Over the past eighteen months, almost every leader has issued the same instruction: get employees to start using AI. Most of these instructions produce the same outcome. Usage is there, but demonstrable value has yet to follow. According to Harvard Business Review, only 1 in 3 organisations actually sees a return on its AI investment (Harvard Business Review, 2026).
This is not a matter of poor execution. It is the product of an approach that held true for years: buy the software, make sure people can work with it, and the returns follow automatically. With AI, individual adoption and value have become decoupled. Because the barrier to entry is so low, usage builds on its own, but the returns come from the process around it. As long as an employee banks that time saving inside a process chain that has not been redesigned, the gain stays with the employee, and more adoption yields no value, only a higher bill (also read: The AI Budget Paradox: Why Traditional IT Budgeting Fails for AI).
The question is no longer whether AI matters. That debate has been won. The consensus now lies one level down, in the belief that the road to value runs through adoption. Many organisations reason as follows: roll out licences, train employees, remove barriers, measure usage, and value follows. This sequence worked for decades with the enterprise software we know. In an ERP system or a CRM, the process sat inside the software, so implementing the software forced the process redesign. Underuse by employees was then the main constraint.
So when the redesign still came bundled with the rollout, managing for usage was enough. With AI, that forced redesign step is missing. The initiative to redesign the organisation and its processes now sits with the organisation itself, and so the step often gets skipped. Research by the Census Bureau finds that in more than a third of companies where employees use AI, there is no formal company-level adoption at all (Census, 2026).
“Firms that are operationalising AI are pulling ahead. Those that aren’t are starting to take on real risk, across talent, clients, and financial performance.”
Steve Hasker, President and CEO of Thomson Reuters
What the employee gains, the organisation fails to capture.
The evidence comes from Denmark, where researchers from the University of Chicago studied 25,000 workers across 7,000 employers. The study reports an average time saving of 2.8 per cent from AI use (Humlum & Vestergaard, 2025). By contrast, in controlled experiments, the same professions achieve time savings of fifteen per cent or more on individual tasks. So where does that gain go?
Time saved within a process is not a financial gain if it evaporates. The real ROI emerges only once all these small savings add up to something the organisation or the process can actually use: a task that is eliminated, a workflow step that disappears, or an extra service the organisation can deliver. That takes a change in the process itself. When that change fails to materialise, the gain stays with the individual, spent on one more email or a meeting that ran over.
That is why the entire organisation needs to be rethought from the ground up. Speeding up one step for one individual does nothing for the throughput of the entire workflow. The claims handler who drafts twice as fast produces no value, only a longer queue at the assessor's desk. Nothing is actually faster. The waiting has simply moved.
T-Mobile shows how it can be done. The company strengthens its customer service not with a chatbot that lets employees search faster, but by deploying AI that pulls real-time data and context at the first point of contact. This takes away the look-up work while preserving the personal touch. It lets T-Mobile give customers the answer they need on the very first contact.
The core idea behind T-Mobile's process redesign is that the human and AI workforces reinforce each other. Employees focus on what they do best—namely human contact, empathy, and problem-solving—while AI supplies information at a speed no human could match. Since then, the company reports a 60 per cent higher NPS and a 39 per cent reduction in churn.
The question is therefore no longer adoption, but how you organise the company so that those substantial time savings pay off at an organisational level. The key lies in redesigning the organisation, and with it the process.
