Workday’s 50-to-20 AI Cut: Why Aneel Bhusri Killed 30 Agents
Most CEOs don’t have an AI strategy. They have a pile of demos, a committee and 30 projects nobody is brave enough to kill.
Most CEOs don’t have an AI strategy. They have a pile of demos, a committee and 30 projects nobody is brave enough to kill.
Aneel Bhusri came back to run Workday in February 2026 and found 50 AI agents in development. He cut that number to 20. That is not a technology story. It is a leadership story — and an expensive reminder that focus is mostly the willingness to disappoint people.
Workday’s real AI move was subtraction
Bhusri, Workday’s co-founder, returned as CEO after Carl Eschenbach stepped down in February. The company did not bring the founder back because it needed another polished keynote. It brought him back because enterprise software is being reordered by AI, and Workday had to convince customers, employees and investors it was not about to become a very profitable relic.
His response has been refreshingly unglamorous: reduce the number of bets, give the remaining work clear ownership, and demand a hard commercial reason for every project.
The number matters. Going from 50 agents to 20 means 30 initiatives no longer get to consume engineering time, management attention, customer goodwill and internal political oxygen. That is the hidden cost most executives refuse to count. Every project you leave alive tells the organisation, “This might matter.” Before long, nobody knows what actually does.
Bhusri has also described consolidating accountability: one owner for agents, one for the AI platform and one for the tools. Again, obvious. Also rare.
I have watched businesses turn simple priorities into a dog’s breakfast because six executives own a thing “together.” Shared ownership is usually a polite term for nobody having enough authority to make a call — or enough exposure when it goes wrong.
If your AI plan has a steering committee but no single throat to choke, you do not have a plan. You have a calendar invite.
Why the founder came back
The background matters because this was not a victory lap.
Bhusri had previously led Workday for years, handed the top job to Eschenbach, then returned after a rough period for the company’s shares and rising anxiety over its position in an AI-heavy enterprise-software market. Eschenbach left the CEO role and Workday’s board, though he stayed on as a strategic adviser.
That kind of reversal tells you something useful about boards. When the market changes fast enough, a board often stops asking, “Who is the best professional manager?” and starts asking, “Who has the authority to break the furniture?”
Founders have advantages in that moment. They understand the original product instinct. They can call a legacy project nonsense without having to spend three weeks protecting the feelings of the executive who sponsored it. And employees will often grant them more permission to make painful changes because they believe the founder is protecting the company rather than polishing a CV for the next job.
But do not romanticise it. Bringing back a founder is not a strategy. It is a high-stakes admission that the existing strategy was not moving quickly enough.
Workday has genuine scale to protect. In fiscal 2026, it reported $9.552 billion in total revenue, including $8.833 billion in subscription revenue. Its fiscal 2027 guidance called for roughly $9.925 billion to $9.950 billion in subscription revenue, with a non-GAAP operating margin of about 30%. This is not a scrappy startup experimenting from a garage. It is a large public company trying to change its engine while flying at altitude.
That is exactly why the 50-to-20 decision is worth studying. Big companies rarely die because they lack clever people. They die because clever people are permitted to work on too many clever things.
AI has made weak management easier to spot
AI is exposing a management problem that existed long before ChatGPT: companies confuse activity with progress.
A board asks what the AI strategy is. The CEO asks every division to contribute. Each division produces a pilot. Every pilot gets a dashboard. Someone creates an internal innovation day. Six months later, there are dozens of prototypes, plenty of slideware and no meaningful customer outcome.
Then management says AI is complicated.
No. Choosing is complicated. Technology is just where the failure becomes visible.
Bhusri’s point is not that 20 is a magic number. For your business, it might be two projects or 200. The principle is that every initiative must clear a higher bar than internal enthusiasm. Does it solve a painful customer problem? Is the return measurable? Is there an executive owner with the power to ship it? What are we stopping so this can win?
Workday’s stated focus is practical: AI tools for recruiting, contracts, workforce management and employee self-service. Its bet on Sana, the AI company Workday acquired for about $1.1 billion, is part of an effort to make a prompt-based interface that can connect employees to information and workflows across the business.
That is a better framing than “we need agents.” Customers do not wake up wanting agents. They wake up wanting fewer broken processes, fewer tickets, faster decisions and less admin.
The minute your team starts describing the technology rather than the commercial pain it removes, pull the handbrake.
The overlooked lesson: focus is a people decision
The obvious interpretation is that Workday is becoming more disciplined about product investment. Fair enough. The less obvious one is that Bhusri is setting a new social contract inside the company.
When a CEO kills 30 projects, they are telling the organisation that effort alone will not protect a project. Seniority will not protect it. A fashionable acronym will not protect it. Customer impact and strategic fit are the tests.
That changes behaviour far beyond the AI team.
People start bringing clearer proposals because vague ones will die. Managers stop collecting headcount for pet projects because there is no sanctuary in a sprawling portfolio. Product leaders have to say what they will stop doing, not merely what they would like to start.
This is where plenty of leaders get soft. They believe ruthless prioritisation means becoming an arsehole. It does not. You can treat people well while being brutally clear about work that should end. In fact, it is kinder.
Keeping a team on a low-value initiative because you cannot make a decision is not compassionate. It is wasting a chunk of their career.
The best operators I know create psychological safety around disagreement, then create absolute clarity once the call is made. Debate hard before the decision. Execute hard after it. Do not confuse an open door with permanent indecision.
Don’t copy Workday’s founder-return playbook
Here is the contrarian bit: the lesson is not “bring back the founder.” Most companies should not do that, and plenty of founders are brilliant at starting a business but terrible at running a mature one.
Bhusri himself has been candid about being drawn to the product and technology side of the job. His return works only if the company pairs founder urgency with operational discipline. That means capable presidents, clear layers of accountability, hard financial targets and a board willing to judge results rather than nostalgia.
The dangerous version of a founder comeback is cosplay. Everyone cheers the return, dusts off the old values deck and mistakes sentiment for a turnaround plan.
The useful version is this: the CEO uses their authority to simplify the business, reset standards and allocate capital to a very small number of outcomes that matter.
You do not need to be a founder to do that. You need courage, commercial judgement and a calendar that is not run by whoever shouts loudest.
What this means for you
If you run a business, lead a division or manage a team, steal the useful bit tomorrow morning.
First, list every AI, automation, product or growth initiative currently consuming meaningful time or money. Not the ones you remember. All of them.
Second, force each owner to answer four questions on one page:
- What specific customer or operating problem does this solve? - What number will prove it worked by a specific date? - Who owns the result personally? - What are we stopping to fund this properly?
Third, cut harder than feels comfortable. If you cannot name the winner, the initiative is not ready. If it has three executive owners, it has none. If the payoff is described as “strategic learning,” be very suspicious unless you have deliberately ring-fenced a small experimental budget.
Finally, tell your people why you killed the work. Do not hide behind phrases like “portfolio optimisation.” Say: “We are stopping this because it is not important enough, and I want your talent on something that is.”
That is what leadership looks like in an AI boom. Not chasing every shiny tool. Not pretending every employee’s favourite experiment is strategic. Choosing the few things that can genuinely move the business, then having the nerve to let the rest go.