Walk through most enterprises today and you will find AI everywhere. Copilots in email, assistants in the CRM, a chatbot on the intranet. Employees are genuinely faster. And yet when the CFO looks for the impact in the numbers, it is hard to find.
McKinsey's latest research puts the gap in plain terms. Eight in ten respondents say AI has improved their personal productivity. Only 37 percent report any positive effect on EBIT, and just 6 percent qualify as AI high performers. Bain estimates that as much as 90 percent of enterprises are still focused on deploying tools and narrow use cases, which produces what it calls micro productivity: real for the individual, invisible on the income statement.
Making each step of a broken process faster does not fix the process. It produces the same outcome, sooner.
01Why faster is not the same as better
When AI is added to an existing workflow, the workflow keeps its old shape. The same handoffs, the same approvals, the same queues between teams. Each person saves time, but the work still waits at every boundary. The time saved is absorbed by the system rather than released as value.
McKinsey also finds that coordination roles have grown roughly 1.5 to 2 times as fast as line roles over the past five years. Much of modern work is moving information between people. That is precisely the work agents now do well, but only if the process is rebuilt to let them do it.
02What the leaders do differently
The pattern across McKinsey, BCG, and Bain is consistent. Nearly three quarters of AI high performers substantially redesigned their workflows, compared with about one quarter of everyone else. Bain reports that enterprises running AI as a business transformation, rather than an IT experiment, saw EBITDA growth of 10 to 25 percent. BCG's rule of thumb is that only 10 percent of the effort belongs to algorithms and 20 percent to technology and data. The remaining 70 percent is people, organization, and process.
In practice, redesign means four things:
- Start from the outcome, not the org chart. Define the result the process exists to produce, such as a closed month, a resolved claim, or a fulfilled order, and design backward from it.
- Decide what moves to agents. Map which tasks agents take on, which handoffs disappear entirely, and where human judgment is essential. McKinsey calls this an orchestration chart.
- Set an ambition large enough to force change. Bain's observation is blunt: incremental goals produce incremental change. A target of 10 percent faster invites automation. A target of five times faster forces a rethink.
- Give one senior leader real authority. Redesign crosses functions. It needs an owner with control over funding and talent, not a committee.
03Where to begin
Do not start with a platform decision or a list of fifty use cases. Start with one process that matters to the P&L and that crosses at least two teams. Order to cash, the financial close, claims, procurement, and customer onboarding are typical candidates. Ask a simple question: if we were building this today, with agents available, what would it look like? Then measure the redesigned process against the old one in financial terms.
That single, visible result does more to unlock the next investment than any strategy deck. It is also how the 80 percent who feel more productive become part of the 37 percent who can prove it.