This guest contribution is by Dr Toby Velte, Affiliate Faculty member at London Business School (LBS), who led Transformational Leadership in the Age of AI, an executive education programme jointly developed by London Business School, Google and Deutsche Bank. Completed by the bank’s top 200 leaders, including the Management Board and Global Leadership Circle, the programme helped build AI fluency, confidence and a shared understanding that AI alone does not create value: leadership does. The difference between experimentation and enterprise-wide impact lies in leaders’ ability to scale adoption, steer change and embed AI responsibly across the organisation. In this piece, Dr Velte explores how leaders can do exactly that.
The best leaders working with AI today have arrived at a single realization: the problem is no longer awareness; it’s conversion to value. And at institutional scale, value comes from deploying AI safely. The organizations that thrive will not necessarily be the ones with the most advanced models. They will be the ones whose executives can navigate the structural, cultural, and governance shifts that safe scaling demands.
AI is a different kind of change
Most technology programmes optimise an existing workflow. AI does something else. It reconfigures how value is created, how decisions get made, and who is accountable when a decision turns out to be wrong. That difference is why so many AI programmes stall after a promising pilot.
Leaders must redesign organisational architecture deliberately. Legacy silos clash with workflows that are cross-functional by nature. Cultural norms that reward compliance and categorically punish failures will squash all experimentation. And incentives that reward experimentation alone will not survive a regulator’s questions. A balance must be struck, and only leadership can make that trade explicit through the creation of organisational structures with incentives that drive the desired outcome, and cultural values that support both.
Culture cannot be left to evolve on its own. Executives need to build an environment where risk is understood and quantified rather than simply feared, so it can be weighed against the upside of a given initiative. Without that, AI work stays trapped in pilot purgatory, blocked not by technical failure but by the absence of a clear risk and benefit mapping. Scaling does not require leaders to accept more risk. It requires them to understand risk with greater precision.
The work itself is shifting
Waves of cognitive automation are already hitting businesses. Routine analysis, data reconciliation, information synthesis, and other lower complexity tasks are moving quickly into the domain of AI assistants. What remains, and what leaders must actively design for, is human judgment, situational interpretation, and client relationships.
This is not a skills gap. It is a management transformation, and it can only originate at the top. Leaders have to redraw task allocation, redefine what good performance looks like, and model how teams work alongside the AI systems they now supervise.
Some tasks will be automated outright. Most will be augmented. The human still does the work, but now with insight from an assistant that has read every relevant document and compared all transactions.
Automation delivers speed, accuracy, and operational efficiency. Augmentation delivers productivity, proprietary solutions, and new forms of value creation including novel offerings the institution could not have previously produced at all. Leaders who treat AI only as a replacement tool capture the first and miss the second, greater opportunity. The manager of the future does not simply allocate tasks; they curate human and machine working systems.
Calibrating risk is a core executive job
AI introduces risks the industry does not yet fully understand, including the risk of standing still with AI while competitors do not. Perhaps the most critical leadership function now is calibrating risk tolerance at the institutional level.
Safe scaling does not mean always playing it safe. It involves setting executive-defined boundaries within which speed and oversight operate together. Traditional stage-gate governance, where every project passes through the same committee at the same intervals, will strangle AI innovation. It applies heavy scrutiny to trivial use cases and no more than that to consequential ones.
The alternative is a dynamic control framework: continuous monitoring, automated compliance checks, real-time audit trails, and accountability pushed to the teams closest to the work product. This is not about removing guardrails. It is about engineering a governance system that does not have to treat every AI project identically, moving low-risk deployments through quickly while concentrating senior human attention on the hardest cases. Most companies already have model risk management and multiple lines of defence. What is usually missing is the triage logic that decides quickly and consistently which project goes down which lane.
Done well, control becomes the accelerator rather than the brake. Governance at AI scale must be built in parallel with deployment by executives who understand that risk and speed are not opposites but interdependent variables.
The race is won by leadership
The AI advantage will not go to the company that buys the most technology. It will go to whomever can shift culture, redesign organisational structure, and rebuild governance for both speed and resilience. Those are leadership tasks. The question is not whether your organisation can afford this transformation. It is whether you can afford to wait.
Home
This guest contribution is by Dr Toby Velte, Affiliate Faculty member at London Business School (LBS), who led Transformational Leadership in the Age of AI, an executive education programme jointly developed by London Business School, Google and Deutsche Bank. Completed by the bank’s top 200 leaders, including the Management Board and Global Leadership Circle, the programme helped build AI fluency, confidence and a shared understanding that AI alone does not create value: leadership does. The difference between experimentation and enterprise-wide impact lies in leaders’ ability to scale adoption, steer change and embed AI responsibly across the organisation. In this piece, Dr Velte explores how leaders can do exactly that.
The best leaders working with AI today have arrived at a single realization: the problem is no longer awareness; it’s conversion to value. And at institutional scale, value comes from deploying AI safely. The organizations that thrive will not necessarily be the ones with the most advanced models. They will be the ones whose executives can navigate the structural, cultural, and governance shifts that safe scaling demands.
AI is a different kind of change
Most technology programmes optimise an existing workflow. AI does something else. It reconfigures how value is created, how decisions get made, and who is accountable when a decision turns out to be wrong. That difference is why so many AI programmes stall after a promising pilot.
Leaders must redesign organisational architecture deliberately. Legacy silos clash with workflows that are cross-functional by nature. Cultural norms that reward compliance and categorically punish failures will squash all experimentation. And incentives that reward experimentation alone will not survive a regulator’s questions. A balance must be struck, and only leadership can make that trade explicit through the creation of organisational structures with incentives that drive the desired outcome, and cultural values that support both.
Culture cannot be left to evolve on its own. Executives need to build an environment where risk is understood and quantified rather than simply feared, so it can be weighed against the upside of a given initiative. Without that, AI work stays trapped in pilot purgatory, blocked not by technical failure but by the absence of a clear risk and benefit mapping. Scaling does not require leaders to accept more risk. It requires them to understand risk with greater precision.
The work itself is shifting
Waves of cognitive automation are already hitting businesses. Routine analysis, data reconciliation, information synthesis, and other lower complexity tasks are moving quickly into the domain of AI assistants. What remains, and what leaders must actively design for, is human judgment, situational interpretation, and client relationships.
This is not a skills gap. It is a management transformation, and it can only originate at the top. Leaders have to redraw task allocation, redefine what good performance looks like, and model how teams work alongside the AI systems they now supervise.
Some tasks will be automated outright. Most will be augmented. The human still does the work, but now with insight from an assistant that has read every relevant document and compared all transactions.
Automation delivers speed, accuracy, and operational efficiency. Augmentation delivers productivity, proprietary solutions, and new forms of value creation including novel offerings the institution could not have previously produced at all. Leaders who treat AI only as a replacement tool capture the first and miss the second, greater opportunity. The manager of the future does not simply allocate tasks; they curate human and machine working systems.
Calibrating risk is a core executive job
AI introduces risks the industry does not yet fully understand, including the risk of standing still with AI while competitors do not. Perhaps the most critical leadership function now is calibrating risk tolerance at the institutional level.
Safe scaling does not mean always playing it safe. It involves setting executive-defined boundaries within which speed and oversight operate together. Traditional stage-gate governance, where every project passes through the same committee at the same intervals, will strangle AI innovation. It applies heavy scrutiny to trivial use cases and no more than that to consequential ones.
The alternative is a dynamic control framework: continuous monitoring, automated compliance checks, real-time audit trails, and accountability pushed to the teams closest to the work product. This is not about removing guardrails. It is about engineering a governance system that does not have to treat every AI project identically, moving low-risk deployments through quickly while concentrating senior human attention on the hardest cases. Most companies already have model risk management and multiple lines of defence. What is usually missing is the triage logic that decides quickly and consistently which project goes down which lane.
Done well, control becomes the accelerator rather than the brake. Governance at AI scale must be built in parallel with deployment by executives who understand that risk and speed are not opposites but interdependent variables.
The race is won by leadership
The AI advantage will not go to the company that buys the most technology. It will go to whomever can shift culture, redesign organisational structure, and rebuild governance for both speed and resilience. Those are leadership tasks. The question is not whether your organisation can afford this transformation. It is whether you can afford to wait.
How helpful was this article?
Click on the stars to send a rating