Podcast: change management in an era of AI disruption

{"text":[[{"start":4.75,"text":"In a specially recorded conversation, FT project publishing editor Oliver Ralph talks to Matilde Guilhon of Skema Business School, Michael Watkins of IMD Business School and FT global education editor Andrew Jack about the challenges facing managers as they deal with AI, and how business school research sheds light on the strategies that can be used. "}],[{"start":28.85,"text":"The discussion (below) is part of the Management Insights series, which highlights the latest business school insights on the key questions facing today’s business leaders. "}],[{"start":null,"text":""}],[{"start":37.85,"text":"Transcript"}],[{"start":38.95,"text":"Oliver Ralph: AI is clearly the hot topic at the moment, creating huge opportunities for businesses but also daunting challenges for the management teams tasked with implementing the technology. Matilde, what are executives that you’re teaching and working with saying to you about what they’re facing? "}],[{"start":56.550000000000004,"text":"Matilde Guilhon: I think we should focus on three tensions. The first one is the tension between pace and readiness, since boards are demanding speed while workforces are unprepared. The second one is centralisation. Who actually owns AI strategy? Is it the centre or are we focusing on the business units? And third, and this is probably the most consequential, is the tension between efficiency and trust. The most obvious return on investment cases involve, of course, reducing headcount. And those are precisely the ones most likely to destroy employee confidence if it’s handled poorly. "}],[{"start":95.25,"text":"There is interesting research by [Sebastian] Raisch and [Sebastian] Krakowski on what they call the automation augmentation paradox. Organisations face a tension between using AI to replace human tasks and using it to enhance human capability. The strategic choice between the two is not purely technical, but rather political, because it signals to employees what the organisation believes they’re worth. "}],[{"start":121.05,"text":"OR: Andrew, what are people saying to you about the main challenges? Where are researchers focusing their attention on this topic? "}],[{"start":129.5,"text":"Andrew Jack: There’s an intensification of challenges. Since Covid, a lot of those things that seemed a bit abstract are actually directly affecting the operations of businesses, notably conflicts and geopolitical tensions, including [US president Donald] Trump’s tariffs. And then all of that’s been overlaid by AI as a disruptive force. "}],[{"start":150,"text":"But [there’s] a disconnect between the peer pressure [on] employers and companies to embrace AI and then the gap towards how to meaningfully use it. Just how much is it saving and what are the potential risks and implications for the workforce? "}],[{"start":164.6,"text":"OR: Matilde, a lot of your research has looked at previous episodes of tech disruption. Do you think AI disruption is similar to previous examples? Are there lessons that can be drawn from how businesses dealt with disruption in the past? "}],[{"start":177.95,"text":"MG: For each wave of threat of disruption or disruption, what is similar is the fear of job displacement and the need for reskilling. But with AI, what’s different is the breadth and the speed. "}],[{"start":189.6,"text":"AI hits knowledge workers at the core of what they do, which is reasoning, drafting, and analysing, whereas the previous waves mainly disrupted transactional or operational work. So, I think it’s interesting here to go back to classic theory on the separation between what is competence-enhancing and what is competence-destroying. Previous tech waves were largely competence-enhancing for knowledge workers. They gave you better tools to do what you already did and knew how to do. But with AI, it’s more ambiguous. It can be competence-destroying or competence-enhancing, and it really depends on how it’s deployed. It is precisely this ambiguity which makes it psychologically harder than the previous tech waves. And it’s psychologically harder because for knowledge workers, it really threatens their sense of professional identity. "}],[{"start":239.39999999999998,"text":"OR: Michael, how are managers approaching the challenge of AI, which is both an organisational and professional challenge for them, but also potentially a personal challenge? "}],[{"start":249.29999999999998,"text":"Michael Watkins: I think the first point is the profound identity challenges that flow from this technology. And especially because we’re seeing it moving so quickly. So that which we still feel anchored in today could easily just be taken away tomorrow. "}],[{"start":263.5,"text":"There are a couple of things that I would have you consider with regard to the implementation of AI. One is the impact on teams, which I think has not been focused on enough. There’s a tendency for all of us to engage in these almost echo chamber-like conversations with AI. And they like to make us happy, right? And so I find sometimes I have to stop and say, please stop being nice to me, kindly be a little bit more rigorous in this. But when you have a leadership team where individuals are making these interactions with AI and bringing the results to the team, there’s a real danger of distortion happening, a kind of entrenching of positions happening on leadership teams. So to me, one of the most profound challenges we face is how do we help teams leverage AI in a way that actually increases the sharing, increases the integration that happens."}],[{"start":310.05,"text":"Maybe the biggest debate today about AI is what will eventually the job impacts be. And that ranges from job apocalypse on one side to utopian, we’re all going to engage in fascinating activities on the other. The reality is somewhere in between. The evidence is unclear. As the cost comes down, I think we’re going to continue to see an erosion of employment. But maybe the biggest question we face today is what is the employment and therefore social impact [from] this technology going to be over the next three to five years. "}],[{"start":339.3,"text":"AJ: The other thing I wonder about is whether it drives flatter organisations and also almost inverts the hierarchy of power in business and other organisations, when it’s perhaps the younger staff who have maybe greater mastery or at least instinctive use of these technologies than the C-suite. And how that changes the power dynamics and the decision-making. "}],[{"start":359.7,"text":"MG: I think we are witnessing a change in the hierarchy, and especially in organisations where you have younger people able to build agents, rather [than] . . . just use AI in a very informational and conversational way. So definitely I think there is something to do within teams. And I do not believe at all in a champion, a so-called design champion in an organisation that will teach others the best practices, but I think it should be a team effort using AI. "}],[{"start":385,"text":"MW: What I would add is that organisational design for the last 150 years has been constrained by human cognitive limitation. The reason we have functions, the reason we build silos, the reasons we need to think about integration fundamentally arise because of limitations on our cognitive capabilities. And networks of agentic AI systems are really blowing up that constraint."}],[{"start":409.25,"text":"What it means is we’re moving from a design point of view, from thinking in terms of functions and specialities to thinking in terms of horizontal flows and the primacy of those horizontal flows. And that changes profoundly how you think about designing organisations. It does make them flatter. It does raise the issue of overall decision-making and governance. It raises fascinating questions about who’s responsible. The governance questions I think are absolutely fascinating. "}],[{"start":437.4,"text":"OR: The speed of AI disruption is such that it’s very difficult now to say what the reality is going to be, even in six to 12 months’ time. And so as a manager, particularly in a very large and potentially slow-moving organisation, you need to change, but implement change for a future that is fundamentally unknowable. "}],[{"start":456.75,"text":"MW: The two biggest questions of AI are net impact on employment and speed of implementation. And those two will determine largely how big the overall social impact is. "}],[{"start":467,"text":"How do you manage when the frontier is receding in front of you in terms of capability all the time? And the answer is that you have to be able to, first of all, sense and respond more rapidly than you did before. Planning cycles for many organisations are far too long to stay tracking with what’s happening. And then you also need to think in options terms. What options are we going to try and develop for ourselves? Once we know which trajectory things are going, we can begin to accelerate down that trajectory. It’s those sorts of things, the sensing, responding cycle time, that’s so critical, the ability to engage in experimentation, but also even more critically put your money on certain options."}],[{"start":508.75,"text":"MG: It’s very difficult to design a strategic plan nowadays. It’s not so sustainable to plan something if we don’t know what’s going to come in the coming years. So we are moving more to scenario planning. "}],[{"start":520.05,"text":"And this is precisely what the best consulting firms are doing. They are not helping companies now to design their three- or five-year strategic plan, but rather they’re imagining a lot of scenarios, which is made possible by the large amount of data we have today. So, establishing an AI strategy that would be valid and sustainable for a few years is not the best idea, but the ambition should be clear . . . something regarding the values we want to pursue as a company regarding the use of AI. "}],[{"start":549.75,"text":"But this is related also to the kinds of competitive advantage we are trying to build and what role we want humans to play in the core processes and the data position as well. "}],[{"start":560.15,"text":"In high-velocity environments, the notion of a durable strategy built around the fixed position we need to defend is a liability. "}],[{"start":568.25,"text":"MW: [Strategy expert] Rita McGrath has written about transient advantage, and the extreme form of the argument says the only durable form of advantage in the future is adaptation. Now that’s clearly too strong a statement because we know that there will remain important sources of competitive advantage. A global mining company is still going to have a set of defendable moats around its business. So we don’t want to go to the extreme of that argument. "}],[{"start":592.3,"text":"But it’s clear that advantages of almost every other sort are becoming more transient. And so we really need to rethink, as leaders of enterprises, what are the sources of competitive advantage going to be that we’re going to invest in. And the ability to adapt clearly is one of them. I think both about the impact on competitive advantage, but also the impact on organisational design. And when you’re bringing those two things together it’s a bit of the wild west at the moment. "}],[{"start":620.9,"text":"OR: The idea that, rather than a typical strategic plan, what you need to have is a set of principles around which you build a lot of scenarios, requires a real sea change in the way that senior managers and boards think about the future of their businesses. Do you think boards are ready to make that scale of shift in their thinking?"}],[{"start":638.35,"text":"AJ: I doubt it. You certainly get the impression that AI before and tech in general was one among many different priorities that would get up to the C-suite or the boardroom. Certainly there seems to be far greater insight and prioritisation [now] of the technology. There’s clearly a demand that perhaps has not yet been met at that leadership level."}],[{"start":657.95,"text":"MG: There is also a high risk of AI strategy as a distraction. And that is a version of AI transformation that consumes enormous leadership bandwidth, while actual adoption stays very low. And I think there is here a capability trap. On one side there is the appearance of progress substituting for actual progress as the metrics are about activities — the number of strategy decks produced under AI [for example], or how vendors are evaluated with algorithmic management. We are not focusing really on the outcomes, how has work actually changed, because we lack these new metrics. We focus a lot on the activity but without having the proper tools to measure it. "}],[{"start":696.8000000000001,"text":"OR: Michael, you mentioned earlier that a lot of pilots were underperforming. Is it that the pilots are underperforming or is it a question of measurement? Are companies measuring the right things when it comes to their AI experiments?"}],[{"start":709.2,"text":"MW: I think the measurement of outcomes depends a lot on the inputs you’re talking about. For example, the current cost of compute really obscures the picture a lot. One of the most important things that has to happen if AI is going to be more deeply implemented is the cost of compute has to come way down. It’s certainly going to, it’s just a question of what the timeframe is going to be. "}],[{"start":731.75,"text":"My view of why pilots are failing though, or underperforming, is a little bit different. I think they’ve been defined in some sense too narrowly. You know, “let’s work on this part of some part of a process” or “let’s automate a particular piece” and, if you buy into my idea that organisational design is now mostly about flows and not about functions, then pilots need to focus on flows and not functions. And only when you start to think about how [to] really apply agentic AI across an entire flow through an entire business, like the customer journey, do you start to see the real benefits. "}],[{"start":766.15,"text":"I was teaching a digital programme at IMD [in May], and I surveyed the participants. Only 40 per cent thought their organisations were ready in the sense of being able to sense and respond rapidly enough to what’s going on. And so that means there’s 60 per cent out there that are kind of dinosaurs awaiting the arrival of the asteroid. And I worry a lot about that. "}],[{"start":786.35,"text":"AJ: The other thing I wonder about is dependency. What are the guardrails that leadership teams need to think about in terms of their reliance on a third party, both for cloud, of course, for data, but also they’re building tools and applications on top of platforms that are controlled out of either the US or China at a time that we’re seeing all this segmentation and new regulation. Do you think that creates a new level of uncertainty through dependency that [organisations] somehow need to adjust to or anticipate?"}],[{"start":814.85,"text":"MG: The question of tech sovereignty is becoming more and more a concern for large companies. What is problematic to me is that it seems to be something that is [well] handled by big companies for the simple reason that these big companies are operating in different regions with different regulations, and the simple fact that they need to comply with different data regulations makes them obliged to tackle this tech sovereignty issue. The problem is more with the SMEs that are not fully equipped to address these issues. And I think they are the ones who need more help, and especially from academics, from researchers. "}],[{"start":849.65,"text":"MW: This is raising pretty big questions about the boundaries of firms, and also about potential disintermediation of firms. Who really controls the value creation in the system? Who controls the interface with the customer becomes a big issue. "}],[{"start":864.75,"text":"[There are] small plucky companies out there that are taking a portion of a value chain and automating it far faster than the large firms can deal with. So I think you’re hitting some of the really profound questions about the impact of agentic AI and [over] who reaps the value from this, including, potentially, the companies that control the models. I think it is a really important question that leaders should be grappling with. "}],[{"start":889.85,"text":"The transcript has been edited for brevity and clarity"}],[{"start":899.45,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1780716751_7889.mp3"}