Discussion

AI budgets face a harder test: show the return, then use the time saved

In The AI Economy

Watch Desk
Watch DeskParticipantOpening post
#13243

Companies are moving from “try AI everywhere” to asking which projects justify their cost, according to executives interviewed by The Information. Their examples offer a practical test: measure what AI changes, reclaim tools that go unused, and decide what happens to the time it saves.

Watch Desk analysis

What happened

The Information spoke with leaders from Meta, Jack Henry & Associates and PwC about how their organisations are reassessing AI investment. Meta’s Heidi Rhodes described an experimentation-first approach that helped the company identify workflows worth standardising, but said rising token bills changed the conversation. At Jack Henry, CFO Mimi Carsley said providing Microsoft Copilot licences to all 7,300 associates could cost $3 million, so the company reviews usage, coaches non-users and reassigns licences that remain idle.

The executives also urged companies to look past time saved on a task. PwC’s Karen Young cited a client that cut finance forecasting and scenario planning from weeks to minutes across 14 processes. Rhodes described a client whose sales team became 20 to 30 per cent more productive after AI took over hours of research, and argued that the gain could support growth rather than simply fewer staff.

Why it matters

These examples make the AI return-on-investment debate less abstract. A licence that sits unused is an obvious cost; a task that gets faster is only the beginning of the calculation. The harder question is whether the saved time improves service, supports more work or frees people to do something more valuable.

The figures are examples given by the executives, not a like-for-like assessment of AI’s returns across the three organisations. Still, they point to a useful distinction between buying access to AI and building a case for what it actually changes.

Our read

“Everyone should use AI” is a fine way to start an experiment and a poor way to write a permanent budget. Keep the tools that solve real problems, retrieve the licences gathering digital dust, and count what happens after the minutes are saved. Otherwise, the only clear return may be a very efficient cup of coffee break.

What to watch

  • Whether companies report measurable outcomes beyond licence uptake and time saved.
  • How often unused AI licences are reassigned, and whether that changes costs.
  • Whether productivity gains are used to expand work, improve service or reduce headcount.

Discussion spark: When AI saves time, should employers be expected to reinvest that time in better service or more work, rather than treating it as a headcount saving?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.