By Saurabh Sharma, Chief Operating Officer, ProHance.ai
Here is an uncomfortable thing to say to a room full of CIOs: most enterprises are deploying AI without any real understanding of how work happens inside their own operations today. The copilots get bought, the licences get counted, the pilots get announced, and then the returns refuse to show up. If that sounds familiar, you are in large company.
Every enterprise now has an AI strategy. Far fewer have an operating model to match it, and the gap is starting to show. Deloitte’s 2026 State of AI in the Enterprise study found that nearly half of organisations, 48 per cent, have introduced AI tools without redesigning the workflows or roles those tools sit inside, and only about one in eight, 12 per cent, have gone the whole way and rebuilt their operating model around AI at scale. The rest are running new technology through old organisational plumbing. The strain shows up as stalled pilots, murky ROI, and employees who were handed a copilot but never told what their job is supposed to look like now.
This is the blind spot. The tool stack is rarely the bottleneck. Visibility is.
Consider what “the workforce” even means in a large Indian enterprise or GCC today. Walk into a delivery floor in Bengaluru, Hyderabad, Pune or Gurugram and the people at the desks are only part of the picture. Alongside them sit outsourced teams in another city, contract staff hired for a single project, and a fast-growing population of software copilots and AI agents that draft, reconcile, respond and increasingly decide. Four kinds of worker now share the same processes: employees, vendors, copilots and agents. Most organisations have governance and measurement for one of them. Very few have it across all four.
India has more riding on getting this right than almost anyone. We are now home to more than 2,100 global capability centres employing around 2.36 million people, a base that has grown by nearly a third since 2021 and generates close to $100 billion in revenue, according to the latest NASSCOM-Zinnov numbers. These centres stopped being cost-arbitrage back offices long ago. They now run the finance close for Fortune 500 firms, the fraud checks on millions of card transactions, the revenue cycle for American hospital chains. When that much of the global economy’s operational core sits on Indian floors, whether we can actually see and govern the work is no longer a management nicety. It is a question of national competitiveness.
Yet most enterprises manage this reality with a patchwork of disconnected systems. Attendance lives in one tool, project status in another, vendor performance in a self-reported spreadsheet, and AI usage almost nowhere. Each tells a partial story; none of them talk to each other. Leaders see outcomes, the closed ticket, the shipped release, but rarely the friction behind them: the idle hours, the rework, the handoffs that stall between two vendors, the workload piling onto one team while another sits underused. Value leaks quietly through the gaps nobody is looking at. Add AI, and the gaps widen, because a co-pilot’s contribution shows up on no org chart and much of it is shadow AI that employees adopted on their own, outside any policy.
Which is why AI is not a procurement problem, and treating it like one is the first mistake. Buying the right model, the right copilot and the right automation guarantees nothing. The organisations pulling ahead are not the ones with the most sophisticated tool; they are the ones who can answer basic questions about their own operation. Who does what? Where is the effort actually going? Is capacity being lost quietly? Where does a tool genuinely remove friction, and where does it just add another layer of review that nobody asked for? That is what workforce intelligence really means. Not a dashboard, but the ability to connect effort to outcome across every kind of worker, and to tell the difference between a team that is busy and a team that is productive.
The people’s question turns out to be harder than the technology one. Microsoft’s 2026 Work Trend Index makes the point bluntly: organisational factors, culture, manager support and the way talent is deployed, account for more than twice as much of AI’s real impact as individual effort does, roughly 67 per cent against 32 per cent. Read that again, because it flips the usual AI narrative on its head. The model you license is the easy decision. The hard ones are how skill gaps get spotted and closed, how roles shift, and how managers are equipped to lead through the change. None of that can be steered without visibility into how work is actually changing on the ground.
The labour market adds urgency. Look at the roles that did not exist eighteen months ago: prompt specialists, AI adoption leads, forward-deployed engineers, workflow architects. AI is creating new jobs and reshaping old ones faster than annual workforce planning can track. Enterprises without real-time intelligence learn about these shifts in hindsight, in an audit, a budget review, or a resignation nobody saw coming because no one noticed a role had quietly become unsustainable months earlier. Enterprises that can see it happening reskill before a gap becomes a crisis, rebalance before burnout sets in, and redesign a role before it goes obsolete rather than after.
A word of caution, because in India the instinct is often to reach for monitoring. This is not about counting keystrokes or logging hours. That is the crudest possible response and it usually backfires, breeding mistrust and gaming. The prize is subtler: using the same quality of data that leaders already trust for finance and customers to decide where work should be redesigned, where people can be reskilled instead of replaced, and where AI is genuinely creating headroom for higher-value work rather than just more output to review. That is the real shift underway in serious AI-first organisations, workforce data moving from an HR reporting function to core operating infrastructure.
For India’s GCCs there is a specific prize. They have already climbed from executing tasks to owning outcomes. The next rung is owning the intelligence about how those outcomes are produced, and exporting that discipline back to global headquarters as a genuine differentiator. The centres that can prove, with evidence, that they run cleaner, better-governed, AI-augmented operations will win the mandates. Those still relying on gut feel and a monthly slide deck will slowly lose them.
The technology is already here, and it is not the hard part. The differentiator now is leadership: the willingness to see your own operation clearly enough to direct AI with intent, measure what it actually changes, and turn adoption into lasting value. AI will not fix an operation you cannot see. It will only make the blind spots more expensive.