The people you trust built the thing you need to replace
Every organisation hands its transformation to the people it trusts most. That is why so many transformations stall.
The structural reason is simple and uncomfortable. Authority displacement is governance failure. Structural transformation means moving decision authority: away from functions that have held it, towards a new operating model, often towards systems that now make decisions people used to make. The people asked to lead that move are the ones whose authority, reputation and networks were built on the model being replaced. Organisations give credibility to tenure and track record, and both were earned inside the legacy. So the transformation is governed by the structure it is meant to dismantle. Incentives do not fix this on their own: a reward for change with no consequence for resisting it is a bribe, and people take the money and perform the change while the operating model stays where it was. What works is outsider-induced change: someone with explicit authority over the operating model and over the personnel decisions, for a defined term, who then leaves.
This paper goes further than the usual call for "strong leadership". It sets out the mechanisms that make internal transformation stall, what an outsider's mandate has to contain, the strongest objection to the whole idea, and the order in which boards should act.
One thing it does not do is quote the claim that 70 per cent of change programmes fail. That figure has circulated for decades, and a review of the published sources behind it found no valid and reliable empirical evidence to support it. The argument here does not need a failure rate. It needs a mechanism.
Credibility is a claim on the old model
Start with how an organisation chooses who leads its transformation. It looks for people with authority, credibility and deep knowledge of how the place works. It finds them where they always are: among the executives who built the current operating model or rose to the top of it.
Their credibility is real. It is also a claim on the old model. Their expertise is expertise in running it. Their networks were formed by it. Their reputations were made by its results. When they are asked whether the model needs structural replacement, they are being asked to judge something they cannot judge from the outside, because they are not outside it.
This is not a question of character. It is a question of position, and there is evidence for it. In a large-scale survey of top executives, Donald Hambrick and colleagues studied what drives "commitment to the status quo", meaning an executive's attachment to the organisation's current strategy and leadership profile. They found that an executive's tenure in an industry was a pronounced driver of that commitment, with significantly more effect than tenure in the organisation. Commitment was also higher when the firm was currently performing well. And incumbent chief executives tended to believe their successors should be just like them.
Each finding closes an exit. Industry tenure matters more than company tenure, so hiring a senior executive from a competitor imports the same paradigm under a different badge. Good current performance raises commitment, so the organisation is least able to change itself at exactly the moment when change is cheapest. And the incumbents' preference for successors like themselves means the legacy selects its own continuation, unless someone outside it intervenes.
Picture how this plays out. A regional insurer decides its claims operating model needs to be rebuilt around automated triage. The board asks its longest-serving operations executive, the person who designed the current claims process and knows every exception in it, to lead the review. Six months later the review concludes that the model is fundamentally sound and needs better tooling, a new workflow platform and some targeted automation. Nobody lied. The executive applied the best judgement available from inside the model. That judgement is the problem.
The selection process makes it worse. A transformation lead is usually chosen by the chief executive or a sponsoring executive, and the choice has to survive the existing leadership team. A candidate who threatens that team will not survive the selection. A candidate the team trusts will. The organisation therefore picks the person least likely to disrupt the people who picked them, and then announces that it has chosen someone with the credibility to drive change. Both statements are true. They describe the same person.
Legacy governance is an immune system
The second mechanism is governance itself. Legacy governance was designed to preserve a working system. It does that job well. Steering committees with a representative from every affected function, risk and architecture boards, budget cycles, approval matrices: each was built to stop the organisation from making careless changes to something that works.
Point that machinery at a structural transformation and it does exactly what it was built to do. Every function that will lose authority sits on the steering committee. Every one of them has a legitimate concern to raise, and every concern becomes a condition. The transformation office reports to an executive whose own domain is being redesigned. Status reports go green, because amber invites scrutiny and red invites blame. The programme produces more committees, more controls and less movement. That is not sabotage. It is an immune system attacking a foreign body.
Ford in 2006 is a public example of what that looks like from the inside, and of what it takes to break it. Alan Mulally arrived from Boeing that year and found a company posting massive losses. He introduced a required weekly meeting, the Business Plan Review, at which senior executives had to disclose problems honestly, using a red-coded slide for a big one. That ran against decades of history at Ford, where telling the truth could get you fired. So nobody went first. Mark Fields, a senior executive already in place when Mulally arrived, was the first to test whether the new boss meant it, by showing a red slide. He was not punished for it. The public account of that moment describes it as the day Ford's dysfunctional culture began to crumble.
The red slide is usually told as a story about culture. It is a story about authority. An insider could not have made red safe, because an insider would have been part of the system that made it dangerous. The outsider could change the governance instrument because he owed nothing to the old one, and because he held the authority to decide what happened to people who reported red, or kept reporting green.
Note what Mulally changed first. Not the organisation chart. The decision forum: the place where the truth about the operating model was reported, and the rules for what happened next. Structural transformation starts by replacing the governance that protects the legacy, not by reorganising underneath it.
AI makes the immune response stronger, not weaker. When a model starts making decisions that experienced operators used to make, such as approving a claim, pricing a quote or routing a case, the accountability framework rarely moves with it. The operations leader is still accountable for the outcome. The analyst is still accountable for the numbers she reports. But the decision is now made by a system they did not design and cannot fully question. Accountability without authority produces a predictable response: another review layer, a manual check on the model's output, a shadow process that restores human control over a decision the official system has nominally automated. Each of these is reasonable on its own. Together they rebuild the legacy around the new technology. Legacy governance does not reject AI. It absorbs it and neutralises it, one sensible control at a time.
Incentives without consequences are bribery
The standard response to resistance is to pay for change. Add transformation milestones to the bonus scheme. Create a long-term incentive tied to the new operating model. Celebrate early adopters.
These incentives are not useless. But on their own they buy the appearance of change, for a structural reason. A milestone is something the organisation can declare: a platform launched, a new structure announced, a process redesigned on paper. The operating model is something the organisation has to live: who actually decides, whose numbers matter, which exceptions still go to the old authority. People are rational. If the reward attaches to the milestone and nothing happens to those who keep the old model running underneath it, they will hit the milestone and keep the old model. That is not cynicism. It is the cheapest way to meet the target while protecting their position.
A carrot without a stick is a bribe. It pays people to say yes without changing what yes means.
The stick in structural transformation is not punishment. It is the personnel decision. Who stays in the role. Who moves. Who leaves. As long as the leaders of the legacy model know that their positions are safe whatever happens to the transformation, the rational strategy is to outlast it. They have usually seen it work before. Transformation leads rotate, programmes are rebranded, sponsors move on, and the people who quietly kept the old model running are still there when the next programme starts. Waiting is not resistance. It is experience. The moment they know that someone with real authority will decide their roles based on what the operating model actually becomes, the calculation changes.
Japan Airlines shows what it looks like when that authority is explicit. JAL filed for bankruptcy protection in January 2010 with debts of 2.3 trillion yen. At the request of the Japanese government, Kazuo Inamori, the founder of the electronics and ceramics company Kyocera, became chairman in February 2010. The restructuring plan called for about 15,600 job cuts, roughly a third of the workforce, with a government-backed turnaround fund investing in the carrier.
The job cuts were the stick, and they were large. But the structural move was the operating model. Inamori introduced Amoeba Management, breaking the airline into small units that each tracked their own revenue and costs, alongside a written company philosophy built on the one he had used at Kyocera. Accountability became visible at a level the old structure had never measured. JAL relisted on the Tokyo Stock Exchange in September 2012, two years and eight months after the bankruptcy filing.
Nobody at JAL needed a transformation bonus to understand that the operating model had changed. The authority over the model and over people's roles had visibly moved.
The mandate: authority, a term and an exit
If the outsider is the mechanism, the mandate is the design. Most failed outsider interventions fail here, because the organisation brings in an outsider and then withholds the authority that would make the outsider useful.
A real mandate has four parts.
Authority over the operating model. Not authority to recommend one. The right to decide how decisions are made, which functions hold which rights, which governance forums exist and what they control. When Lou Gerstner arrived at IBM in 1993, the first leader recruited from outside the company since its founding, there was already a plan in motion to split it into separate businesses. The insiders had made the operating-model decision. Gerstner reversed it, and later called keeping IBM together the most important decision he ever made. Only someone with authority over the operating model could have overruled a plan the organisation had already adopted.
Authority over personnel. Who stays, who moves, who goes, including at the top. Without it, the outsider is an adviser with a large title, and the organisation knows it. This authority does not have to be used widely to work. It has to be credible, which usually means it is used visibly and early, at least once, on someone senior who was expected to be safe. The organisation watches the first personnel decision more closely than any strategy document. If the board overrules it, or lets it be negotiated away, the mandate is finished, whatever the paperwork says. That is why the outsider should report to the board, not to the executive whose model is being replaced.
A defined term. Structural change has a beginning and an end. The mandate should say how long, and what "done" means: the operating model installed, the governance replaced, a successor team in place.
An exit. The outsider leaves. This is the part organisations least expect and most need. An outsider who stays long enough stops being an outsider. Their decisions become the new model, their appointees become the new establishment, and their judgement starts to defend what they built. The research on commitment to the status quo applies to them too. The outsider who stays becomes the next legacy.
The public cases show the exit in practice. Inamori became chairman emeritus of JAL in February 2012, about two years after he arrived. Mulally handed Ford to Mark Fields, the executive who had shown the first red slide, who became chief executive on 1 July 2014. Gerstner was succeeded as chief executive in 2002 by Sam Palmisano, who had joined IBM in 1973 as a sales representative. In each case the successor came from inside, but from an inside the outsider had already changed.
The exit is hardest when the outsider has succeeded. Results improve, the board grows comfortable, investors associate the recovery with one name, and the natural instinct is to ask them to stay. That is the moment the risk turns around. A successful outsider who stays becomes the person whose credibility is a claim on the model they installed, which is exactly the condition this paper started with. The exit has to be agreed at the start, written into the mandate, and enforced by a board that understands it is protecting the next transformation, not ending this one.
There is a distinction worth making plain. Advisory authority is not authority. Consultants, transformation offices and interim programme directors can diagnose, design and recommend. They cannot decide who stays. A board that hires advice when it needs authority has bought a better description of the problem it will keep.
The strongest objection: outsiders lack context and destroy knowledge
The case against outsider-induced change is serious and should be stated at full strength.
Outsiders do not know the business. They do not know why things are the way they are, which apparently pointless processes hold the operation together, which customers will leave if a habit is broken, or which quiet expert keeps a critical system running. Institutional knowledge is mostly tacit: it lives in people and relationships, not in documents. An outsider with the authority to replace people can destroy in months what took decades to build, and by the time the damage is visible, the people who could have warned them are gone.
The evidence gives this objection real weight. When Ron Johnson moved from building Apple's retail business to run J.C. Penney, he ditched coupons and most sales events in favour of everyday low prices. Revenue fell 25 per cent, from $17.26 billion to $12.98 billion, and the board removed him after 16 months, bringing back his predecessor. Research on chief executive succession points the same way. A study of 193 departing CEOs by Yan Zhang and Nandini Rajagopalan found an inverted-U relationship between the level of strategic change and firm performance: moderate change helped, large change hurt, and both effects were more pronounced under outside CEOs than inside ones. Joseph Bower of Harvard Business School argued that the best leaders are inside-outsiders: people from within who have kept enough detachment from local traditions to see with an outsider's objectivity. And one of the most discussed technology turnarounds of the past decade was led by an insider. Satya Nadella had worked at Microsoft since 1992 when he became its third chief executive.
So the objection is right about something. Outsiders do fail, sometimes badly, and they fail in a recognisable way.
The pattern is worth naming, because a board can watch for it. The outsider arrives with a model from somewhere else and installs it whole. Customer and operating evidence is treated as resistance. The people who could explain why things work as they do are removed before they are heard. Change is measured by how much has been changed, not by what the operation now does better. Each of these is a choice about method, not a consequence of being an outsider.
But look at what the evidence actually separates. The objection treats context and commitment as one thing. They are two. Context is knowledge of how the system works. Commitment is investment in keeping it that way. The Hambrick research found that commitment to the status quo is driven by tenure in the paradigm. Context can be acquired. Commitment cannot be removed. The design problem is to get the context without the commitment, and that is solvable: keep the people who hold the knowledge, and take away their veto.
The successful outsider cases did exactly that. Mulally's best-known move at Ford was not a purge. He changed the forum in which the existing executives reported, and in time handed the company to the one who had told the truth first. JAL kept flying throughout its bankruptcy, run by the people who knew how to fly it. The knowledge that was lost in those turnarounds was mostly knowledge of how to operate the old governance: how to keep a slide green, which committee to lobby, which exception route avoided scrutiny. Losing that knowledge is not a cost. It is the point.
The J.C. Penney case is not a failure of outsider status. It is a failure of method. Johnson had authority over the operating model and used it at the far end of the inverted U: the largest change, made fastest, with the least evidence. According to an executive quoted at the time, when colleagues argued for testing the new pricing first, his answer was that it would be rolled out, not tested. The lesson is that an outsider's mandate must include a requirement to test and learn, and a scope that matches the Zhang and Rajagopalan finding: deep change to authority and governance, not a gamble on everything at once.
And Nadella fits the model rather than refuting it. He came from the cloud and enterprise business, at the edge of the Windows franchise that had defined Microsoft, and the board changed around him: Bill Gates stepped aside as chairman when Nadella was appointed. Bower's inside-outsider is still an outsider in the sense that matters. "Outsider" describes a structural position, outside the paradigm and its network of authority. It is not a passport. When an organisation has a true inside-outsider and gives that person explicit authority, use them. Most organisations, when they look honestly, do not have one in a position to take the job. Their most credible candidates are credible precisely because they are committed.
What boards change, and in what order
The sequence matters, because each step creates the conditions for the next.
First, name the problem as structural, and name the leadership as part of it. The board has to say plainly that the current operating model cannot be fixed by its current owners. This is the step organisations resist most, which makes it the best test of whether the transformation is real. If the board cannot say it, stop and save the money.
Second, write the mandate before choosing the person. Specify authority over the operating model, authority over personnel up to and including the executive team, the term, the definition of done and the exit. If the board is unwilling to grant those rights, it does not want an outsider. It wants a sponsor for the existing plan.
Third, choose for distance from the paradigm, not just from the company. An executive from a direct competitor may carry the same commitments under a new badge. Look for people outside the industry's assumptions, or for a genuine inside-outsider from the periphery of the business, and pair either with deep operating context.
Fourth, keep the knowledge and remove the veto. Retain the people who know how the operation actually works. Give them a formal channel to warn and to challenge, and no power to block. Require the outsider to test changes before scaling them, and to report what the tests showed, including the ones that failed. The board's job here is not to second-guess each decision. It is to check that the outsider is learning faster than they are changing.
Fifth, replace the governance before the structure. Install the new decision forum and the new rules about what happens to bad news before redrawing the organisation chart. Then make the personnel decisions visibly, early, and on the basis of how the new operating model is actually working.
Sixth, build the succession and enforce the exit. From the first year, identify the insiders who will run the new model, and move them into the roles where they will prove it. When the definition of done is met, the outsider leaves on schedule, and the board, not the outsider, owns the choice of successor.
Boards should also stop four things. Stop appointing the architect of the legacy model to lead its replacement. Stop funding transformation offices that report to the executives whose authority is being redesigned. Stop paying milestone bonuses with no consequence attached to the operating model that actually runs. Stop buying advice when the problem is authority.
This is a governance argument, not a leadership fashion. AI is making it urgent, because AI moves decisions faster than legacy governance can absorb, and every model that reaches production redistributes authority somebody currently holds. That theme runs through why AI transformation fails. The organisations that absorb the shift will be the ones that let someone outside the legacy decide how authority moves.
You cannot fix legacy governance from inside legacy governance. And the one who fixes it has to leave before they become it.