Bolt Fired HR. Microsoft Rewired It. IBM Automated It. So Is the Structure Broken?

By Angela Lane & Sergey Gorbatov We sat down for a LinkedIn Live last week with one question on the marquee: Is the HR structure broken? The week answered before we did.

On May 19, 2026, Bolt CEO Ryan Breslow took the stage at Fortune's Workforce Innovation Summit and announced he had fired the entire HR department. His line: "We had an HR team, and that HR team was creating problems that didn't exist." Bolt's valuation had collapsed from $11 billion to roughly $300 million. He cut 30% of the headcount. What's left of HR is now called People Ops, with a remit limited to training and basic employee support. Breslow had written on LinkedIn that "HR is the wrong energy, format, and approach."

A few weeks earlier, Microsoft Chief People Officer Amy Coleman pushed a different button. She consolidated engineering HR, merged HR4HR with Culture and Inclusion into a new People and Culture unit, joined People Analytics with the Employee Experience team, and brought in a new VP of Workforce Acceleration focused on skilling, redeployment, and human-agent collaboration. The Chief Diversity Officer is out. No theatrics; just a memo.

And IBM did a third thing entirely. CEO Arvind Krishna told the Wall Street Journal that AI has replaced the work of "a couple hundred" HR roles. Its AskHR agent now automates 94% of routine HR tasks, including pay statements. Customer satisfaction on internal HR support rose from –35 to +74 on a net promoter–style measure. Total headcount went up, not down, because Krishna redirected savings into software engineers and salespeople.

Three reactions to the same pressure. Three different verdicts on the same model. So which one is right?

The model they're all reacting to is 30 years old

David Ulrich published Human Resource Champions in 1996, on the back of the globalization-era restructuring that had already rewired the rest of the business. Big supply chains, big shared services. He proposed an equivalent for HR: pull transactions into a service center, pull expertise into Centers of Excellence, put Business Partners next to the business. Three legs. One stool. Over 70% of large enterprises base their HR structure on some version of the Ulrich model.

The promise was beautiful: operational excellence, deep expertise, strategic partnership. The execution has been variable on a good day, theatre on a bad one.

What broke isn't the model

A colleague recently described his global HR stack to us: over 400 platforms across the footprint, none of them integrated. Workday was supposed to solve this a decade ago. It didn't, because every layer added its own tool to solve its own problem, and nobody owned the integration. Operational excellence became operational sprawl.

Centers of Excellence were supposed to deliver depth. Many became depth without context, producing beautiful programs the business never asked for and could not deploy.

Business Partners were supposed to be strategists. Ask one for the strategy. You will get answers ranging from "what's a strategy?" to "come back next year." A few have one. Most became the glue stitching CoE design to Ops delivery, then explaining the seams to the line.

Then Employee Experience appeared as a horizontal layer on top, because the three legs hadn't delivered an experience. We are still not sure why EX is a thing, but some make it a thing.

This is a capability failure and an accountability failure dressed up as a structural debate. Nobody asked Business Partners for a strategy, so they didn't write one. Nobody asked CoEs to develop product management, service design, financial acumen, or AI literacy, so they didn't. Nobody owned the platform sprawl, so it grew.

Where the same structure works

Prudential kept all three legs. Then they pulled people out of those legs and built cross-functional teams that get deployed to the top three people-and-organisation problems the business names. Same model, new operating rhythm, and it works.

Haier went the opposite extreme. After firing 12,000 middle-managers, Haier divided itself into 4000 independent "micro-enterprises", with HR as one of several internal marketplaces the micro-enterprises buy from or ignore. Look at the structure cold and you wonder how anything ships. Look at the numbers and they ship. The model runs on brutal clarity about who creates value for whom.

IBM took a third path. They didn't fight the model. They automated the bottom of it, redirected the savings to capabilities AI can't replicate, and the HR NPS rebounded once the rollout was fixed. The Ulrich model, augmented.

Same structural ancestry. Three viable strategies. The question isn't "which structure" but "which execution."

The criteria for what's next

AI doesn't respect functional boundaries. It cuts end to end. The CoE that designs a program no longer needs to hand it off to Ops to roll out and to Business Partners to land. The CoE can ship direct to leader, direct to employee, with the Business Partner holding the trust relationship and the standards. This is already happening in companies you would name.

So what survives the next cut?

• Wisdom. AI gives you knowledge in seconds. Wisdom is knowing when to scale a coaching bot to all employees and when a struggling commercial director needs a human in the room. The Business Partner role survives if it carries judgment and political nerve. It dies if it carries summaries.

• Conscience. AI can run a goal-setting cycle. It cannot tell a senior leader they are about to do something wrong. HR sits with audit and compliance as one of the few functions with standing to call out right and wrong. Don't outsource that to a chatbot.

• End-to-end ownership. Value gets created across the chain. If your KPIs reward only what happens inside your leg of the stool, you are measuring the wrong thing.

• Capability over credential. Most CoEs were not built for product management, service design, financial acumen, AI literacy, or cost-to-serve thinking. The honest answer for closing that gap: rotate people, retrain people, or replace people. Pick.

Our verdict

Cracked in most places. Broken in some. The best thing they ever did in a few.

Same architecture, different outcomes. The variable is whether anyone in the building can articulate what they want from HR, and whether anyone is held accountable for delivering it.

Breslow at Bolt is wrong about what his HR team was doing. It wasn't inventing problems. He hadn't built an HR function that could solve any. Firing it felt decisive. It hasn't fixed the company; Bolt's valuation is still down 97% from peak.

Microsoft is doing something more useful. They asked what capabilities matter most for the next decade and built structure around capability, not the other way around.

IBM did something different again. They automated what AI does better, kept the wisdom on humans, and grew the company.

None of these models saves you from the harder work.

A question for next time

We had this conversation as two people who came up through Centers of Excellence and the business. We brought our biases.

The next LinkedIn Live brings the room. Business Partners in the middle of this transition. CoE leaders defending their turf. Ops leaders who live with 400 platforms. We'll put them on the same call and watch what happens when they have to face each other.

The question we'll open with:

If you built HR from scratch tomorrow, knowing what AI can now do for free and what only humans can do because no machine ever will, what would survive the cut, and who in your organisation has the standing to make it?

Bring your answers. Bring your scars.

The opinions expressed here are those of the Authors and not the organisations with whom they are proudly associated.