AI changes the tasks first, the titles last

When AI enters a function, the sequence is almost always the same. First the tasks change: what the person spends their day doing quietly re-composes around what the model now handles. A reconciliation analyst spends less time reconciling and more time investigating exceptions. The model carries the volume; the human carries the judgment. Same seat, different work, and no org chart has recorded any of it yet.

Then, much later, the title changes to describe the new reality. In between there’s a stretch where the job on the chart and the job being done have drifted apart, and nobody has written it down. That stretch is structural. Headcount plans are built annually and defended quarterly; task composition shifts continuously. So for 2 or 3 quarters, the real work of a function can run well ahead of the structure meant to describe it. People are doing new jobs under old names.

An org chart is a record of decisions that have already been ratified. A new box appears when a manager has argued the case, secured the budget line, and won the approval, a process that runs on the annual planning clock, not the pace at which the work is actually changing. So the chart is a lagging indicator by design. It’s where a change goes to be confirmed.

The pressure behind this lag isn’t abstract. India’s Net Employment Outlook was +68% for Q2 2026, the strongest reading since 2008, yet 82% of employers reported difficulty finding the skills they need, according to ManpowerGroup’s Employment Outlook Survey. When intent is that high and supply is that tight, the tasks inside a role re-compose fast, because managers reach for whatever capability is closest rather than waiting for a headcount cycle to bless a new hire. The work moves under the pressure. The chart waits for the paperwork.

This is why a role can look identical on the org chart while the job inside it has quietly become something else. The label is stable, so the drift is invisible to anyone reading the structure instead of the work. And the people most likely to read the structure, the parent organisation, the finance partner, the workforce planner working off last year’s template, are exactly the ones setting the reqs.

The danger in the lag is that it hides an unmet need. When the investigation-and-judgment layer of a workflow grows and no one is hired to own it, the work doesn’t disappear. It gets absorbed by whoever is closest and most senior, until they are stretched past the point of doing either job well. In regulated environments, BFSI GCCs especially, that’s exactly where quality slips show up: a control layer thinning out at the same moment the volume through the model is rising.

So the cost of a lagging org chart is a senior person quietly carrying two roles, a governance function budgeted as if the productivity story never created new work, and a mis-hire waiting to happen on the next req. Because the title still carries the old skills list, you keep sourcing the profile that fit the job as it was. The candidate arrives competent for a role that no longer quite exists and short on the judgment the changed work actually needs. The mismatch doesn’t surface until they are in the seat and the gap is expensive to unwind.

The way past it is unglamorous but reliable: look at the actual composition of the work before you write the req. What in this role is now assisted by a model, and what does that free the person to spend their time on instead? Usually the routine share shrinks and the judgment share grows, which means the profile you want is more senior in instinct than the title implies, even though the title hasn’t moved.

This is a habit more than a technique. The centres that do it well interrogate the work behind every req in an AI-affected function rather than trusting the title to still mean what it meant a year ago. They watch what is automating and what is intensifying inside their own teams, and they treat that as the real hiring signal. It costs a conversation up front and saves a mis-hire that’s expensive to unwind later. Do it consistently and you also catch the thing that pure headcount planning misses: the need that has grown inside an existing role without ever earning its own box.

Everyone eventually updates the title. The advantage belongs to whoever reads the task change before the title does and hires against the real job first. The market hires when the box appears; the sharpest centres hire two quarters earlier, when the work changed but the chart hadn’t caught up. That head start is often the difference between staffing the change and scrambling after it.

The discipline is simple to state and easy to skip. Before the next req in an AI-touched role, ask what the work has become. Watch the composition of the job, budget for the judgment layer the productivity is creating, and hire into the emerging need before it has been formally recognised. Get that right and your team is fit for the function as it actually operates rather than as the chart describes it, which, in a function AI is reshaping, is where you want to be.