This spring, Wakefield Research surveyed 300 C-level executives at U.S. companies with at least $500 million in revenue, on behalf of the AI consultancy KUNGFU.AI. A big majority, 74%, said that they had projected more confidence in their AI strategy than they actually felt.
Odds are that three in four of the executives presenting AI plans to your board and committing your capital to them were performing a conviction they did not hold.
That is “fake it till you make it” dressed up as executive leadership. It is easy to understand why, though. For most of a leader’s career, confidence was a reasonable stand-in for competence. If you understood the business and spoke with conviction, you were thought to be demonstrating mastery.
AI changed the conditions of leadership.
There is no established playbook for many of the decisions companies are making now. The tech is moving faster than most organizations can train people to use it, and the risks that come with it are still emerging. The business case is often unproven.
Your talent system still rewards the leaders who sound like they know all the answers. The leader who says, “I don’t know yet. Here’s how we are going to find out,” inspires less confidence than the leader who sounds certain. That certainty may rest on nothing anyone has tested.
Your talent system still rewards the leaders who sound like they know all the answers.
Your organization may be buying into confidence over evidence.
A few weeks ago, a board member of a publicly traded company asked me which AI tools his company should be evaluating. Nobody had yet named the business need those tools were supposed to fill. Tool first, need second.
That sounds irrational, but it makes sense once you look at what a tool purchase does for the buyer. Buying a tool feels like progress. It provides a roadmap of steps to take, a potential leap over the competition, or something concrete to check off a list of concerns.
And the alternative is a much harder sell. Start with an unresolved business problem and admit what you do not know. Build something small and put it in front of users. Change direction based on where it breaks.
That work is roll-up-your-sleeves messy, done in stages, with testing and revision at each one. In the room, it reads as hesitation.
The same dynamic explains why so many companies wait for competitors to go first. Let another group spend the money and absorb the failures, then adopt whatever they prove works. That feels prudent and unpunishable.
But it also puts a ceiling on what can be achieved. A best practice is, by definition, something a competitor has already proven. If your strategy is to wait for proof and then copy it, your best possible outcome is parity.
Certainty also arrives for everyone at the same time. The moment that a move is safe enough to copy, everyone can see it.
If your strategy is to wait for proof and then copy it, your best possible outcome is parity.
Even the certainty being purchased is often counterfeit. In the same Wakefield survey, 79% of executives said that they would fundamentally change how their organization invested in AI if they could start over. Nearly four out of five admitted quality assurance or testing steps were shortened or skipped to move faster. Leaders are projecting confidence they do not truly feel, while organizations sometimes remove the very testing that would show whether that confidence is justified.
This is what fake-it-till-you-make-it performance looks like when it reaches the P&L. Capital gets committed at full scale before the organization has earned the confidence to execute it. Then customers run into the problems the executive team did not take the time to test for.
The gap between declared and demonstrated capability runs the full height of the organization. In April, the learning platform Acorn surveyed 1,224 professionals at organizations with 1,000 or more employees and found that 77% of executives believe their managers are prepared to guide AI capability conversations. Only 9% of individual contributors agree. A gap that size is a warning that capability is being reported at the top long before anyone has asked to see it demonstrated.
Your talent systems may be making the same mistake. Ratings, performance reviews, and succession plans were all built around a world in which the right answer existed. They show who is good at executing a known playbook. They tell you much less about who can lead when the playbook no longer applies.
AI is where that gap shows up. The leader who always looked decisive may simply be very good at executing known answers. The leader who is willing to admit that there is more to learn is more valuable now.
The senior HR leaders closest to this can already name the pattern, even when their systems cannot. A chief people officer at a Fortune 500 company recently told me which leaders concern him most when it comes to AI: the ones who plant a flag and commit before they have enough information to justify the commitment.
You can usually spot performed confidence once you know what to look for. The plan has an answer for everything. There are no meaningful unknowns, no assumptions recommended for testing, and projections precise enough to suggest certainty when none exists.
Ask what evidence would make the team stop or change direction. If the answer is more effort rather than a named result, you are hearing commitment, not evaluation. Listen to the vocabulary as well. When the presentation runs on the vendor’s terms instead of your company’s operating terms, the confidence was bought, not built.
Your company already knows how to test claims. Talent acquisition teams do it every day. If a candidate says that they have AI expertise, you do not take that statement at face value. Ask what the candidate built, what it produced, what went wrong, and what they learned.
Then ask why an eight-figure AI investment has a lower bar to meet. It should not.
Apply the recruiter’s standard to your executives. Before the next stage of money is released, require that something has actually been built. Start from a business need you have named, build the smallest working version that touches it, and come back with what it produced and what it taught you. That is a third option between betting big and waiting for a competitor to go first: small builds that show what works before the full budget is committed.
The size of the build is irrelevant. A leader who has put something real in front of users can tell you exactly where it broke, what that cost, and what needed to change. A leader who has only bought a plan cannot. Make evidence the price of the next round of funding.
The test fits inside a meeting you already run. In the next AI review, pick the initiative’s biggest unknown and ask the team to explain it. Then watch what happens.
Honest uncertainty gets more specific under pressure. Performed conviction gets more emphatic or more abstract. The leader with real footing reaches for a number, a date, or the test already underway. The performer reaches for adjectives.
Your leaders are watching what gets rewarded. Reward certainty and you will get more certainty. Fund the initiative that came back with evidence, and state clearly that the evidence is why you funded it.
Honest uncertainty gets more specific under pressure. Performed conviction gets more emphatic or more abstract.