Managing people and AI agents together: A new leadership mandate

agentic AI, and 15 per cent of day-to-day work decisions will be made autonomously through agentic AI by 2028.

This signals a clear shift in leadership. As part of ‘s August editorial series, “Future leadership: Redefining leadership for an AI, human and business first world,” this story explores one of the most consequential shifts redefining leadership today: the emergence of AI as an active member of the workforce.

As intelligent agents increasingly work alongside employees, leadership is evolving from managing people alone to orchestrating human and machine capabilities together, raising new questions around accountability, judgment, organisational design, and the future of work.

India Inc has already embraced this shift. Tata Sons Chairman recently stated that the company will have as many AI agents and workers as its human workforce within the next three years.

During the earnings call, Rajeev Jain, Vice Chairman, Bajaj Finance, confirmed that their AI team is growing from 230 people to 400 people. The firm has deployed 27 AI bots across different functions and has already rolled out 17 AI applications for employees.

The company further plans to expand these to 118 AI applications over time, while Customer AI for business partners is expected to go live during the second quarter.

Managing AI is no longer a separate activity

The emerging reality is clear: AI will reshape leadership, shifting the focus to managing human and machine capabilities in tandem while unlocking meaningful organisational capacity.

Rajiv Naithani, Chief People Officer, Persistent Systems, doesn’t see it as a choice to make between managing technology and managing people. In his opinion, engaging with AI is no longer a separate activity competing with people leadership.

For Persistent Systems, introducing AI was not primarily a technology journey; it was a leadership journey. About 30 per cent of every employee’s performance assessment, including managers and HRs at the firm, is now centred on ‘AI Readiness and Impact.’ This is evaluated across three dimensions: Learn, Use, and Impact.

Similarly at the Export Trading Group, the AI’s impact varied across different layers of work. While 50 – 70 per cent of transactional tasks, such as employee queries and offer letter generation, have been automated, AI has enabled about 20 – 30 per cent of analytical work, including attrition analysis, workforce planning, and pay equity.

According to Prashant Pillai, Global HR Transformation Projects, analytics & AI – GCC & COE, Export Trading Group, an entirely new category of managerial work has appeared with AI evolution at work: Call agent oversight, or managing the digital bench.

An agent needs a different kind of supervision than a person. It doesn’t need coaching or motivation, but verification, setting up guardrails, and handling its exceptions. About 5-10 per cent of the managers’ KPIs at the firm are around the expected AI usage.

An interesting case was when the team deployed a global resume screening agent. The initial instructions were to shortlist a Grade Point Average (GPA) of 3.0. But as Indian universities have a Cumulative Grade Point Average (CGPA) or percentage system, candidates with CGPA above 4 or 50 per cent marks were also getting shortlisted. This taught the team that it is essential that the instructions are culturally aligned, and humans are in the loop to verify the data.

Beyond these transactional jobs, leaders collectively agreed that people continue to lead the conversations when it comes to decisions that shape careers and experiences, such as hiring, promotions, employee relations, terminations, leadership coaching, organisation design, or culture.

According to Avisek Dasgupta, CHRO, The Talent Care, the next phase of AI maturity will not be defined by the number of AI agents deployed. It will be defined by how much high-quality human capacity those agents genuinely release back into the organisation.

“The real dichotomy is not people versus AI. It is automation versus organisational capability. In reality, every automation first relocates work before it reduces work,” he said.

If an AI agent performed the work of two HR Operations executives but still required an experienced professional to review, challenge, and correct every output, the organisation hasn’t created additional capability. It has simply shifted cognitive load from execution to supervision.

When AI taught process discipline

The few tensions that Persistent Systems consciously navigated were speed versus judgment, scale versus empathy, and consistency versus clarity.

Interestingly, AI prompted the firm to become much more disciplined in documenting and governing the processes.

One instance that reinforced this learning came early in the implementation journey, when Naithani asked Pi-Assist (HR assistant) a simple question about the organisation’s promotion process. It responded immediately and very confidently, but the answer wasn’t entirely correct.

“The issue wasn’t the AI. It was the policy documentation, which left room for interpretation,” Naithani said.

This incident led the HR team to introduce a practice where process owners regularly challenge the agent with real employee questions and continuously improve its knowledge base.

“As a result, Pi-Assist consistently operates at around 99 per cent. It has reduced the routine HR email traffic by around 95 per cent. It handled an average of more than 33,000 employee queries every month, served over 8,000 employees, resolved 99 per cent of queries, and typically responded in less than 15 seconds,” Naithani said.

The shift to becoming product managers

Earlier at Vahan, managers spent a significant part of their week updating trackers, preparing reviews, coordinating interviews, or chasing follow-ups. What surprised Kartik Rao, Chief People Officer, Vahan, was how the team’s role evolved from thinking like ‘process owners to product managers.’

While the final decisions continue to rest with people, nearly 60 to 70 per cent of the operational HR workflows at Vahan are either AI-assisted or AI-executed at this stage.

Over the last few weeks, the HR team built three production-grade in-house HR products, including an ATS, an AI-powered employee lifecycle bot and a performance management system.

Initially, HRs approached AI as a coding assistant, but the breakthrough happened when it was treated as a coordinated team of specialised agents. Different AI agents owned development, testing, quality assurance and deployment, continuously validating each other’s work before anything reached production.

This dramatically changed the company’s economics.

“By building these systems ourselves on a lean technology stack, we reduced our annual HR technology cost for recruiting and performance management from nearly ₹48 lakh to less than ₹25,000, while retaining the capabilities we needed,” said Rao.

From technology discussion to organisational design

Experts argued that concerns around AI were not limited to just adoption, but extended deeply into psychological safety, accountability, and context.

Pillai pointed out that people constantly have a fear of going obsolete if the work gets replaced by agents. The mistrust that AI may miss the cultural context, like sensitivity to Indian workplace norms, is another apprehension.

Even Dasgupta felt similar ambiguity. While working on an AI-assisted game-based learning initiative, he was excited that AI could significantly reduce the effort required from instructional designers and e-learning developers. But the output proved otherwise.

While AI had produced content quickly, the instructional depth, learner engagement, and contextual relevance still required experienced professionals to rethink and rebuild large parts of the solution.

This raised a critical question: If AI generates the first draft, who owns the quality of the final learning experience?

“The person who built the AI is not accountable for the learning outcome. The instructional designers are expected to sign off on work that AI produced, while the manager remains accountable for both the AI agents and the human team,” Dasgupta said.

These experiences highlighted a larger shift. AI is no longer just a technology layer. It is an organisational design question, forcing leaders to rethink accountability, capability, and the evolving relationship between human expertise and machine intelligence.

Managing AI agents alongside people brings in new coordination challenges, new failure modes, and new expectations of oversight.

Leaders must learn to set boundaries for autonomous systems, calibrate trust in their outputs, and ensure that human judgment remains firmly in the loop where it matters most.

The task is no longer just aligning teams, but continuously shaping how humans and AI interact to produce reliable outcomes.