Artificial Intelligence

AI speeds up work but does not always improve it. Here are the critical points

Although AI increases work speed, much of the time saved is spent on corrections, highlighting the need for organisational and training change

by Gianni Rusconi

27 April 2026

5' min read

Translated by AI

Versione italiana

5' min read

Translated by AI

Versione italiana

Four out of ten Italian workers claim to save up to one day a week thanks to artificial intelligence. Yet, behind this seemingly unequivocal data hides a paradox: a significant share of that time (about half) is reabsorbed by activities of correction, verification and rewriting of imperfect output. This is what emerges from Workday's global research 'Beyond Productivity: Measuring the Real Value of AI', presented a few weeks ago at the opening of the Californian multinational's Innovation Lab in Milan. The trends to reflect on are well expressed by two percentages and are as follows: while 92% of employees claim to be more productive thanks to AI, up to 40% of the time saved is, however, spent on rework, with one employee in two spending between one and two hours a week correcting results generated by algorithms deemed to be of inadequate quality. In other words, we are in the presence of a kind of 'apparent productivity', in which the speed of task execution increases, but does not always translate into a real improvement in the work process.

From enthusiasm to maturity of use

Italia is still at an early stage in the AI adoption cycle, and this is confirmed by the fact that only 29% of workers use this technology on a daily basis, although the perception of its benefits is generally very high. However, according to Fabrizio Rotondi, Country Manager of Workday Italia, a cautious reading of these indicators is necessary. "Comparing Italian data with global and US data, we can say that in our country we are still in the hype phase, while in the US there is already a higher level of scepticism towards artificial intelligence. In Italia,' the manager continues, 'we are mainly intercepting the most enthusiastic users, often young people, who have not yet developed a full awareness in the management of productivity increased by technology'.

This imbalance is bound to diminish over time, as adoption becomes more mature, but it brings with it a first critical issue: the widespread use of AI does not automatically coincide with an effective use of these tools. In fact, it is always the quality of the data that determines its value, so much so that one of the main risks associated with the massive adoption of generative models is what Rotondi calls 'shallow AI', i.e. systems that generate fast but insufficiently reliable outputs. If many projects fail, this is Workday's conviction, it is because they work on unclean or unstructured databases; it is no coincidence that one of the American company's 'musts' is to avoid this phenomenon, and the recipe for doing so is to resort to a cloud-based database built on the platform's more than 11,000 customers and the approximately 1,400 billion transactions managed each year. Without a solid database, ultimately, the risk that companies run is that of multiplying productivity without increasing accuracy, highlighting all the limitations (and the problem of reprocessing is obviously among them) of an incomplete integration between technology, processes and information.

Organisation does not keep pace with technology

Another key element of the issue concerns the misalignment between tools and organisational models. In most companies, roles and processes have not evolved at the same speed as technologies, and this is made clear by the comparison that sees employees today using tools released on the market in 2025 within organisational structures anchored to 2015 screens. A dichotomy that only increases the cognitive load and individual responsibility of people, who are forced despite themselves to more outputs to manage and more controls to carry out.

The subjects most exposed to the rework phenomenon are in fact the youngest workers (and theoretically the most prepared), while at the same time the most assiduous users turn out to be also those under the greatest pressure, with over 77 per cent of the sample of workers interviewed declaring that they check the results produced by AI with the same, if not greater, attention than human work. "The real issue," Rotondi pointed out in this regard, "is the change management. Companies often fail to grasp the opportunities of AI due to inertia or the difficulty of rethinking models and processes, and this confirms that the turning point is organisational before being technological'.

Agents arrive, need to reinvest in people

The discussion on the impacts of AI in the workplace must necessarily refer to an evolutionary scenario that goes beyond the use of individual tools and points towards a deeper integration between people and intelligent systems (agents). "We see a future in which artificial intelligence will be the new interface for business processes, with agents embedded in the body within the organisation. The central point,' Rotondi adds, 'will be the governance of these assistants, and therefore how and when to activate them, in which areas and to what level to delegate activities and tasks to them'. The model that is emerging is precisely that of structured collaboration between humans and digital agents and faithful to the 'human in the lead' principle, in which the person not only acts as supervisor but has full responsibility for managing processes and decisions.

However, this evolutionary shift in employees (and the managers who lead them) also implies a change in the skills required: skills are shifting from specialised content to transversal abilities such as leadership, critical thinking, relationships and decision-making. In this context, even human fallibility takes on a new value and, according to Workday's Country Manager, "can become an asset, especially in terms of strengthening empathy and engagement skills".

A managerial challenge before a technological one

Returning instead to perhaps the most important summary of the research, that which quantifies the time 'saved' through the use of AI, it is interesting to note that the difference between the companies that achieve results and those that lag behind is not so much the degree of adoption of the technology per se, but precisely the way in which employees spend this time. In this sense, it turns out that the most advanced organisations reinvest these 'savings' in developing skills, reviewing corporate roles and improving decision-making processes; in contrast, many companies continue to allocate resources primarily to technology or increasing workload.

"The great opportunity we have," Rotondi concludes, "is to help people use AI effectively, especially in activities that require critical thinking and creativity. The reference goes directly to the new Innovation Lab in Milan, a space for co-creation between companies, partners and experts to develop concrete use cases where design thinking is done and prototypes of intelligent agents are built to be tested in the field. The challenge to be overcome (for Workday and the thousands of Italian companies that are already experimenting with AI) is that of knowing how to govern its use through a parallel evolution of processes, skills and organisational models: the risk of generating mere partial efficiency is real and the task of management is instead to transform the benefits induced by technology into quality of work, decision-making capacity and competitive advantage.