Opinion

Why AI will transform how we understand leadership

As generative AI continues to grab headlines, executive coach Douglas Board explores whether it could reshape how we understand and evaluate leadership

Female Speaker Giving A Presentation During Business Seminar At Convention Center.

Over the next five years, generative AI could transform how we understand senior leadership, for better and for worse. 

My experience is in recruiting and coaching senior leaders. My focus is on the C- and D-suites, meaning the very top executives and the tier of leaders below them. At the end of last year I became interested in whether generative AI might change our understanding of leadership at this level. To explore this fully, I enrolled in an executive education programme offered by Insead. I came away convinced that AI will have an impact, but perhaps not in the ways some might expect. 

How AI can learn from our communications

To begin with, generative AI could radically change the way we understand executive leadership by analysing communication patterns and emotional cues in real time.

There are three reasons why this is now possible. Firstly, data is now as plentiful as grains of sand in a desert. When it comes to picking out patterns in all this data, humans can spot big, obvious trends, but contemporary computing power can see much harder-to-spot patterns – and fast. 

Secondly, when this data comes from the textual, audio and video communications which constitute executive life, generative AI has been able to learn from how we communicate with each other, so it can, theoretically, be trained on human behaviours like humour and sarcasm. 

Thirdly, the pandemic pushed so much workplace communication online that machine-learning algorithms suddenly had access to a bigger pool of this audio-visual data than ever before.

Before Covid, most sensitive C- and D-suite conversations were face-to-face. Who was in the room and what was recorded were tightly controlled. Covid changed this. Today, hybrid meetings are normal, even for sensitive topics. Those words, silences and facial expressions are digitally captured. They now constitute more grains of sand in the data desert. 

Could AI be used to manufacture charisma?

AI’s ability to process the data gleaned from online communications means that it can, in theory, start aping how great leaders communicate with their staff. And one of the first leadership attributes at risk is also one of the most human: charisma.

There have been highly publicised examples of the technology being used to help leaders have more meaningful connections with their workforce. One example is Publicis Groupe CEO Arthur Sadoune sending 100,000 New Year videos, using AI to personalise these to each colleague’s passions, be that skydiving, wakeboarding or ballet. If great leaders know how to connect with their people on a deeper level, what’s to stop mediocre leaders from mimicking this behaviour with the help of AI?

C-suiters using AI as an emotional autotune to optimise their daily messages may well become routine during 2025. If that happens, it might give boards options to “make or buy” charisma, widening the pool of individuals appointable to these roles. Even so, charm wouldn’t be dislodged from the place it has in our various idealisations of good leadership. How could machines disturb something so deeply woven into how we respond to each other? AI can only really be helpful in this way if our communications remain online. Any manufactured charisma would surely be exposed the minute a leader had to engage with someone in real time.

Three limitations for machine-learning algorithms

And operating face-to-face is not the only limitation for using generative AI to transform executive leadership. 

Three key things affect what machine learning can do with a vast amount of data: accuracy (how reliable was the data-capture process?), completeness (is there any data missing which could create bias or unrepresentativeness?) and legal status (who is allowed to own and process this data?).

To illustrate this, let us imagine an example. 

Out of the 250+ global companies with more than 100,000 employees, imagine one called Titan. Think about all the email, instant message, diary, audio and video traffic between the (say) 30 individuals in Titan’s board and C-suite, the 500 in its D-suite, and the 10,000 employees across the company with whom those leaders most frequently interact – seeking to direct, motivate, coach or challenge. How does this dataset stack up against our criteria?

Accuracy: the audio of in-person participants in hybrid meetings may be patchy, depending upon microphone quality, but the system knows who is in the room and can pull from other existing audio and video data to help clean this up. On top of this we have a wealth of other communication data: the messages sent, the messages drafted but not sent, facial reactions and the time lapses before those reactions. All in all, a pretty accurate dataset.

Completeness: everything is captured online, so the dataset is also pretty complete. 

Legalities: Titan owns this data and has a clear employer purpose for using it: improving leadership effectiveness. Although there ought to be safeguards in place to protect individuals’ privacy, this can be dealt with by limiting who has access to the raw data and any findings the AI might generate.

How GenAI can reveal important leadership traits 

And the usefulness of these findings could be significant, taking us beyond superficial measures of leadership effectiveness. Take this example. Suppose you wanted to research the importance of high-quality listening when assessing great business leadership. With sufficient budget and access to Titan’s communications data, before generative AI you could have tasked machines to look at which leaders spoke less during meetings or asked more questions. Were there any signs that they were more impactful than colleagues who stayed in broadcast mode? 

Silence can mean boredom or multi-tasking and some leaders ask questions to test or undermine, rather than to gain new knowledge. Generative AI will allow us to move beyond drawing conclusions based on which leaders talked in a meeting and which did not. These algorithms could look for leaders who are present in discussions when new evidence or ideas are aired. These leaders may or may not ask questions, but AI will allow boards to see whose subsequent actions or communications show them using what they have heard to good effect.

The result is that the next decade will be the first in human history in which we will have the data, the processing power and the machine-learning tools to study large-scale leadership in all its sprawling, real-time messiness. 

 Douglas Board is an executive coach, visiting professor at the University of Chichester and author of “Elites: can you rise to the top without losing your soul?”