Lead­ers ex­tend hope when AI pro­jects fail to de­liv­er on prom­ises

Once a month, researchers from CBS write a column in Børsen, where they give readers a current and research-based perspective on the challenges managers face. 

In this column, Ursula Plesner and Frank Meier, Associate Professors at the CBS Department of Organisation focus on what leaders do when new, hyped technology doesn't achieve the expected results.

With the spectacular recent advances in Artificial Intelligence (AI), hopes and promises abound in organisations. This raises the question of how leaders can work more consciously with the mechanisms behind promises and hopes. We can learn a lot from looking at AI-projects if we want to understand how promises function in an organisational context.

Many promises are made at a time when AI is expected to solve problems that businesses have been struggling with for decades. The hope is that AI can help address the climate crisis, the recruitment crisis, and other global challenges on a large scale. Simultaneously, there are hopes that new intelligent technologies can enhance organisational efficiency and the quality of products and services. According to consulting firm McKinsey, up to 60-70% of all work tasks can be automated with the help of AI and other digital technologies.

Here, hope becomes the expectation of fulfilment attached to promises. With numerous and accelerating promises, it is not surprising if some cannot be fulfilled or are only met in a modified form, with changed or reduced ambitions.

This resonates with what Gartner – another influential consultancy firm – describes with their so-called hype cycle, a model that illustrates the standard progression that new technologies undergo: At the tail end of the first innovation stage, expectations rise to sky-high levels and reach a peak of inflated expectations. Then comes disappointment when it turns out that these expectations cannot be met. This is the 'trough of disillusionment' in Gartner's terminology. After that, there is a slight rise, where actual technological progress is observed, and finally, a realistic level is reached where the technologies prove useful in practice.

Gartner's reports have repeatedly mapped promising technologies onto this model, which often provides an excellent overview of how a particular technology has experienced hype at one point, only to reach a more realistic level and transition into everyday use. It goes without saying that generative AI is currently at its peak of inflated expectations, according to Gartner.

However, these macro analyses do not provide insights into how leaders handle the relationship between ambitious promises and their altered, delayed, scaled-down, or even unfulfilled delivery on a day-to-day basis. But managing promises is an important leadership task, so let's delve into how leaders do that in their daily practice. We've studied this in Danish organisations experimenting with AI.

AI is characterised by being a multifaceted entity, encompassing both spectacular generative AI like ChatGPT or DALL-E, which can generate text and images in seconds, and more subtle machine learning that aids in pattern recognition in vast amounts of data. The latter is also involved in complex decision-making, and in all cases, the data situation, coding, and use of output are rarely straightforward.

Since AI is an umbrella term for many different technological solutions to various problems in different organisations, it is not an easy management task to determine how AI can be employed to solve tasks and problems in an organisation in a manner specific to the organisation's mission and sufficiently adapted to its operational mode. There is a flood of ideas, and promises are continuously made when it comes to securing funds for development. Given the hype cycle, it is unsurprising that many of these promises cannot be fulfilled.

So what happens when project managers are held accountable by sponsors? When results need to be delivered? When lessons learned are to be shared?

What doesn't necessarily happen is that management focuses on whether or how the promise has been fulfilled, risking disappointment inside and outside the organization. Nor does management necessarily take specific responsibility for which parts of the promise have been fulfilled and to what degree. More often, we see managers doing something else, namely renegotiating the promises in different ways, which often succeeds in prolonging the hope that the technology will ultimately fulfill the promises. This is not about evading responsibility as a leader; on the contrary, there is a lot of subtle leadership work involved in extending hope between the time the promise is made and its final realization.

Firstly, we see leaders re-commit to the promise – 'we'll get there even if we don't stick to the schedule'. Secondly, we see leaders making new promises or highlighting activities and results that were not promised – 'over the next few months, we'll prepare to deliver x and initiate activity y'. And thirdly, in the case of AI, we see leaders formulating new approaches to fulfill the promise, redefining AI (given its complexity) – 'this can also be interpreted as AI'.

In other words, we do not find that management is distancing itself from the promises and the promising projects – what is referred to as de-coupling – but rather that the parties enter into new binding relationships through the extension of hope. This propels the projects forward.

Extending hope is legitimate and meaningful leadership work that – amid difficulties and potential disappointments with inflated promises – keeps employees and stakeholders engaged and mobilized. In this way, the organization's trajectory is maintained, and it continues to orient itself toward its future, which is the horizon to which hope is attached.

We believe this mechanism is central to understanding how organizations persist over time despite setbacks and disappointments.