Why Train Humans at All?
Rethinking the Workforce in the Age of AI Agents
Earlier this year, ServiceNow and NVIDIA announced a collaboration to advance agentic AI. That refers to AI that doesn’t just assist with tasks, but takes on full roles. Their focus is improving the reliability and performance of AI agents before they are deployed so they can operate with greater trust in business environments.
It sounds technical. But the implications are deeply strategic.
This isn’t just about upgrading automation tools. It actually signals a fundamental shift in how we think about workforce design, employee capability development, and the operating model of the modern workplace.
Shortly after the ServiceNow and NVIDIA announcement, Kore.ai, a not-yet-public, but significant company financnially backed by NVIDIA, also unveiled its enterprise-grade multi-agent orchestration framework.
Add that to ServiceNow’s March Yokohama release, and it becomes clear that the AI conversation has moved well beyond simple task automation. We’re now seeing the emergence of full-scale AI workforce coordination.
Then came Accenture.
At a May advisor day, it unveiled its own multi-agent orchestration engine, built around a growing catalogue of specialised digital workers. If there was any lingering doubt, this should end it. The shift isn’t just about automating tasks, it’s about rearchitecting the workforce itself.
If you are still trying to make sense of what’s really happening you are not alone. The daily drumbeat of AI announcements is disorienting.
But here’s how I’ve started to make sense of it. This isn’t just about more intelligent tools. It’s about a new operating model for the workplace. And the common thread running through many of these moves, NVIDIA, is providing a surprisingly useful reference point.
Once known for powering AI through its GPUs, NVIDIA is now embedding itself into the platforms where AI agents will live and operate. Rather than building all the agents itself (though it is doing that too), NVIDIA is laying the tracks they’ll run on. It’s positioning itself as the infrastructure, orchestration, and execution layer for the next era of enterprise work. Think of it like Intel, or even Windows, for AI agents.
In a future where every enterprise manages hundreds (if not thousands) of AI agents, from finance assistants to customer service bots to procurement specialists to marketing gurus, someone needs to power the ecosystem. NVIDIA wants to be that AI Workforce Operating System.
So that’s the important backstory. But here’s the real headline.
In an agentic workforce, why are we still training humans for roles that no longer make sense? That is no longer a question for futurists. It’s a strategic imperative for every executive team because AI has moved from being a tool applied to work, to becoming a participant within the work.
Microsoft now allows AI agents built in Copilot Studio or Azure AI to be assigned managed identities within Entra Active Directory. That effectively gives them permissions, access controls, and security protocols just like human employees.
That’s not workflow automation. That’s workforce integration.
Think about the people in your business. And the unfilled positions, and the ceaseless requests for addition headcount. For decades, we’ve worked within the natural constraints of human capital. We hire, we train, we try to retain. We’ve built complex onboarding and performance frameworks, hoping each employee stays long enough to return that investment.
And when they don’t, we explain it away with generational or cultural clichés. Millennials lack loyalty. Gen Z wants purpose. But these are distractions. The real problem isn’t about who is leaving. It’s about why we keep designing roles that only work if people stay. If we can even get them in the first place.
There’s a structural mismatch between the time, effort, and cost required to build human capability and the speed, precision, and resilience that today’s organisations demand. Part of that is because constant disruption is a key part of modern business.
Take frontline public service roles, like those at Services Australia1. Proficiency takes years. Yet turnover is constant. Staff leave each month. Many actively look for a way out. Not because they aren’t capable, but because the roles have become unsustainable. They are emotionally demanding, under-resourced, and burdened with unrealistic KPIs. The employee liabilities are significant.
My wife, a qualified lawyer and grief counsellor, once worked for a major government-funded training and employment provider. The policy and services onboarding was brutal. The emotional toll was heavy. The turnover? Predictably high.
In my own time in city government, I saw the same pattern. Our call centre staff handling parking disputes needed deep policy and legislative knowledge. But above all, they needed resilience. Their wellbeing was tested daily.
These are all examples of where we are asking humans to fill roles that no longer align with the way work, or life, actually functions.
This isn’t an argument to replace people with AI. It’s about facing the fact that some roles have become incompatible with human sustainability. AI isn’t here to displace. I think it is a critical tool that is here to preserve what matters, by taking on what no longer does.
So do we just keep training? Keep rehiring? Keep burning money? The question is why? Not philosophically. Operationally.
Why continue investing finite resources training people for roles they’re statistically unlikely to stay in? Once, there was no alternative. But now, AI agents can do that work without sick days, without churn, and with total consistency. This isn’t a rhetorical question anymore. It’s a strategic decision.
This doesn’t mean removing humans from the equation. In many roles, you can’t. But it does mean reallocating people to where they add the most value and using AI agents for structured, repeatable, emotionally draining tasks that don’t retain people anyway. That’s the opportunity we keep missing. AI isn’t a replacement. It’s a reallocation engine.
And the platforms enabling this transition like ServiceNow, Kore.ai, Salesforce, Microsoft, NVIDIA, Accenture aren’t just selling software. They’re building the infrastructure for a new kind of workforce called the Agentic Workforce.
Some agents assist. Others act autonomously. They’re not dumb bots. They’re skilled digital teammates. And, if we’re honest, they’re already better than humans at a lot of things. They don’t get tired. They don’t resign. And they learn fast. Really fast.
This is why system integrators like Accenture and Infosys are pivoting. Their whole labour-arbitrage model used to scale by hiring. Now it will scale by orchestration.
Practically though it is not just about the global BPOs and SIs and MSPs feeding the global hunger for outsourcing and skilled labour. Every workforce plan now needs a new line item: Agentic Entities (ID-assigned).
Microsoft’s Entra Agent ID makes this real. Agents now get unique, managed identities. They’re secure, auditable and governed under policy. They are not rogue bots. They’re recognised colleagues managed like staff, not software.
It also addresses the next looming threat which is agent sprawl, the unchecked spread of AI agents without governance. Entra Agent ID offers a framework to contain that risk. In short, you don’t just deploy agents, you manage them. That changes everything.
So tear up your legacy training programs. Rethink your workforce assumptions. The platforms, data pipelines, and orchestration layers now being built are equivalent to laying rail during the industrial revolution. By comparison, your HR onboarding playbook is probably still riding horseback.
For decades, we’ve designed learning programs to shrink the gap between new hires and peak performers. But in the age of AI agents, performance isn’t something you build over time. It’s something you configure, deploy, and tune.
So this isn’t about whether AI will replace jobs. That’s a distraction. The better question is:
Why are we still training humans for roles they were never meant to stay in, when agentic AI is ready to work today?
Because the real shift isn’t about humans versus AI. It’s about creating teams where both thrive. Where humans focus on what only they can do. And AI takes on the work we never should have asked people to do in the first place.
This isn’t the future of work. It is already here. It is time to let go of the old assumptions. It is time to redesign your operating model. To modernise your core. To choose your hybrid platform architecture (HYPA) wisely.
The real leaders in this new era won’t be the ones who resist AI. They’ll be the ones who strike a balance by blending the best of human and agent to build something stronger than either could alone.
https://www.anao.gov.au/work/performance-audit/performance-management-the-australian-public-service
https://www.paulfletcher.com.au/media-releases/media-release-2023-aps-census-result-toxic-culture-at-services-australia-exposed
https://www.sbs.com.au/news/article/longer-wait-times-staff-shortages-centrelink-struggles-to-keep-up-with-surge-in-demand/bam4zmnnl
https://www.thesenior.com.au/story/8086433/centrelink-faces-staff-shortage-as-wait-times-increase/