Ryan Breslow Fired HR. AI Is About to Do It at Scale.

Ryan Breslow’s decision to eliminate Bolt’s entire HR department triggered predictable controversy. The real story is more consequential: AI has crossed the capability threshold where it can automate the majority of what the modern HR function does, at lower cost and without the political drift that has plagued the function for a decade or more.

The $500+ billion global HR industry grew on conditions, cheap capital, intense DEI pressure, regulatory complexity, talent scarcity, that are either changing or no longer justify the administrative apparatus they produced. AI is now the forcing function. The executives who understand this are already compressing their people functions. The rest are scheduling another engagement survey.

“HR Team Was Creating Problems That Didn’t Exist”

The Bolt Story Is Not What You Think

In March 2025, Ryan Breslow returned to Bolt – the payments company he founded in his Stanford dorm room in 2014 – as chief executive. He inherited a company that had fallen from an $11 billion valuation in 2022 to roughly $300 million: a collapse of nearly 97 percent. He also inherited, in his words, a ‘sense of entitlement that had festered across the company.’

His diagnosis was unambiguous. The HR team had become the institutional home of that entitlement. The department had been rebranded ‘people ops’ in a prior attempt at reform. That rebranding failed. ‘We had an HR team, and that HR team was creating problems that didn’t exist,’ Breslow told Fortune’s Workforce Innovation Summit. ‘Those problems disappeared when I let them go.’

He eliminated the department entirely.

“We need a group of people who are very oriented around getting things done, and there is just a culture of not getting things done and complaining a lot.”Ryan Breslow, CEO, Bolt – Fortune Workforce Innovation Summit

The public reaction fixated on the optics, the culture-war implications, and the management style. But the more important question went unasked: what was HR actually doing at Bolt, and why wasn’t AI already doing it? That is the question every CEO needs to be asking right now.

The Bloat Equation

The HR industry is not a marginal operation. In the United Kingdom, for example, the profession employs more than 500,000 people and has grown by 83 percent since 2011. In the United States, the Bureau of Labor Statistics counts approximately 800,000 HR professionals, a figure that excludes the sprawling ecosystem of HR consultants, DEI contractors, employee engagement firms, and workforce analytics vendors that attach themselves to corporate people functions at considerable additional cost.

This expansion occurred during a specific era: the post-2008, low-interest-rate, talent-scarce period in which companies were flush with capital, competing aggressively for talent, and facing intensifying regulatory and reputational pressure around diversity, equity, and inclusion. HR expanded to meet those demands. The problem is that it kept expanding after those conditions shifted, and it expanded into territory with limited connection to business performance.

Whitehall documents released in 2025 under the Freedom of Information Act documented 200 HR events and meetings across seven UK government departments during working hours in a single month. These included listening circles for specific employee groups, themed music sessions, and two-hour DEI workshops. Whether one agrees with the underlying goals is beside the point. An administrative function consumed 200 organizational meetings in one month without generating a measurable unit of economic output – that is the point.

This is Goodhart’s Law in its most advanced form. When HR compliance metrics, DEI participation rates, and engagement survey scores became official organizational targets, they stopped functioning as indicators of health and became ends in themselves. The function optimized for its own metrics rather than for business outcomes. Process theater replaced performance. A pathology was born.

The Principal-Agent Collapse

Classical economics provides a precise framework for this pathology. In principal-agent theory, a principal, the business owner or board, hires an agent to manage a defined set of tasks on their behalf. When this relationship works, agent interests align with principal interests. When it fails, the agent’s interests diverge: typically in the direction of the agent’s own expansion, self-preservation, and values rather than the principal’s objectives.

What HR was retained to do: attract and retain talent, ensure legal compliance, manage compensation design, and build the organizational capability that drives competitive performance.

What large HR functions increasingly became: internal institutions with their own ideological programs, self-determined metrics, internal coalition politics, and an expansion logic that had decoupled entirely from business results. The UK government HR data is not an outlier. It is the trajectory made visible.

The traditional remedies for principal-agent drift are tighter monitoring, clearer incentive structures, and harder accountability. AI offers a structurally more robust solution: replacement of the agent’s administrative functions with a system that cannot drift politically, cannot accumulate internal power, and reports directly to the data the principal cares about. You cannot capture an algorithm with organizational politics.

What AI Can Now Do That HR Would Not

Here is the disruption the Bolt story obscures: the legitimate functions of the HR department are now largely automatable. Not theoretically. Actually, today, with tools already deployed at scale inside large organizations. The gap between AI’s current capability and most HR teams’ current workload is significantly smaller than most CEOs expect.

Recruitment screening and candidate matching is fully automatable now. Tools from Eightfold.ai, Beamery, and Workday’s AI layer screen and rank candidates against defined job requirements with greater consistency and less demonstrated bias than human recruiters, in seconds, at scale, at a fraction of the cost of a recruiting team.

Employee query resolution is already automated inside many large organizations. Conversational AI handles the vast majority of routine HR interactions: benefits questions, policy lookups, leave requests, payroll inquiries. ServiceNow, Leena AI, and Microsoft Copilot integrations across enterprise HRIS systems handle millions of these interactions monthly. No coordinator required.

Compliance monitoring and pay equity analysis runs continuously on AI platforms. Tools including Syndio and Knoetic provide real-time pay equity analysis across entire workforces, flagging disparities before they become legal exposure. This is work that previously required a team of analysts and an external consulting engagement. AI delivers it continuously.

Workforce analytics, real-time attrition risk modeling, productivity pattern analysis, skills gap identification, and workforce planning, are now automated capabilities inside Visier, Workday, and SAP SuccessFactors. A single analyst oversees what previously required a team of ten.

The functions AI cannot yet replace are clear: identifying future talent markets, designing organizational structures for competitive advantage, managing complex executive transitions, and building the cultural architecture of a high-performance organization. These require judgment, experience, and contextual intelligence that current AI does not replicate.

The problem is that most HR departments spend the overwhelming majority of their time on the automatable work, wrapped in process complexity that makes it appear more difficult than it is. That is the opportunity AI now makes visible, and the exposure it makes unavoidable.

The Lean People Ops Stack: What Replaces the Department

The blueprint for the post-HR organization is not speculative. It is being assembled, function by function, by companies that understand where AI capability has arrived. The emerging model operates in four distinct layers, each with a defined scope and a clear owner.

At the transactional layer, conversational AI handles all routine employee interactions around the clock. An employee has a benefits question at 11pm. They ask the AI. The AI answers from the policy database and logs the interaction. No coordinator, no ticket, no next-business-day response time.

At the compliance and analytics layer, AI systems run continuously against workforce data. They flag outliers, generate regulatory reports, model attrition risk, and surface compensation anomalies. A single analyst manages a workload that previously required a full department.

At the talent acquisition layer, AI handles sourcing, initial screening, scheduling, and candidate communication. A smaller team of senior human recruiters focuses exclusively on relationship-intensive roles and executive hiring, the work that requires contextual judgment and relationship credibility that AI cannot yet provide.

At the strategic layer, a small, senior people function makes organizational design decisions, manages succession, runs executive development, and maintains the cultural operating system. This team is lean because AI has absorbed everything beneath it. The 30-person HR department at a 500-person company compresses to 6. The 200-person HR function at a 10,000-person enterprise compresses toward 40. These numbers are already being achieved at AI-native technology companies and selected enterprise early adopters.

The Breslow Model and What It Actually Signals

Breslow’s move was not clean. Eliminating an entire HR function in a single action, without pre-built AI infrastructure, creates real legal and operational exposure. Compliance functions cannot simply disappear. In many jurisdictions, specific HR processes are legally mandated and cannot be delegated to an algorithm without careful transition planning. The abruptness of the Bolt approach reflects the urgency of a turnaround situation more than a replicable template for stable-operating organizations.

But the underlying strategic logic is sound. If your HR function has become a self-referential bureaucracy optimizing for its own existence rather than business performance, another rebranding exercise will not fix it. The solution is radical compression, AI-enabled redesign, and a redefinition of what people operations actually means in a world where AI handles 70 percent of what HR has historically done.

The executives who will execute this well are not the ones firing their HR teams in a press conference moment. They are the ones quietly mapping which HR functions AI can absorb, deploying the tooling, redesigning the organizational structure, and transitioning their people function toward genuine strategic work. They will have smaller teams, better data, faster decisions, and significantly less organizational drag.

Three Decisions Every CEO Faces Now

The Bolt story forces a set of decisions that most leadership teams have been deferring. AI capability is real. Cost pressure is real. The organizational drag from over-staffed, politically drifted HR functions is real. The question is not whether to act. It is how.

Audit the function against AI capability. Map every task your HR team performs against what current AI tools can actually do today. The gap will be significantly smaller than most CEOs expect. That gap is the strategic plan.

Separate compliance from administration from strategy. Legal and regulatory compliance functions carry real liability if eliminated without replacement. Cultural and organizational design work is genuinely valuable and largely irreplaceable by AI today. Administrative and process functions are the compression target. Start there.

Build the AI infrastructure before reducing headcount. The Breslow approach, eliminate first, build later, is viable only in turnaround situations where operational survival is the binding constraint. In stable operating conditions, the sequence reverses: deploy AI tooling, validate against the existing workload, transition roles, reduce through attrition. This sequence protects legal compliance and preserves institutional knowledge while capturing the structural cost advantage.

The organizations that execute this sequence correctly will have a structural cost and speed advantage that compounds over years. Those that ignore it will find themselves defending a function that costs more than its AI-enabled replacement and delivers measurably less.

The Bottom Line

Ryan Breslow made a move that most CEOs are thinking about and have not yet made. He executed it too bluntly, too publicly, and without the AI infrastructure that would make it sustainable. But he was right about the underlying problem, and right about the direction of travel.

The global HR industry grew on a set of conditions, cheap capital, intense DEI pressure, regulatory complexity, talent scarcity, that are either changing or no longer justify the administrative apparatus they produced. AI has now arrived at the capability threshold where it can replace the majority of that apparatus with systems that are faster, cheaper, more consistent, and impossible to capture by internal political dynamics.

The Breslow moment is not an isolated management story. It is the first public signal of a restructuring that is already underway at AI-native companies and will move through the broader enterprise market over the next three years. The competitive gap between organizations that adapt and those that defend the status quo will be substantial.

The executives who understand this are already building leaner people functions. The rest are scheduling another engagement survey.