AI is driving a new wave of restructures. Here’s how to make your next restructure ‘successful’

A synthesis of more than 30 years of research into organizational restructures to help you avoid 'failure'

A few weeks ago, I posted about organizational change frameworks, leading with the statement ‘When Bain reports that 88% of change interventions fail to deliver on their original ambitions, and McKinsey, BCG and Gartner all land in similar territory, it’s worth asking what’s not working.’ The essay was about change frameworks, but the responses that came back were almost entirely war stories about restructures.

Right now, restructuring seems to be everywhere. The trigger, more often than not, is reported to be AI - the pressure to reshape operating models, redesign roles, and build organizations fit for a different pace and a different kind of work. And alongside the announcements and the shareholder responses, there’s the stories about how people are being treated through these processes. The terminations by email. The roles eliminated in ways that left people feeling like line items rather than team members. The restructures that were declared successful by the executive and experienced as something quite different by the people who lived through them. These stories matter, morally, and because they point to a practical question that doesn’t get asked clearly enough - do restructures actually deliver the value they’re supposed to deliver? And do they do it in ways that are sustainable - for the organization, and for the people inside it?

Many senior leaders commissioning restructures probably sense that the intervention won’t deliver everything promised – but proceed anyway, because they aren’t sure that there is another viable way to drive the business improvements they need. This is where the research can help you get closer to the success you need.

After 20-odd years of working inside organizations navigating structural change, I could pull you together a practitioner’s guide to restructuring well. This would be written based purely on what I’ve observed works - and what has failed in ways that were entirely preventable. That guide would be useful to you. It would also be overly influenced by my experiences - shaped by the industries, the leaders and the moments that I was there for.

So, instead, I went down a rabbit hole of research. The collective evidence that has been produced since the 1990’s. Pulling together the findings from the researchers who’ve studied restructuring across hundreds of organizations, with a degree of objectivity that practitioner experience alone can’t provide. Plus the reports from the consulting houses (who tend to be the most vocal about low success rates for organizational restructures).

If the process of restructuring is so fundamentally flawed, then why is it one of the most leveraged tools to drive a different business outcome?

The research is considerably more insightful (and much less dramatic) than the 70%-88% failure-rate that gets thrown around in LinkedIn posts. It doesn’t confirm the worst-case nightmare scenario and it doesn’t confirm that all restructures will work. What it does is identify are the specific factors that separate the restructures that create value from the ones that don’t.

These factors are neither mysterious nor particularly expensive to act on. They’re just been waiting to be applied consistently.

The dramatic headlines are misleading

Consulting firms put the failure rate of change initiatives somewhere between 70% and 88%. If those numbers were reliable, restructuring would be the organizational equivalent of playing Russian roulette - an expensive habit with a poor expected outcome that organizations keep repeating for reasons that defy rational explanation.

Turns out that those statistics don’t have a defensible empirical basis.

Mark Hughes, writing in the Journal of Change Management in 2011, traced the five most-cited published instances of the 70% claim to their original sources. None of them rested on systematic empirical research. The earliest substantive origin - Hammer and Champy’s 1993 work on business process re-engineering - explicitly described its figure as “unscientific.” The claim has been bouncing around consulting reports, HBR articles and even peer-reviewed journals ever since, occasionally citing its own previous citations as evidence.

This is, if nothing else, a useful reminder about the research credibility of statistics that feel intuitively true.

What the peer-reviewed literature shows is more nuanced. The average return on a generic structural move is near zero or modestly negative. But those organizations that include the factors that matter record a better outcome.

The difference between a restructure that creates value and one that destroys it is not random. It’s a matter of how you go about designing and delivering it.

A note on how the evidence was assessed

Before you get to the good stuff – a clear word of caution.

No research is perfect – no research is completely defensible. No research is error-free. And what I’ve done with this essay is to consolidate and sort a wide range of findings and rank them. I’ve tried to make sense of what exists, so that you can make better decisions about driving change in your organization. I’m a psych, with training in statistics – but I am not an expert researcher – just an expert practitioner who likes to make sure she makes informed choices.

Here’s how I’ve assumed importance and impact -

• High = multiple meta-analyses and/or causal identification, with results replicated across independent research teams.

• Moderate = consistent findings from longitudinal panels or well-designed correlational studies.

• Low = cross-sectional, single-source, small-sample or perceptual evidence.

• Causal evidence in this space is rare – most of the findings are correlational. I won’t bore you with a detailed definition, but basically, don’t assume that if you do A, you will DEFINITELY get B…. Rather, if you do A, you’re more likely to get B, but not because A causes B.

• I can’t guarantee that I have included every single piece of research from everywhere.

• The research field is overwhelmingly Anglo-American. So, if you are operating in a global organization, don’t automatically assume that these factors are always consistent in different socio-cultural environments.

• Major papers are noted in this essay – but not ALL the references, because this isn’t an academic paper.

The point? “What works” in restructuring is more accurately described as “what is consistently associated with better outcomes.”

The full source list does include findings from major consulting firms - McKinsey, Bain, BCG, Deloitte and others. These are deliberately weighted lower than peer-reviewed academic research, for a few reasons. Consulting firms research the services they sell. Their samples skew toward organizations already engaged in structured change programs. Their outcomes are almost always self-reported by respondents rather than independently verified. And their methodologies are proprietary - so you can’t critically dig into their methodologies the way you can a peer-reviewed study.

That said, consulting research covers territory the academic literature doesn’t. Operating model redesign at scale. AI-driven workforce transformation. The practitioner conditions that produce outcomes in real organizational contexts under real time pressure. When consulting and academic evidence point in the same direction - which they do on several of the factors below - that convergence is worth paying attention to. It means two structured bodies of research have reached the same conclusion, which should give us more confidence.

Do restructures ‘work’?

If you do a restructure hoping that it will directly effect your financial performance, you might be disappointed. The peer-reviewed literature on restructuring - across downsizing, mergers and acquisitions, and organizational redesign – suggests that structural moves, on average, do not improve financial performance.

Steel and House (2024, 905 effect sizes across roughly 40 years of data) found that post-downsizing performance centres on zero or negative. A 2024 meta-analysis by Eshghi and Astvansh (78 studies, 34,594 layoff announcements) found a pooled investor negative reaction to restructures. The most important moderator in that analysis - why the restructure was happening. Reactive layoffs drew a further market penalty.

Investors don’t reward proactive restructuring. They simply stop punishing it.

The M&A literature tells the same story. King, Dalton, Daily and Covin (2004, 94 studies) found that acquirers do not gain on average, and that the factors everyone assumes matter - whether the acquisition was related or conglomerate, how it was financed, and whether the acquirer had done this before - showed no significant effect on performance. None of them.

It’s worth thinking about that for a moment – A lot of M&A strategy conversations revolve around precisely the factors that the research suggests don’t work. What then is the value of M&A for your business?

The implication isn’t that restructuring is futile. It’s that a generic restructure - one not designed around the specific factors that give you a good outcome - has a poor expected return.

How do you avoid ‘failure’?

These factors are ordered by evidence strength, strongest first. They aren’t absolute guarantees – but the more of them you build into your implementation, the better your odds.

1. The timing has to be proactive, not reactive

The most consistent finding across the downsizing literature is that restructuring driven by strategic intent - when the organization has room to move, before performance has deteriorated to crisis point - produces significantly better outcomes than restructuring from a burning platform. Love and Nohria (2005) found downsizing improved performance only when firms had high pre-existing slack, used broad-scope redesign rather than pure headcount cuts and acted proactively during stable performance.

For organizations restructuring in response to AI - building new operating models before the competitive pressure becomes existential rather than after - this is a good argument for acting now. Not because there’s a prize for moving early but because the alternative carries a larger cost.

2. The redesign has to go deeper than the org chart

This is the single most consistent theme across both consulting and academic evidence, and it maps directly to what practitioners observe repeatedly in the field. Structural changes without corresponding changes to processes, decision rights, incentives and behaviours fail to deliver.

Burton, Lauridsen and Obel (2002) found that even a single misfit between structure and its contingencies produces measurable ROA loss. A new org chart drawn on top of old decision rights and old incentive systems isn’t a redesign. It’s a reorganization of a system that will continue producing the same outputs.

3. Procedural fairness is a commercial consideration, not a ‘soft’ one

Van Dierendonck and Jacobs (2012) found a strong correlation between fairness and affective commitment after restructuring - for both the people who stayed and the people who left.

For those of you interested – affective commitment is the degree to which an employee genuinely wants to be part of the organization - not because they feel they have to stay, or because leaving would be financially painful, but because they feel emotionally connected to what the organization is trying to do and believe in it. It’s the difference between an employee who shows up because they want to and one who shows up because the alternative is worse. In a restructuring context, it matters because it’s affective commitment - not normative or continuance commitment - that predicts whether people will go above and beyond, absorb the disruption constructively and still be genuinely engaged on the other side of the transition.

The nuance that most organizations miss is that procedural justice (how decisions were made and communicated) matters more than distributive justice (who got what outcome). People can accept outcomes they don’t like when they believe the process was fair. What they don’t forgive is opacity.

The practical detail the researchers emphasise explicitly - survivors judge procedures as fairer when their direct supervisor delivers the message, not a centralized HR function. The urge to manage consistency centrally and efficiently tends to strip out the thing that actually builds trust in the room.

Trevor and Nyberg (2008) quantified the downstream cost. Their research showed that a 1% headcount reduction predicts approximately a 31% rise in voluntary turnover among survivors. The talent loss that follows a poorly managed restructure is often larger, and more expensive, than the headcount reduction that preceded it.

4. Communication is the one factor with genuine causal evidence - and it’s looks different to what you think it would

Schweiger and DeNisi (1991) remains a compelling causal study in the change management literature. A longitudinal field quasi-experiment comparing two merging plants found that a structured, realistic communication programme causally reduced uncertainty, dissatisfaction, distrust and turnover intent.

Not a polished town-hall narrative. Not a carefully managed announcement. Not the version where everyone is “cautiously optimistic.” A structured, realistic sharing of what is known, what isn’t yet known and what the process for finding out will be. That specificity matters, because many change communications plans are built around reassurance rather than reality. They manage the message rather than the uncertainty. And in doing so, they produce exactly the conditions - information gaps, distrust, rumour - that the Schweiger and DeNisi programme was designed to prevent.

Realistic communication, delivered early and consistently, is not a soft nice-to-have. It has been proven to improve restructure outcomes.

5. Integration depth matters more than integration speed

King, Wang, Samimi and Cortés (2021, 220 studies, 19 variables) found that integration depth significantly predicts managerial-assessed performance. Integration speed, by contrast, was non-significant when pooled across studies.

The “move fast and decisively” prescription that can circulate in executive conversations is not supported by the meta-analytic evidence. The supported prescription is to match depth to the logic of the specific change - deeper where operational coherence or synergy is the goal, lighter where capability needs to be preserved.

This is saying that whether you move fast or slow doesn’t independently predict whether you succeed. Speed is not the variable that matters. What matters is how thoroughly you’ve actually redesigned the thing - how far the integration goes into the real operating model, not just the surface structure.

Here’s the thing - speed is the wrong question entirely. Organizations spend enormous energy debating how fast to move. The question worth asking isn’t “how quickly are we integrating?” but “how deeply are we actually redesigning - and is that depth matched to what this specific change needs to deliver?”

BUT - deeper isn’t always better. Deep integration when you’re trying to preserve an acquired capability - say, an innovation team whose value comes from operating independently - destroys the thing you paid for. So focus on deliberateness over pace - have you thought through what level of redesign this change actually requires, and are you doing that level of work? Speed becomes relevant only after that question is answered.

So, if your restructure conversation is primarily about timeline and pace, you’re probably optimising for the wrong variable.

6. Codified learning beats accumulated experience

Zollo and Singh (2004) found that deliberate codification - building explicit methodologies, templates and structured approaches from prior change experience - substantially improves outcomes and is more important than tacit experience alone.

Having done several restructures does not predict success. Having extracted and systematically learned from those restructures does.

This distinction matters more than it might appear. Most organizations treat each restructure as a standalone project - they bring in external support, run the process, and move on. The expertise walks out with the consultants. The organizational memory of what worked and what didn’t lives in the heads of the people who were closest to it, degrades as those people move on, and has to be rebuilt from scratch the next time. That pattern is not a resource problem. It is a design choice - and the research suggests it is a costly one.

The alternative is treating restructuring capability as something the organization builds and owns permanently. Not a checklist. Not a framework borrowed from a consulting firm’s methodology library. A genuine internal capability: the diagnostic tools to identify where the design problem actually sits, the structured approaches to address it, the accumulated institutional learning about what works in this specific organizational context, and the expert support to apply judgment when the situation doesn’t fit the template.

The organizations that get genuinely faster and more effective at structural change are not the ones with the biggest transformation budgets. They are the ones that decided to own the capability rather than rent it.

(Side note – this is the reason that I built Orby – because in a world that requires your business to adapt constantly, it should be easier to build and maintain the internal organizational design and change capability that compounds over time. Each restructure less disruptive than the last. The learning becoming an asset rather than an overhead.).

7. Human capital is a strategic asset, not a cost line

Cascio, Chatrath and Christie-David (2021, 37-year NYSE panel) found that pure employee restructuring among financially healthy firms generally destroys value, while asset restructuring outperforms it on average.

The talent that leaves in the wake of a poorly managed structural change carries institutional knowledge, judgment, and relationship capital that doesn’t appear on a balance sheet and can’t be rebuilt quickly. In AI-driven operating model redesigns - where the judgment about what to automate, what to protect, and how to sequence change is critical - this loss is particularly costly. AI is trained on what has been. The people who understand why the organization works the way it does are the ones who know which of those patterns are worth preserving IN THE FUTURE.

This has a direct design implication that most restructures don’t consider. If the primary mechanism for capturing efficiency is reducing headcount, the organization is simultaneously reducing the human capital that makes the remaining structure function. The short-term cost reduction is visible and immediate. The capability loss is invisible and gradual - until it isn’t. The roles that look redundant on a org chart are often the ones carrying the informal operating model - the workarounds, the relationship networks, the institutional memory of why certain decisions get made the way they do. Strip those out and the formal structure that remains is operating without the connective tissue that made it work.

8. What do the consulting houses highlight as best practice?

I’ve already caveated that the consulting houses research can’t be interrogated in the same way as the academic peer-review research. But there are still excellent insights to be gained from the people who spend a majority of their time supporting businesses to deliver new ways of working.

Restructures that are scoped broadly - addressing operating model, roles, decision rights and ways of working rather than headcount alone - consistently outperform those that focus primarily on cost reduction.

Leadership behaviour during the transition, particularly visible sponsorship and consistent communication from senior leaders, is identified across virtually every major consulting study as a primary predictor of whether the change holds.

Capability building (investing in the skills and structures that enable the organization to absorb and sustain the change) is consistently identified as the factor most frequently under-resourced relative to its impact.

And the organizations that treat AI-driven redesign as an operating model question rather than a technology implementation question are consistently identified as the ones generating durable competitive advantage.

Where these findings align with the peer-reviewed academic evidence - on breadth of redesign, communication quality, and capability retention – I’d suggest they strengthen the case considerably. Where they diverge - particularly on speed of implementation, where consulting firms tend to recommend moving faster than the meta-analytic evidence supports, perhaps the academic literature is the more ‘independent advisor’.

9. Are you clear on what ‘good looks like’?

This one is not from the research directly…. But it is from implied in many of the articles so it’s critical to call it out.

Organizations spend months debating HOW to restructure. They argue about timing, communication plans, org charts. What almost never gets settled clearly is what success actually looks like. And that absence is costly.

The research backs this up in a frustrating way. “Success” is variously operationalised as short-run shareholder returns, sustained operating performance, cost targets achieved, synergies realised, speed of stabilisation, employee retention and commitment, or strategic intent delivered (trust me, there’s more definitions) - and these measures frequently point in different directions.

So a restructure that looks successful by one criterion can simultaneously look like a failure by another, yet most practitioner claims (and a significant share of consulting-house benchmarks) collapse these competing definitions into a single pass/fail verdict without disclosing which measure they’ve working towards.

This definitional looseness is not just a technical inconvenience - it is the reason that headline “failure rate” statistics (including the widely cited but poorly sourced claim that 70% of transformations fail) are essentially meaningless. Fail at what? Over what timeframe? Compared to what alternative? Without knowing what they’re measuring, the statistic tells you nothing actionable.

So, before you design your next restructure, get brutally clear - What does success look like for YOUR organization? Not in general. For you, specifically. Then design everything around delivering that outcome - and defend that definition when people try to shift the goalposts mid-process.

Because if you don’t know what you’re aiming for, you won’t hit it.

One last point - No one has really researched whether the right org design was chosen in the first place

Everything covered in this article addresses how to execute a restructure well. The timing, the depth, the fairness, the communication, the capability-building. What the research does not address is whether the structural design you’ve chosen is the right response to the problem you’ve diagnosed in the first place.

The closest the literature comes is contingency theory - the idea that structure should fit the organization’s environment, strategy, size and technology (we love alignment and fit-for-purpose for good reasons). Burton, Lauridsen and Obel (2002) provided the most concrete evidence of this, finding that even a single misfit between structure and its contingencies produced measurable ROA loss. That tells you misfit is costly. It doesn’t tell you whether organizations correctly identified their misfit before restructuring, or whether the design they chose actually resolved it.

The empirical research simply assumes a restructure has been decided and asks how to execute it well. The starting question - should you be restructuring at all and is your design the best available response - is treated as the province of strategy and judgment rather than empirical research. There is no published meta-analysis, and very limited primary empirical work, that directly examines whether organizations correctly diagnosed their structural problem before acting, or whether the design adopted was the optimal available choice.

So while the restructuring research can meaningfully improve your odds of executing a restructure well, it cannot tell you whether the restructure you have chosen to execute is the right structural response to begin with.

We’ll pick this up in another essay in the future.

What this means for restructuring in an AI context

Businesses have been restructuring forever. But never at the same rate and with the high levels of visibility that we are experiencing now.

There’s two reasons you should want your business to restructure successfully.

Firstly because you can’t afford to waste the time and resources on something that doesn’t give you better business performance.

Secondly, because your people deserve better – if you are going to change how they work and whether they continue to work for you, it’s worth knowing you’ve tackled it well.

Plus, AI is creating genuine pressure to redesign how organizations operate. Decision rights that made sense in one structure may not make sense in another. Roles built around tasks that can be automated need to be rebuilt around the judgment and coordination that can’t. The pace of change is accelerating, which means the organizations restructuring right now are restructuring into more uncertainty than any previous generation of leadership teams has managed.

The research is clear on what that context demands - proactive rather than reactive timing, broad redesign rather than narrow cutting, procedural fairness from direct supervisors rather than centralized communications, depth matched to purpose rather than speed for its own sake and deliberate capability-building that turns this restructure into organizational learning rather than an isolated event.

The leaders who will navigating this well will do it deliberately - diagnosing the specific design problem before reaching for the structural solution and building the internal capability so the next one is less disruptive than this one.

That’s means not treating change as a project outcome. Treating it as an organizational capability. Building the diagnostic muscle, the design methodology, and the change capability as a permanent internal asset rather than a periodic intervention.

Because the alternative - lurching from one restructure to the next, each one starting from scratch, none of them quite delivering - is not a strategy. It’s just expensive.

Additional resources

DM me if you’d like access to either of the following:

• Full research source list - 250+ sources across consulting, peer-reviewed academic literature, meta-analyses and systematic reviews. Evidence strength ratings and direct DOI links throughout.

• Research limitations – how confident you can be about the findings reported in the papers.

The pattern the research describes - where organizations that build genuine internal change capability outperform those that call in external support for every major transition - is exactly the gap Orby is built to close. Not as a consulting engagement that arrives, delivers and leaves. As an ongoing system for building the internal organizational design and change capability that compounds over time. If you’re navigating a restructure right now and want to understand how Orby works, heyorby.com is the starting point - or reply directly and we can work through where you are.