People Matters Logo

From tools to tasks: Why AI-led workforce transformation has to start with work

• By Branded Content Team
From tools to tasks: Why AI-led workforce transformation has to start with work

Most organisations have given employees access to AI.  But far fewer have redesigned work around it. 


At a People Matters × Nuvepro roundtable in Mumbai, senior HR and transformation leaders looked at what it takes to close that gap. The theme, "AI-Led Workforce Transformation: Redesigning Work", was held in Mumbai on 30 September 2026.

CHROs, HR transformation heads and L&D leaders from manufacturing, financial services, insurance, telecom, infrastructure, pharma and technology talked candidly about where AI is changing work, what happens to roles when it does, and why giving people tools does not make them ready. Giridhar L V, CEO and Co-founder of Nuvepro, delivered the opening keynote.

Start with the work, not the tool, 

Giridhar, who is also a co-author of The Agentic Enterprise – Organisation Reinvented: Redesigning Work, Roles, and Teams for the Age of AI Agents, argued that AI-led transformation goes beyond technology adoption; it requires organisations to rethink and redesign how work gets done.

He proposed three interconnected lenses: the workflow (how work is broken down, executed, handed off, and increasingly automated), the workforce (how roles, skills, responsibilities and career paths evolve), and the workspace (how digital and physical environments adapt to the new way of working). At the centre of this analysis is the task, which he called the "atomic" unit of work. Nuvepro's own analysis started from the US O*NET standard of about 900 occupations and broke down around 100,000 job postings into close to five million tasks. A single cashier role can carry 30-plus tasks, he noted. Workflows, unlike job descriptions, are specific to each company.

When clients asked Nuvepro to "create an AI engineer", his reply was: "That is a title. Tell me what this person is supposed to be doing first." Only then can an organisation understand the tasks involved and decide what to automate, what to augment and what to keep with humans.

Leaders tested the idea live. One broke a recruiter's role into about 15 distinct tasks. The room agreed that AI opportunities sit at the level of tasks, not job titles. 

Organisations are starting in different places. In one, every function head must bring at least one AI deployment that changed a workflow to the management review, and say what the team has stopped doing as a result. "Just by giving away the tools, it's like a gym membership," a leader said. Others are building safe platforms for experimentation, or naming AI champions in each function whose projects must show business return.

There was caution too. One leader warned that unlimited enterprise access to AI models is expensive. "Once those token limits start coming," organisations will find out whether their automations really save money.

Accountability will remain human

Many tasks can be automated, one participant said, but the harder question is who is accountable when something goes wrong. Another put it bluntly. When work goes well, people credit AI. When it fails, "you cannot blame AI. It is you."

Giridhar linked this to entry-level hiring. Organisations that replace freshers with agents should ask who will be accountable when things go wrong, and where future leaders will come from. He called downsizing in response to AI short-sighted, noting that some organisations had already reversed such cuts. Beyond upskilling, he pointed to what the book calls "re-rolling", meaning roles that are deliberately redesigned. AI is also creating new work. Developers, for example, are increasingly expected to verify AI-generated output.

Governance is following. One leader shared that their CEO had asked HR to draft an AI agent governance and accountability policy. The reasoning was that HR manages people today and may soon manage people and agents together.

From access to proof: Learning from what didn't work

In one of the evening's most candid moments, Giridhar described Nuvepro's own path. First came a company-wide Copilot mandate, with usage dropping off after a couple of weeks. Then a wider coding-assistant rollout in which licences were disabled after the tool was used outside its intended scope. "One step forward, one step back," he said, before inviting the room to "touch grass" and share their own setbacks.

The examples were telling:


  • Credibility over polish. AI-built leadership presentations fell apart after a few follow-up questions. The message to teams became: if it is on your slide, you must be able to explain it. "Don't outsource your brains."

  • Same prompts, same answers. A hackathon produced near-identical solutions from different teams. "AI can be an enabler but not a decision maker," one leader concluded.

  • Technology without role redesign. One organisation automated screening and shortlisting, but recruiters carried on working as before and hiring managers had not been trained. The recruiter role had to be redefined before the system delivered.


Put five people in a room and they may agree 20% of the time. Give them the same AI tools and they agree 80% of the time, so the diversity of human thinking gets lost. The edge will come from what people build on top of models, the "harness", not from prompts alone.


That is why the conversation ended on practice and proof. The Nuvepro team described an approach that starts by mapping roles, tasks and workflows, identifying high-impact AI opportunities, and recreating the new work in simulated, safe environments. Employees can then practise and demonstrate their readiness through hands-on execution and assessment before proven workflows move into implementation. Moving into production also means passing governance and technology reviews, because "automation, somebody has to own it."

The takeaway

No one claimed to have a blueprint, but the room largely agreed on the order of things. Understand the work before choosing the tool. Keep accountability human as roles are redesigned. Give people safe places to practise. And treat readiness as something employees demonstrate, not something they complete. 


This is the essence of work transformation: redesigning workflows, evolving workforce capabilities and adapting work environments to enable a new way of working. As the closing remarks put it, workforce transformation in the age of AI "is not simply an HR, technology or learning agenda; it sits at the intersection of all three."