AI at Work: Why workflow redesign matters more than tool adoption

AI has moved from experimentation to execution, but many organisations are still grappling with a key question now – is it actually changing how work gets done?

For HR and business leaders, the next phase of AI is less about how many tools employees use and more about whether those tools improve productivity, sharpen decision-making and free people to focus on higher-value work.

That was the focus of a leadership conversation featuring Premita Singh, CHRO, Havells; Varun Gautam, Senior Corporate VP and Global Head, Talent Management, Persistent Systems; and Rajani Athreya, Senior Director – HR, South Asia, Kantar, moderated by Asha Krishnan, Vice President, General Atlantic.

The discussion highlighted a growing consensus among business and HR leaders: AI is neither something organisations can simply opt out of nor a technology that should be layered on top of existing processes without rethinking them. Its real value lies in redesigning work around what humans do best while allowing intelligent systems to take on tasks where they can add greater speed, scale and precision.

One of the biggest shifts is that AI's impact is no longer confined to technology or knowledge-based roles.

At Havells, AI is being viewed not as a technology transformation but as a business transformation. With a workforce of around 40,000 spanning manufacturing, sales, corporate functions and other roles, the company is looking at AI across the value chain.

The starting point is the business problem rather than the technology itself. As Premita explained, the organisation asks what friction customers are experiencing, how consumer experiences can be improved and how AI can help the business become more proactive, serve customers better and take a longer-term view.

That approach has led to AI applications across manufacturing, sales, distribution and R&D.

Havells has 16 manufacturing plants, most of which operate on Industry 4.0 principles. AI and digital technologies are being layered into these operations to enable predictive maintenance, identify potential machine failures before they happen and improve quality through vision analytics.

The company is also using digital twins to simulate processes and assess potential product quality before an actual product is created. The impact extends to the company's retail network. Retailers can use applications to assess inventory and request materials based on data and AI-driven insights rather than relying entirely on intuition or historical patterns.

In R&D, Havells has also created digital personas and agents that can support engineers in different roles, including leadership, testing and quality.

Even HR is being used as a testing ground for a different approach to learning. Premita cited an AI-enabled open-book examination in which students are allowed to use AI while solving problems. The objective is not to reward memorisation, but to identify those who can use technology while demonstrating stronger problem-solving ability.

The discussion also challenged the assumption that AI-driven productivity necessarily means reducing headcount.

At Havells, the focus is instead on using AI to increase revenue generation and deepen engagement with the ecosystem. By enabling the same people to spend less time on routine work, the organisation can create capacity for higher-value activities and potentially expand the top line.

A similar shift is visible in General Atlantic's portfolio companies. Asha shared the example of PhonePe, where AI is being used to improve the productivity of field teams involved in merchant acquisition. Instead of managers manually determining how field employees should structure their daily routes, AI can help define their beats and make their activities more targeted.

The longer-term implication is significant for managers. If AI takes over more administrative and planning work, managers can spend more time coaching and supporting their teams.

That could eventually change managerial spans of control, allowing leaders to manage larger teams without sacrificing the quality of leadership.

The important question, therefore, is not simply “How many people can AI replace?” but “What higher-value work can people do when AI removes some of the work around them?”

For Rajani, this distinction is critical. AI tools can improve individual productivity. An employee may use AI to draft an email, summarise a document, conduct research or complete a task faster. But individual productivity gains do not automatically translate into enterprise-level productivity.

That requires organisations to rethink the workflow itself. At Kantar, the approach is to fundamentally redesign processes so that AI can be embedded naturally into them rather than simply adding an AI tool to an existing process.

Kantar, a marketing, insights and analytics company, has around 4,000 people in India. As a data- and knowledge-intensive organisation, it is increasingly looking at where agents can take on parts of workflows while humans retain responsibility for judgement, relationships and trust.

This has even influenced how the organisation thinks about its own workforce.

Rajani described a cultural shift where HR is no longer simply a “people function”. It is increasingly viewed as a people and agents function, reflecting the reality that agentic systems are becoming part of how teams operate.

At Kantar, employees are encouraged to think of their teams differently. A team could consist of two human employees and three AI agents.

The point is not the number of agents, but the mindset behind it: the agentic workforce is no longer a future concept. It is becoming part of the present workplace.

Kantar's approach is not to use AI simply to optimize existing processes. The focus is on identifying where human judgement matters most and redesigning workflows around it.

For example, Kantar is using agents to help employees prepare for client pitches, review briefs and handle parts of the work that can be supported by intelligent systems.

That gives employees more time and mental bandwidth to focus on areas where humans remain critical, judgement, discernment, relationships and trust.

The same principle was illustrated through another General Atlantic portfolio company, KFintech. The company had a team of around 50 people handling highly manual processes related to areas such as share transfers and KYC. These employees had built up significant experience and could intuitively assess requests, categorise them and determine what needed to happen next.

Rather than simply placing an AI tool on top of that existing workflow, the team broke down the process to identify the specific points where judgement was required.

Agents were then designed around those individual “slivers of judgement”, creating a six-agent orchestrator model.

The result was a fundamental redesign of the workflow rather than a technology layer added to it. According to the discussion, the model helped increase turnaround time threefold, reduce the number of people involved by half and maintain accuracy at 96%.

The lesson is important: AI impact comes from reimagining how work flows, not simply adding AI to the work that already exists.

But redesigned workflows need people who can work effectively within them. For Persistent Systems, where the business is highly dependent on technology and people capability, establishing an AI baseline cannot follow a one-size-fits-all approach.

The appropriate level of AI intervention depends on the customer, engagement, scope of work, technology environment and engineering skills involved. For Varun, building AI capability therefore has to happen at both the individual and enterprise levels.

At Persistent Systems, certifications – whether externally recognised or developed through internal academies – can help establish whether employees have developed the required capabilities.

Varun noted that a significant part of his role in talent management is focused on learning and on establishing capability baselines across different industries and business segments.

He said organisations cannot expect employees to adapt to AI simply because the technology has been made available to them. Capability needs to be deliberately built.

So what makes someone more likely to succeed in an AI-enabled workplace? Varun identified adaptability and curiosity as two critical qualities.

Being able to use an AI tool is not enough. Employees need to be willing to change how they work, experiment with new systems and continuously learn.

That also needs to be combined with human capabilities such as empathy and the ability to remain in the loop when decisions require human judgement. The need for these capabilities will vary across industries.

Technology companies may be more comfortable experimenting rapidly with AI, while sectors such as healthcare may take a more cautious approach because of the sensitivity of the data involved. Banking can also vary significantly depending on the business model and use case.

The technology may be the same, but the tolerance for risk, the regulatory environment and the consequences of mistakes can be very different.

The discussion ultimately moved the AI conversation away from a simple human-versus-machine narrative.

The emerging workplace is likely to be one where humans and intelligent systems work together, but not necessarily on the same tasks.

AI can take on repetitive, administrative and data-heavy activities. Agents can support workflows and accelerate analysis. Humans can then spend more time on judgement, problem-solving, relationships, coaching and decisions that require context and empathy.

For HR, this means the role is changing as well. HR leaders will have to help organisations identify where AI can genuinely create value, redesign work around it, build the capabilities employees need and manage the cultural shift that follows.

That requires more than AI literacy. It requires organisations to rethink jobs, workflows, teams and even what they mean by productivity.

The key takeaway from the discussion is that simply giving employees access to AI will not transform an organisation.

The real opportunity begins when leaders ask deeper questions: What work should AI do? What work should humans continue to own? Where does judgement matter? Which workflows need to be redesigned? And what capabilities will people need to succeed in this new model?

For Havells, that means applying AI across factories, retail, R&D and HR to solve real business problems.

For Kantar, it means redesigning workflows so agents become part of teams while humans focus on judgement, relationships and trust.

For Persistent Systems, it means making learning and capability-building the backbone of AI readiness.

Across these examples, one principle stands out: AI impact is not created by technology alone. It comes from the intersection of technology, redesigned work and human capability.