Digitising an old process is not a transformation, but rather its perpetuation within a system — often with greater speed, scale and at the expense of future change.
For years, companies have been transferring procedures designed for paper, email and organisations divided into functional silos into systems. A form becomes an application, a signature becomes a button, and the manual handover of a document becomes an automated workflow. Technology speeds up the operation, but its underlying logic remains unchanged.
As a result, redundant approvals, multiple checks, data re-entry and the division of responsibility that arose a decade earlier are also digitised. The Harvard Business Review describes this problem as ‘process debt’ — a debt arising from the accumulation of outdated working practices that are disconnected from the customer. A new system can reduce such debt. It can also lock it in for years to come.
AI has laid bare the problem
Artificial intelligence highlights this difference exceptionally clearly, as it can be implemented much more quickly than reorganising an organisation.
According to Deloitte, 48 per cent of the organisations surveyed introduced AI without redesigning the processes or roles in which the technology is to operate. Only 12 per cent made such a change on a larger scale alongside a new operating model.
This is the fundamental difference between automating tasks and changing the way a company operates. A system may be able to produce an analysis in a minute rather than an hour, but the benefit quickly disappears if the document still has to go through four levels of approval. An agent may automatically handle a customer enquiry, but this makes little difference if the process requires data scattered across several systems and the intervention of a different department each time.
McKinsey analysed 25 characteristics of organisations using generative AI. Of these, workflow redesign was the factor most strongly linked to the reported impact of the technology on EBIT. At the same time, only 21 per cent of companies using generative AI reported a fundamental overhaul of at least some of their processes.
The scale of investment therefore contrasts ever more sharply with the scale of change. According to the World Economic Forum, over $250 billion was allocated globally to AI in 2025, whilst only 25 per cent of companies rated the impact of this technology as transformative. The WEF also highlights a fundamental problem: many companies are simply adding AI to existing processes rather than designing their operations around the new possibilities offered by the technology.
Productivity gains from individual tasks are not enough
The difference is also beginning to show in financial results. A 2026 PwC survey of 1,217 managers indicates that around 20 per cent of companies are realising three-quarters of the declared economic benefits of AI. Organisations in this group are twice as likely to redesign their workflows for AI rather than simply adding further tools to their existing way of working.
BCG reaches a similar conclusion based on its own implementations of agent-based AI. Projects involving end-to-end process re-engineering achieved cost reductions of up to 60 per cent, whilst piecemeal automation of individual stages yielded less than 20 per cent.
This is not about automating everything from start to finish. It is about questioning the very structure of the process before digitising it.
This is clearly evident in manufacturing. The Hindustan Unilever plant in Tinsukia did not limit itself to introducing a single application. Over 50 technological applications covered planning, production changeovers and packaging testing. The production plan freeze period was reduced from 14 days to one, the number of product variants supported tripled, and the time taken to test new packaging fell by 84 per cent. It was the architecture of the process that changed, not just its interface.
A digital process can be just as bad as a paper-based one
In 2026, the problem becomes more costly, as AI shifts automation from individual tasks to entire workflows. Deloitte reports that only 5 per cent of organisations consider their processes to be very well prepared for the use of AI agents, although 74 per cent of managers expect that within four years, almost half of business processes will be rebuilt around them.
This changes the economics of transformation. The greater the proportion of the old process that a company automates, the more technology, integration and control will come to depend on its current form. Subsequent restructuring will be more difficult, not easier.
The most costly mistake in digitalisation, therefore, need not be a poorly chosen system. It may well be a perfectly functioning system that perpetuates a process which is unnecessary in its current form.
