Speed Is the Easy Half: What the Consensus Debate Misses About AI-Era Transformation

A recent Harvard Business Review article by Jonathan Rosenthal and Neal Zuckerman makes an argument we hear more and more in boardrooms: consensus-based decision-making is too slow and too distorting for the AI era, and the companies that win will be the ones with the courage to change how decisions get made. They are right. Consensus dilutes accountability, launders bad information into safe-sounding agreement, and turns weeks into quarters. But it misses the harder half of the problem.

Why do most AI-era transformations fail?

Most AI-era transformations fail at execution and embedding, not at the decision. When a value-creation plan stalls, the post-mortem almost never reads "we decided too slowly." It reads "the decision never made it into how the business actually runs." The org chart changed and the behavior did not. The new operating model lived in a deck, not in the weekly cadence. The AI tool got procured and then sat unused because no one rewired the process around it. [QP proprietary data point to insert here: e.g. "Across the transformations we have tracked, X% stalled at embedding rather than at the decision."]

Is faster decision-making enough to win in the AI era?

No. Faster decision-making is now table stakes, not a source of advantage. Cleaner decision structures produce more decisions, but they do not on their own produce more results. A company that decides quickly and executes poorly simply fails faster and at greater volume. That is not agility. That is churn with better optics. The durable advantage is the speed at which a decision becomes embedded behavior, repeatable and durable, without the founders or the consultants in the room.

What does the consensus critique get wrong?

The consensus critique is correct but incomplete in three ways.

First, concentrating authority is necessary but not sufficient. Naming a single owner, as frameworks like OVIS prescribe, fixes who decides. It does nothing for who executes, how progress is measured week to week, and what happens when the plan meets reality. Ownership of a decision and ownership of an outcome are different jobs, and most organizations conflate them.

Second, removing the dissenter is not the same as removing the friction. The point of consensus, for all its cost, was to surface objections before they became failures. Stripping that out without a deliberate replacement does not eliminate the risk. It moves the risk downstream, where it is more expensive to find. AI makes this worse, because a confident model can quietly displace the colleague who would have pushed back. Speed without a mechanism for productive dissent is just faster overconfidence.

Third, AI raises the stakes on embedding, not on deciding. The marginal cost of generating options, analyses, and recommendations is collapsing toward zero. Decisions will get cheaper and faster whether or not leaders restructure for it. What does not get cheaper is the organizational work of changing how people operate. As decisions accelerate, embedding becomes the binding constraint. It is the part of the system that cannot be automated, and therefore the part that determines who actually captures the value.

How should a company structure transformation in the AI era?

Run transformation across three stages, not one: Plan, Execute, Embed.

Plan is where the consensus debate lives. Build a sharp value-creation thesis and a decision structure that puts a clear owner on each move. Speed matters here, and so does designing in the right friction so that real objections get heard once, early, rather than relitigated forever.

Execute is where most plans go to die. Turn the thesis into a cadence: clear owners, hard metrics, short cycles, and visible progress a board can actually track. The test is simple. Can the organization show movement in weeks, not quarters?

Embed is the stage almost everyone skips, and the one that decides whether the value is real or temporary. A change is embedded when it survives the departure of the people who designed it, when the new behavior is how the business runs by default rather than a process maintained by force of will. This is the difference between a transformation and a project.

What is the bottom line for leaders?

Faster decisions are table stakes, available to anyone willing to restructure. The durable advantage belongs to the organizations that can take a decision and make it stick, at speed, repeatably, without the people who designed it standing over it. Rosenthal and Zuckerman are correct that legacy decision-making will not survive the AI era. We would put it more pointedly: decide faster, yes, but the prize goes to whoever embeds faster.

Quadrillion Partners works with private equity sponsors, public company operators, and management teams to plan, execute, and embed value creation. If you are navigating an AI-era operating model change, we should talk.

‍

‍