When technology outpaces the organisation

Imagine a finance team that introduces AI to prepare its monthly management reports. Commentary that previously took several hours to draft can now be produced in minutes. The technology works, the team is impressed and the initial results look promising. But the report still goes through the same review and approval process. Analysts check the AI-generated commentary line by line, managers repeat some of those checks because they are unsure how much they can trust the output, and the document moves through the same approval levels as before. 

The first draft is ready much sooner, yet the final report may not reach management significantly faster. 

This illustrates a problem facing organisations investing in AI. Technology can change the economics of individual tasks without necessarily improving the performance of the processes in which those tasks sit. Improving the entire process requires decisions about responsibilities, controls and ways of working that technology alone cannot make. 

When yesterday's successful model becomes tomorrow's constraint 

At the recent IV Biznes Polska Świat Congress at the Warsaw Stock Exchange, Nobel Prize-winning economist Philippe Aghion discussed creative destruction and the difficulty economies face when moving from adopting existing technologies to creating their own innovations. One of his arguments was particularly interesting: the structures and mechanisms that supported an earlier period of growth can become obstacles to the next one. Established players may have little incentive to challenge the arrangements that made them successful. 

Aghion was discussing economic growth, competition and market structures, but his argument raises a useful question at company level. What happens when the operating model that made an organisation efficient was designed around constraints that technology is now removing? 

For decades, businesses have invested in specialisation, standardisation and centralisation. They have divided work into functions, built shared services centres, introduced detailed procedures and created management structures that make responsibilities and controls predictable. These decisions delivered real benefits, allowing companies to manage growing volumes of work, improve consistency and control costs. AI is now beginning to change some of the assumptions behind those arrangements. 

Consider a process where data preparation, analysis, reporting and quality control are performed by different roles or teams. That division of labour may have made perfect sense when each activity required distinct technical skills and considerable manual effort. Now AI can support several of those activities within the same workflow. A company could simply introduce AI into the existing structure, allowing employees to perform their individual tasks more quickly while keeping every handover and approval point intact. Alternatively, it could reconsider how the entire process should operate, including which activities still require separate roles, where specialist judgement adds value and which controls remain essential. 

The second option is considerably more demanding because it affects more than technology. Buying a new tool rarely requires changing the responsibilities of several departments. Redesigning how work moves between them often does. It can affect budgets, reporting lines, job descriptions, performance measures and management authority. Decisions previously owned by one function may need to be made elsewhere, while some existing activities may no longer be necessary. 

This helps explain why a company can have technology capable of changing a process without yet having the organisational conditions required to change it. 

A successful AI task is not the same as a successful business process 

Another memorable presentation at the Congress came from Dr Konrad Maj, who discussed the rapid development of robotics in China. He described robots performing increasingly impressive demonstrations, from dancing to acrobatics, before asking a rather practical question: why would someone actually buy one? His closing message was simple: “People will pay for a problem solved, not for a backflip.” 

A robot's ability to perform a spectacular movement tells us something about its technical capabilities. It tells us much less about whether it can perform useful work reliably, safely and economically. The same distinction applies to AI. Generating a report, analysing a document or preparing a customer response may demonstrate that a technology can perform a task, but it does not establish whether using that technology improves the business process. 

Return to our finance reporting example. If AI produces the initial commentary in minutes, but the review and approval process remains unchanged, the overall improvement may be limited. The problem is not necessarily that the controls are excessive. Financial processes require appropriate checks, and AI does not remove that responsibility. The problem is that the organisation has not yet decided how those controls should work when part of the analysis is prepared by AI. 

Which outputs require detailed human verification, what evidence should accompany an AI-generated conclusion and who handles exceptions? These questions need to be resolved before the organisation can confidently change the workflow. Otherwise, employees may continue checking everything manually, sometimes duplicating work that the technology was supposed to reduce. 

There are also three different outcomes to consider: a task taking less time, the end-to-end process performing better and the organisation capturing an economic benefit from that improvement. Progress at one level does not automatically produce progress at the next. 

Time saved is not necessarily cost saved 

Suppose AI reduces the time spent on selected activities by 20%. That does not automatically mean the organisation can reduce employment costs by 20%. The tasks may be spread across ten different roles, and employees may still be needed to handle exceptions, make decisions or perform work outside the automated scope. Some of the released capacity may also be absorbed by new quality checks or more complex activities. 

The organisation therefore needs to decide what to do with the capacity created. It could handle higher volumes without increasing headcount proportionately, improve the quality of analysis, take on additional responsibilities or redirect people towards work that previously lacked sufficient resources. In some circumstances, redesigning roles and team structures may also create opportunities for lasting cost reductions. 

These are different business choices, each requiring a different operating response. AI may release employee capacity, but it does not decide how the business should use it. Without an explicit decision, improvements in task productivity can remain difficult to identify in financial or operational performance. 

Who makes the decisions when AI performs part of the work? 

In another Congress discussion, Professor Aleksander Mądry spoke with Jarosław Sroka about the development of advanced AI and the challenge of keeping increasingly capable systems aligned with human intentions. One of his points has direct relevance to business management: even a highly capable system needs an appropriately defined objective. 

Telling AI to improve company performance is not enough. Does improvement mean reducing costs, increasing sales, improving customer experience or lowering risk? Which trade-offs are acceptable? These choices require human judgement, and the same challenge appears in everyday operations. 

Traditionally, a manager organised work across people, set priorities, reviewed quality and intervened when decisions were needed. With AI, that responsibility becomes more complex. Managers must also determine how work is allocated between people, AI and other systems. They need to decide which activities AI may execute, where it should only prepare a recommendation, when human approval is necessary and who remains accountable for the outcome. 

There is another consequence that deserves attention. As AI takes over more routine activities, the work remaining with employees may change in character. People may spend less time on repetitive preparation and more time dealing with exceptions, interpreting ambiguous situations and exercising judgement. A team could therefore perform fewer individual tasks while facing more demanding work. That changes the skills required, the way workloads should be managed and potentially how performance should be measured. 

The manager's role cannot be reduced to approving whatever AI produces. Human oversight needs to be meaningful, and that requires the right information, expertise and authority. If a manager cannot properly challenge an AI-generated outcome or intervene when necessary, assigning formal accountability achieves very little. 

Organisations therefore need to design decision rights, review mechanisms and escalation paths as part of the workflow, rather than adding them after the technology has been deployed. Management routines may also need to change. The information managers review, the exceptions requiring their attention and the way they evaluate quality and capacity will increasingly depend on how work is divided between people and technology. 

These are operating-model questions, not simply technology-adoption questions. 

New capabilities do not always require new hires 

Skills shortages were another recurring theme at the Congress. Executives involved in Poland's major infrastructure investments discussed growing competition for engineers, technical specialists and experienced project professionals. In a separate discussion about the future of work, Dr Joanna Frączek-Broda challenged the way organisations approach professional development later in people's careers. 

Companies frequently invest heavily in employees during their early years, then gradually shift towards expecting them to rely on the expertise they have already acquired. Yet someone in their forties or fifties may still have fifteen or twenty years of professional life ahead. There is little reason to assume that their development is largely complete, particularly when technology is changing the content of existing jobs. 

This also points to a problem in how organisations define their capability needs. When a new technology or business requirement appears, the instinctive response is often to create a new position and start recruiting. Sometimes that is exactly right. A business may need specialist knowledge that does not exist internally and cannot be developed quickly enough to meet an immediate requirement. But a new task does not always justify a new permanent role. It may require changing the scope of an existing position, developing the current team or bringing in specialist expertise for a defined period. 

Consider the relationship between process knowledge and AI expertise. An experienced finance professional understands business rules, recurring exceptions, system dependencies and the consequences of errors. Much of that knowledge may never have been fully captured in process documentation. An AI specialist understands the technology but may not know why a particular process operates as it does. The manager, meanwhile, understands business priorities, service expectations and the decisions that must remain under human control. A workable solution often depends on bringing these perspectives together. 

This is one reason why developing existing employees can be as important as recruiting new expertise. Experienced professionals do not necessarily need to become software engineers or data scientists to contribute to AI-enabled work. Their process knowledge may become more valuable when combined with an understanding of what the technology can and cannot reliably do. 

At the same time, reskilling has limits. Not every new capability is a natural extension of an existing role, and not every business can afford to wait while critical expertise is developed internally. The more useful question is therefore not whether to hire or reskill, but which capabilities need to be acquired, which can be developed and how the two should work together. 

In many situations, combining external expertise with internal development is the most practical answer. A specialist can help close an immediate gap and support delivery while existing employees contribute business knowledge and build new skills through the work itself. For that to create lasting value, knowledge transfer needs to be part of the engagement from the beginning rather than something addressed shortly before the assignment ends. 

The same applies to employee development. Providing training does not mean the organisation has created an opportunity to use the new skills. If people are expected to learn new technologies, contribute to improvement projects and redesign processes while carrying exactly the same workload, development can remain an aspiration rather than a practical reality. 

A company cannot expect continuous adaptation without making room for it. 

Who owns the redesign of work? 

There is no shortage of discussion about AI models, data quality, security and the cost of technology. The decisions about how work itself should change are often harder, particularly when responsibilities are spread across several functions. 

Someone needs to determine which activities should remain with people, how responsibilities will shift, which capabilities the organisation must develop and what happens to capacity released through automation. These decisions do not sit neatly within a single department. IT can assess technology and integration requirements, HR can support workforce planning and capability development, Finance can evaluate investment economics, and operational leaders understand the process and are accountable for its performance. Yet none of these perspectives can solve the problem independently. 

The challenge lies in connecting them and establishing who has the authority to make the necessary decisions. 

At Helprise, we approach these questions from the perspective of people who have built, transformed and run business operations. That experience shapes how we work with clients, whether they need specialist practitioners, additional capabilities, process redesign or support in implementing a new operating model. 

Sometimes the right answer is to bring in an expert who can address a specific capability gap. Sometimes it is to develop the existing team, simplify a workflow or change the allocation of responsibilities. Other situations require a broader redesign of the operating model and support through implementation. The starting point should be the work that needs to change and the business outcome expected from it. 

Consider our finance reporting example one final time. If AI can prepare management commentary in one hour rather than four, but the final report still takes several days to approve, investing in a faster model may not solve the remaining problem. The business may need to change its review process, clarify accountability or reorganise how the work moves between teams. And if the new process genuinely releases employee capacity, management still needs to decide how to use it. 

Those decisions determine whether a company simply performs yesterday's work faster or builds a better way of operating.