Started Too Late, Then Played It Safe: What Boards Owe the Next CEO

Ask a board whether the new CEO can do the job and 83% say yes. Ask the same board whether that CEO is actually connected to them and trusted by them, and 10% say yes. Those two numbers came out of the same survey, from the same directors, about the same people. The 73-point space between them is where most succession failures are actually built.

The numbers come from Korn Ferry's 2026 CEO & Board Survey, released 21 July 2026 — 250 board directors and chief executives surveyed in April 2026, across seven industries, at organisations of 1,000 to 20,000+ employees. It is a small sample and Korn Ferry sells succession consulting, so read it accordingly. But the internal contradictions in it are the useful part, and they are not the kind a vendor invents.

Started too late, then played it safe

Two more findings from the same survey sit next to each other and explain a great deal.

50% of boards say succession planning started too late during their last CEO transition. And 50% say they played it too safe in that same succession decision.

Korn Ferry does not publish the cross-tabulation, so we cannot say those are the same boards. But the causal direction only runs one way. A board that starts late has compressed its own option set, and the only candidate you can underwrite in a compressed window is the one who looks least surprising on paper. Late does not merely produce a rushed decision. It produces a conservative one — because the safe choice is the one that requires the least explanation to the people who did not have time to do the work.

The mechanics behind "too late" are in the same data. 60% of boards agree succession works best as an ongoing process. Only 17% review succession plans quarterly. Nearly everyone knows the right cadence; roughly one in six runs it. And 70% say they did not gather enough 360-degree feedback during the last transition — which is to say seven in ten boards selected a chief executive without systematically asking the people who had worked alongside that person what it was like.

A board that starts late does not just decide in a hurry. It decides conservatively, because conservative is what you can defend without evidence.

Promoting internally does not solve it

The obvious response is that this is an external-hire problem, and the fix is a deeper internal bench. The succession data says otherwise.

The Conference Board, with ESGAUGE, Semler Brossy and Egon Zehnder, tracks CEO successions from SEC Form 8-K filings across the Russell 3000 and S&P 500. In its 2025 edition, 65% of Russell 3000 successions were internal, up from 62% the prior year. Internal promotion is already the majority path. It has been for years.

So the 10% trust figure is not describing outsiders parachuting into unfamiliar boardrooms. In most cases it is describing a person the company has employed for years — and the board still does not know them. That is the distinction almost everyone misses. Being internal to the company is not the same as being known to the board. A CFO can spend eight years presenting quarterly numbers to directors and remain, to those directors, a competent presenter of quarterly numbers. Presenting is not the same as disagreeing. Nobody learns how a person handles being wrong in front of the people who can fire them until it happens.

This is the same failure we priced in The 11% Résumé, moved up a floor. There, search processes measure technical competence — the source of 11% of hiring failures — because it is the part that is easy to see. Here, boards measure track record, because a track record is the part of a candidate that shows up in a board book. In both cases the organisation screens hardest for the variable it can observe, and then discovers the outcome was decided by the one it never tested.

The bench that was flattened out of existence

There is a structural reason the internal option keeps disappointing, and it has nothing to do with succession planning at all.

We covered it in The Missing Middle Rung: span of control moved from 8.2 direct reports per manager to 12.1, which removes roughly a third of first-line manager seats for the same headcount. That rung is where people first learn to be accountable for someone else's work. Remove it and the pipeline still looks full — the same number of directors and VPs exist — but the directors arriving at the top have fewer years of having actually run anything.

In private equity the same shortage shows up as a buying pattern rather than a building one. Altrata's portfolio company data, which we covered in The External Default, puts 74% of US portfolio leadership roles as externally filled. And when the external hire is made against a thin brief, the failure lands on a predictable fuse — the year-two exit.

The AI gap the board has not staffed

One more finding, and it is the one most likely to matter in the next 24 months.

49% of first-time CEOs are confident in managing AI and emerging technology risks. Only 30% of board members say the same.

Read that as a governance fact rather than a technology one. Fewer than half of new chief executives believe they can manage the risk, and the body that oversees them is less confident still. There is no one in the room with conviction on the subject. In that configuration the default outcome is not a bad decision — it is deferral, which in a category moving this fast is itself the decision. It is the same ownership vacuum we mapped in Five Questions That Decide Whether Your AI Spend Becomes Revenue, and it is the most common reason a fractional Chief AI Officer gets hired: not to build models, but to give the board and the CEO a named person whose answer on AI risk they can actually rely on.

The Connection Ledger: five entries a board owes the next CEO

Confidence in capability is cheap and boards have plenty of it. Connection is expensive and boards have almost none. The fix is not more assessment. It is five specific things written down before the vote — what we call the Connection Ledger.

1. The disagreement record. Name a moment where this candidate and the board — or the chair specifically — held opposing positions and worked it out. Not a presentation. A disagreement. If no such moment exists, the relationship is untested, and 83% confidence is confidence in a résumé. Manufacture the moment before the appointment if you have to: give the candidate a live, contested question and watch.

2. The 360, gathered from below. Seven in ten boards say they did not collect enough. The fix is not volume, it is direction. Peer and subordinate input surfaces behaviour under pressure; superior input mostly confirms results. Ask three people who reported to this candidate what happens when the candidate is wrong.

3. The phase call, in writing. Does this company need someone to transform it, or to multiply what already works? These are opposite jobs and they select for opposite people. A leader who has only ever multiplied a working model will not survive a transformation, and a serial transformer will break a company that simply needed to be run well and grown. Write down which one you are hiring for, and which one the candidate has actually done. If the board cannot agree on the answer, the board is not ready to appoint.

4. The month-24 seat. Describe the job as it will exist two years out — the revenue, the headcount, the ownership structure, the regulatory surface. Then ask whether this person fits that seat. Most succession decisions are underwritten against the company as it is on the day of the vote, which is the one configuration guaranteed not to persist.

5. The quarterly review, on the calendar. 60% believe it should be continuous; 17% do it quarterly. Put four dates in the calendar for the next year and give one director the job of holding them. This is the cheapest entry in the ledger and the one that prevents the other four from being needed in a hurry.

None of this is an argument that boards are choosing badly. 83% confidence in a first-time CEO's ability to lead is probably close to right — these are, by and large, capable people. The argument is that capability was never the variable in question.

What the Korn Ferry data describes is an industry-wide habit of underwriting the part of the decision that is visible and skipping the part that determines the outcome. Half of boards started too late. Seven in ten skipped the feedback. One in six runs the cadence they say is correct. And then one in ten can say the person they just appointed is genuinely connected to and trusted by them.

The remedy is not a better assessment product. It is fewer transitions, understood far more deeply — the same discipline that separates a search built on a real brief from one built on a job description, and the same reason the first 90 days after signing decide more than the search that preceded them. A board that can write all five entries of the Connection Ledger before the vote does not need to hope the trust arrives later. It has already been built.