For the past year, many executive teams have treated AI as a productivity story. Faster recruiting. Better case management. Cleaner compensation decisions. More scalable learning. Less manual work.
Those gains matter. But across recent M1 conversations, a sharper theme is emerging: the real constraint is no longer access to technology. It is leadership imagination.
The organizations moving fastest are not simply asking how AI can improve the current HR model. They are asking what the HR model should become if AI can absorb large parts of the transactional burden, connect fragmented systems, personalize support, and make expertise available at scale.
That distinction matters. Optimization keeps the old operating model intact. Imagination starts to redesign it.
At the June 30 NYC meeting, one leader framed the posture plainly: employees are being asked to pursue AI opportunities where the risk of failure may be greater than the chance of success. Others echoed the same leadership requirement. You do not have to know exactly how the vision will be achieved. But you do have to be bold enough to name it, and credible enough that people believe the operating model will actually change.
That may be the defining AI leadership challenge as we enter the second half of 2026.
AI Has Moved From Experiment To Architecture
Earlier this year, an M1 insight described HR as moving to the center of enterprise redesign. The Q2 meetings suggest that shift is accelerating.
In Mexico City, leaders described AI and automation not as a lab exercise but as embedded infrastructure across core HR work. One organization had automated recruiting and hiring for a workforce that hires roughly 45,000 people annually. Another described moving from 1,800 HR employees in 2021 to 1,200 today, while building automated processes across recruiting, compensation, employee data, exits, and learning. The same organization now gives tens of thousands of employees access to HR information through digital tools and is exploring another 20 to 30 percent reduction in HR department size.
The point is not the reduction itself. The point is that AI is forcing a more honest conversation about what HR humans should be doing.
If machines can answer policy questions, process exits, draft development plans, monitor pay decisions, translate learning content, and route employee requests, then the HR operating model cannot remain a lightly automated version of the old one. The function has to decide what work deserves human judgment, what work should be handled by agents or automation, and what new work becomes possible once transactional load falls away.
That is where many leadership teams are still under-reaching.
The Imagination Dilemma
Several members noted a common pattern: executives are using AI to improve existing swim lanes rather than rethinking how value is created. They are asking how to make the current process faster, not whether the process should still exist.
This is understandable. Incremental use cases are easier to approve, easier to govern, and easier to explain. They also preserve the current leadership comfort zone.
But the more advanced conversations are moving elsewhere. In Chicago, agent demonstrations were described less as vendor showcases and more as a way to spark “the art of the possible.” The value was not simply seeing a recruiting, compensation, coaching, or learning agent. It was watching whether HR leaders became more curious, more specific, and more willing to think beyond familiar process maps.
One participant put the value of those sessions in practical terms: they helped leaders see what was in the realm of possibility before deciding how broadly to engage the organization.
That may be the right sequence. Not “buy tools, then transform.” Not “write a strategy, then hope people believe it.” First, stretch the leadership imagination. Then redesign the work.
Some of the scarcest capabilities named in NYC were not prompt engineering or AI literacy. They were imagination, humility, and the courage to be wrong.
Precision Is Becoming A Leadership Discipline
Boldness does not mean treating every problem as an agent problem.
One of the more useful tensions from the June 30 discussion was the reminder that RPA, predictive analytics, generative AI, and agents are different tools with different economics. When someone says, “I need an agent,” the better answer may be, “You need RPA.” Or a workflow redesign. Or better data hygiene. Or a predictive model. Or a human.
That precision will matter more as AI cost becomes more visible. Members are already beginning to talk about token cost-effectiveness and whether every AI-enabled task is worth the compute it consumes. One prediction from NYC was that employees may eventually work within individual token budgets, spending compute deliberately rather than applying AI indiscriminately.
That framing is powerful because it moves AI from novelty to resource discipline. Tokens become a kind of workforce currency. Leaders will need to decide where machine reasoning creates enough value to justify its cost, where cheaper automation is sufficient, and where human judgment remains the better investment.
The next wave of AI maturity may look less like enthusiasm and more like economic clarity.
The Agent Ecosystem Is Coming, But It Will Need A Front Door
London and Mexico City both pointed toward a more complex agent landscape. Leaders are experimenting with in-house agents, vendor-built agents, and specialist external tools that sit on top of HR platforms, policy documents, SharePoint, Workday, SuccessFactors, or other enterprise systems.
One organization described building an HR agent layer over internal data and documents, with security permissions matched to each employee’s role. Others discussed specialist agents for coaching, recruiting, benefits, compensation, learning, and employee servicing. The likely future is not one agent. It is an ecosystem.
That creates a design problem. If employees and managers face 60 tools, adoption will collapse. The better model may be a small number of “front door” agents that understand the request, route it to the right specialist, and know when to hand the work to a human.
This also reframes the debate about whether agents should have names. The emerging view from NYC was pragmatic: naming may help early adoption, but humanization may matter less as usage matures. The concern about defaulting to female names and voices is real, and leaders should treat naming as part of adoption design, not branding theater.
The real question is not whether the agent has a friendly name. It is whether the system helps people get work done without confusion, false confidence, or unnecessary friction.
HR’s Prize Is Not Efficiency. It Is Strategic Capacity.
The most important HR question raised in NYC may have been: what can’t HR get to today because of everything transactional?
That is the point leaders should keep returning to. AI is not only a cost lever. It is a capacity lever. It creates the possibility for HR to spend less time administering the enterprise and more time reshaping it.
That includes work the function has long claimed as strategic but often lacked the bandwidth to do well: workforce redesign, leadership capability, culture change, skills architecture, manager effectiveness, and transformation readiness.
Recent discussions on AI-enabled mental health support pointed back to the same idea. The opportunity is not simply to deploy another tool, but to intervene earlier and scale support in areas HR has struggled to reach consistently.
The real prize is not a leaner HR function. It is a function with more room to do the work only humans, and especially HR leaders, can do.
The Mid-year AI Leadership Question
The operational leaders driving incremental improvement still matter. Not every leader needs to be the visionary. AI transformation will require disciplined operators, careful governors, systems thinkers, and pragmatic implementers.
But the latest bigger signal is that HR leadership now requires a more explicit posture of boldness. Not recklessness. Not hype. Boldness grounded in economic discipline, human judgment, and operating-model intent.
The leaders who will shape the next version of HR are asking different questions:
What work should disappear?
What work should move to agents, automation, or analytics?
What work becomes more human, not less?
Where are we optimizing because reimagining feels too risky?
And if our people do not believe we will actually change the operating model, why would they participate in changing it?
That last question may be the most important one. AI adoption is not just a technology rollout. It is a credibility test. Employees are watching whether leaders are using AI to decorate the old model or build the next one.
The technology is moving quickly. The harder question is whether leadership imagination can move with it.
