Ask most leadership teams whether they are built to run the existing business well or to explore new ones, and you will usually get a confident answer: one or the other. “We’re an execution culture.” “We’re an innovation shop.” It feels disciplined to pick a lane. It is also, according to three decades of organizational research, one of the more expensive mistakes a company can make.

In 1996, Michael Tushman and Charles O’Reilly published research on what they called the ambidextrous organization — a company structured to do two contradictory things at once: exploit its current business (refine, scale, extract efficiency from what already works) while simultaneously exploring new ones (experiment, prototype, tolerate failure in pursuit of what might work next). Their later work, including the 2004 Harvard Business Review article “The Ambidextrous Organization,” showed that the firms which survived major technology and market shifts weren’t the purest exploiters or the boldest explorers. They were the ones that ran both modes at the same time, deliberately kept them organizationally separate, and had senior leadership hold the tension between them rather than resolve it.

Why “pick one” feels right and is still wrong

The instinct to choose a single operating mode is understandable. Exploitation and exploration reward opposite behaviors. Exploitation wants tight processes, low variance, and a bias toward “do it the way that already works.” Exploration wants loose processes, high variance, and a bias toward “try the thing that might not work at all.” Put both under one manager with one set of KPIs and one culture, and the stronger, more measurable discipline — exploitation — wins almost every time. New ideas lose the resourcing fight to this quarter’s numbers, not because anyone decided innovation didn’t matter, but because the organization was never built to protect it.

This is the trap Tushman and O’Reilly’s research was aimed at. Separating exploration and exploitation isn’t a philosophical nicety — it is a structural requirement. Without deliberate separation, the core business will quietly starve the new one every time, because it is easier to measure, easier to defend, and easier to explain to a board.

What structural ambidexterity actually looks like

In practice, the research points to a few concrete moves, not a mindset shift:

Separate units, shared leadership. The exploratory unit gets its own structure, processes, and culture — different hiring criteria, different incentives, even a different physical location if possible — while still reporting into the same senior team as the core business. The separation protects the new venture from being judged by the old business’s metrics; the shared leadership keeps it from drifting so far that it can’t be reintegrated later.

Different metrics for different horizons. A new venture evaluated against the core business’s margin targets will be killed in its first review cycle. Exploratory units need their own success measures — learning velocity, customer validation, option value — not a scaled-down version of this quarter’s P&L targets.

Senior leadership as the integration point. The tension between “run efficiently” and “explore boldly” cannot be resolved at the team level — it has to be actively managed at the top. The leaders who do this well treat the contradiction as a permanent feature of the job, not a problem to be solved once and filed away.

Where this shows up in change and improvement work

This is also why so many process-improvement and operational-excellence programs quietly undercut a company’s own innovation ambitions. A relentless focus on standardizing, eliminating variance, and driving out waste — the core discipline of lean thinking — is exactly right for the exploit side of the business. Applied uncritically to a new product team or an early-stage venture, the same discipline strangles the experimentation that venture needs to find its footing. The skill isn’t choosing lean thinking or entrepreneurial thinking. It’s knowing which parts of the organization need which discipline, and building the structure that lets both run at full strength without one quietly cannibalizing the other.

For a leadership team weighing where to invest change-management effort, Tushman and O’Reilly’s work offers a useful diagnostic question: is a given part of the business being managed — and measured — in a way that matches what it’s actually trying to do? A new venture held to core-business metrics, or a mature business unit run with start-up-style informality, are both signs the structure hasn’t caught up with the strategy yet.