For a while, workplace AI’s sales pitch was beautifully simple: automate the dull bits, free up people for higher-value work, collect the productivity dividend. It had the clean, irresistible logic of a spreadsheet that has never met a manager.
This week’s news makes the messier reality clearer. AI is changing the ingredients of work - tasks, skills, approval chains and career paths - faster than many businesses can redesign the jobs around them. The question is no longer whether AI can create capacity. It is who has a plan for it.
TL;DR
-
Workday wants employers to see how AI changes tasks, skills and career paths before they redesign jobs by instinct.
-
Josh Bersin’s new Workday partnership signals that agentic HR will need operating models and education, not just new software.
-
Anthropic’s own data shows that more automated output can create a fresh human burden: coordination, judgment and attention.
Workday Wants To Make Job Redesign Less Of A Guess
At Workday Rising, the company introduced Work Intelligence, a capability that breaks jobs into their constituent tasks, connects those tasks to skills and helps leaders assess what AI should automate, augment or leave firmly human.
It is available in early access to Workday HCM customers and is designed to inform some awkward decisions: should a business hire, reskill existing people, create a new role or redesign an old one? Workday says the capability can track how employee time changes, which skills are growing and whether workers are actually moving into new roles and responsibilities.
That is a more serious proposition than another dashboard promising “skills visibility.” As UC Today noted in its Workday Rising verdict, Work Intelligence could make task-level workforce planning more practical. The important word, however, is could.
A platform can show that a finance analyst now spends less time reconciling data and more time interpreting it. It cannot ensure the analyst is trained, trusted or given enough room to do the interpretive work well. Nor can it stop a company treating the time saved as a reason to ratchet up targets.
Workday’s own framing is telling. Nikki Helmer, its VP of Talent Management and Insights, said:
“the basic building block of work is shifting from the job to the task.”
The implication is substantial: job architecture, progression and performance management were built for a world in which roles changed more slowly. AI is asking HR to keep up with work in motion.
The Key Question: Who Owns The Time AI Creates?
Josh Bersin’s newly announced partnership with Workday lands squarely on this problem.
The partnership will produce whitepapers, tools, education and advisory services under the banner of HR 2030, helping employers move towards what Bersin calls a “growth-oriented agentic model” of HR. Future plans include embedding Bersin’s Galileo AI platform into Workday.
That may sound like the traditional analyst-vendor handshake, complete with an alarming amount of future tense. Yet the underlying argument deserves attention. If organisations are moving from a workforce of people using applications to a system where people and agents share work, HR needs to govern more than headcount and learning content.
It needs to define where human judgment sits. It needs to make skills development tangible when entry-level tasks are automated. And it needs to decide whether freed capacity is reinvested in better service, innovation and development - or simply absorbed into a larger workload.
The risk is that “agentic HR” becomes a glossy label for a familiar outcome: fewer clear career ladders, more opaque performance expectations and an employee base told that efficiency is somehow its own reward.
Workday and Bersin are right that businesses need a playbook. The question for leaders is whether they will use it to create a credible bargain with employees: AI changes your work, and we will help you move into what comes next.
The Lesson: Automation Still Needs Someone To Read The Room
Anthropic’s new report on how employment is changing inside Anthropic offers a useful corrective to the idea that task automation simply gives people time back.
The company reports code commits per engineer rising fivefold in six months, alongside more contracts processed by lawyers, more invoices processed by accounts-payable staff and more deals closed per salesperson. On one level, the productivity story is hard to dispute.
But its internal data also exposes a coordination tax. Median meeting time per employee rose 20% year on year, while “meta-work” - creating visibility tools and monitoring agent pipelines -rose from 10% of interactive sessions in autumn 2025 to 22% in spring 2026.
One employee described facing “an army of Claudes” demanding attention. Another compared the agents to children needing their juice boxes opened, each presenting their issue as equally urgent. Funny, yes. Also a fairly concise description of what happens when output multiplies faster than a team’s ability to prioritise it.
Anthropic is not a typical enterprise, and it says as much. Its workforce practices and unusually intense AI use will not translate neatly elsewhere. It has also reduced its share of junior technical hires, which is a warning that internal mobility does little for people who never get the entry point.
Still, the lesson travels. The human role does not disappear when AI takes on tasks; it shifts towards setting direction, checking quality, resolving exceptions and deciding which work deserves attention. That is not the absence of work. It is a different, often more demanding kind.
AI productivity is becoming a workforce-design problem because the technology changes what work is before organisations have agreed what people should become. Workday can map the change, Bersin can help write the playbook and Anthropic can demonstrate the friction. The leadership test is turning all that new capacity into better work - not simply faster work with more things flashing for attention.
