engineering, product, marketing, and the Global Business Organisation, the division that manages sales relationships with enterprise clients including recruitment agencies. LinkedIn's stated reason was "building a flatter organisation." The revenue context: LinkedIn was growing at 12% year-over-year when it made the cuts.

At the same time, LinkedIn's AI Hiring Assistant had reached an annualised revenue run rate of $450 million, one of the fastest-growing product lines LinkedIn has launched. The AI Hiring Assistant automates candidate sourcing, personalised InMail outreach, initial screening, and pipeline management. Its own published metrics: 81% reduction in profile reviews needed per role, 66% higher InMail acceptance rate, 1.5 hours saved per role.

Who Is Buying the $450 Million

The $450 million in AI Hiring Assistant revenue does not come from recruitment agencies. It comes from employers: the hiring companies that are LinkedIn's primary direct customers for enterprise hiring products.

This matters for recruitment agencies because the $450 million is being generated by employers who are doing sourcing, screening, and outreach without an agency in the workflow. They are paying LinkedIn for the AI that does what agencies previously did, and they are getting volume and speed metrics, 81% fewer profile reviews, 66% higher InMail acceptance, that make the AI tool look more efficient than the human alternative.

The sales and support teams LinkedIn cut, the Global Business Organisation roles, previously included the account managers who worked with recruitment agencies and enterprise hiring clients. Those roles managed the relationship between LinkedIn and its largest account holders. LinkedIn cut them because the AI platform increasingly does not need a human account manager to generate results for the customer. The $450 million proves the customer is getting enough value from the automated workflow to not need the human support layer.

The Recruiter Licence Paradox

Every recruitment agency paying for LinkedIn Recruiter Corporate is now in the following position: they pay approximately $17,000 to $20,000 AUD per seat per year for access to the LinkedIn database and outreach capability. That payment is revenue for LinkedIn, which allocates it toward the development of AI products. Those AI products include the AI Hiring Assistant, which employers purchase to source and screen candidates without an agency intermediary.

The agency's Recruiter licence fee is, in part, funding the product that makes the agency optional.

This is not LinkedIn's intention: the Recruiter and AI Hiring Assistant products are sold to different customer segments. But the revenue flow is real. The platform charges agencies for sourcing access at the same time it sells employers the AI that automates the sourcing workflow the agency performs.

The InMail cap compounds this. LinkedIn has capped Open InMail sends at under 100 per month per account, down from a practical limit of approximately 800 per month. The agency is paying more for a tool that delivers less volume, while LinkedIn's own AI Hiring Assistant is producing 66% higher InMail acceptance rates for the employers on the other side of the same platform.

What the AI Hiring Assistant Does in Practice

The AI Hiring Assistant operates at the top of the recruitment funnel. It searches the LinkedIn database based on a job description, identifies relevant candidates, generates personalised outreach messages based on each candidate's profile, sends InMails, tracks responses, and surfaces the engaged candidates for human follow-up.

This covers the work that accounts for the majority of recruiter time in a contingency placement model: building the search, reviewing profiles, writing outreach, following up, and managing the initial pipeline. The AI handles these steps at scale and at lower cost per contact than a human recruiter working through the same Recruiter interface.

The work the AI Hiring Assistant does not do: the qualifying conversation with the shortlisted candidate to understand what they actually want in their next role. The counter-offer call when the preferred candidate's employer matches the salary on Friday afternoon. The judgement call about whether the hiring manager's stated requirements match what they will actually accept. The candidate who looked perfect on paper and fell apart at final stage, and needs rebuilding for the next opportunity.

These are the outcomes that require human judgement, built-up market knowledge, and trust relationships. They are also the outcomes that are hardest to price, because they happen in conversations rather than in a database query.

Rebuilding the Fee Around What the AI Cannot Do

The recruitment agencies most exposed to the LinkedIn AI shift are those whose fee is primarily justified by volume sourcing: "we'll get you 10 qualified CVs within 48 hours." That value proposition is directly competed by the AI Hiring Assistant's speed and scale advantages.

The agencies best positioned are those whose fee is justified by outcomes that happen after the shortlist: the hiring manager alignment conversation, the candidate management through the process, the negotiation, and the onboarding. These are the stages where agency expertise produces outcomes the employer cannot achieve with the AI tool alone.

Practically, this means structuring the fee differently. Rather than a single percentage of annual salary at placement, a model that separates the sourcing fee (which AI is compressing) from the process management fee (which AI cannot replace) gives the agency a defensible pricing structure when clients start asking why they are paying full contingency for the sourcing the AI Hiring Assistant can do.

LinkedIn's $450 million is the evidence that this conversation is already happening, at scale, in the employer market. The recruitment agencies that have it proactively, before the client has the AI tool and decides to test it, will be better positioned than the ones who find out when the mandate does not come through.