Why Leaders Want It, Why They Don’t Get It, and What Actually Works

Ask any executive what their organization needs more of, and “accountability” will land near the top of the list. Ask the same leaders whether they’re good at it, and the answer collapses. In a study of more than 40,000 participants, 91% ranked the ability to hold others accountable as one of the top leadership development needs in their organization. In the same study, 82% admitted they had limited to no ability to actually do it. Only 18% of people can successfully hold others accountable for delivering on what was promised.

That gap, between how much accountability is wanted and how often it’s delivered, is the story modern workplace research keeps telling. And the reasons it persists are not about lazy employees or weak leaders. They’re about how the system is designed and what the word itself has come to mean.

Accountability Has a Definition Problem

For most employees, “accountability” is a synonym for “getting in trouble.” Leaders use the phrase “lack of accountability” the way they’d describe a missing piece of equipment, and the message lands as a threat. People learn to avoid visibility rather than seek it.

Harvard Business School professor Amy Edmondson has spent two decades arguing for a different definition. Accountability, in her framing, is psychological ownership: a person’s internal sense of responsibility for the outcome of their work. It’s not a score someone else assigns to them. It’s what they bring to the work themselves.

This reframing matters because it dissolves a false trade-off that runs through most leadership advice. Psychological safety is often treated as the soft alternative to accountability, the thing that gets in the way of holding a high bar. Edmondson’s research is the opposite. Safety and accountability are allies. High accountability with low safety produces anxiety and burnout. High safety with low accountability produces a comfort zone with no performance. Only the combination of the two creates what Edmondson calls the learning zone, where teams consistently perform at a high level.

What Working Accountability Systems Have in Common

The organizations that actually pull this off don’t all use the same framework. Google runs on OKRs. AB InBev and Intel use FAST goals. Healthcare systems and public agencies often use balanced scorecards. Underneath the branding, the mechanics are the same. Five ingredients show up across the research.

The first is clear, written, public expectations. MIT Sloan’s Donald Sull, drawing on metadata from more than 600,000 goals, found that when employees are given the option, more than 90% choose to make their own goals public. Transparency is preferred, not feared, as long as the system around it is fair. Sull’s FAST framework (Frequently discussed, Ambitious, Specific, Transparent) is positioned as the research-backed alternative to SMART goals, specifically because SMART goals are usually set in private and reviewed once a year, which the data shows is far too infrequent to drive accountability.

The second is frequent, low-stakes check-ins. Gallup research finds that employees who receive regular, meaningful feedback are four times more likely to be engaged. Harvard Business Review’s reporting on this is direct: the annual performance review is the wrong unit of time. People need course corrections in weeks, not quarters.

The third is commitments owned by the person doing the work. Edmondson’s research shows that accountability is durable only when the person internalizes the commitment. Leaders who assign tasks without involving people in shaping them get compliance at best and quiet resistance at worst. One practical move documented by the NeuroLeadership Institute: have the team member restate the commitment and the by-when in their own words before the meeting ends.

The fourth is fairness in how performance is evaluated. Ron Carucci’s research, published in Harvard Business Review, found that when people perceive the accountability system as fair, they are four times more likely to be honest about their own mistakes, act fairly toward others, and serve the organization’s interests instead of their own. When the same system feels unfair, those behaviors collapse and information-hiding spikes. Fairness is built through consistent criteria, documented decisions, and visible consequences that apply equally, including to senior leaders. Inconsistent enforcement is one of the fastest ways to kill an accountability culture that was previously healthy.

The fifth is coaching conversations instead of blame conversations. The NeuroLeadership Institute’s joint research with David Rock and Amy Edmondson identifies three habits common to leaders who build accountability without fear. They think ahead about what success looks like. They obsess about commitments, not blame. They anchor on solutions when something goes wrong.

Where Organizations Most Often Fall Short

The pitfalls are rarely about bad people. They’re about predictable system failures. Treating accountability as scorekeeping is the first. Carucci puts it bluntly: the scorekeeping nature of accountability creates a built-in negativity bias, where leaders reflexively hunt for shortfalls. The result is a culture that punishes honesty about mistakes.

Vague or constantly shifting expectations come next. The Workplace Accountability Study found that only 15% of leaders successfully define and broadly communicate their key results, and 30% of employees report priorities shifting frequently enough to create real confusion. You can’t be accountable to a moving target.

Feedback only when things go wrong is another. In the same study, 80% of participants reported that feedback happens to them only when something is broken. That trains people to stay out of sight, not to step into it.

Inconsistent application is the quiet killer. Even when a system isn’t explicitly biased, the perception of unequal treatment alone is enough to trigger the fourfold drop in honest behavior that Carucci documented. And accountability cultures die fastest when employees see senior leaders held to a lower standard than everyone else.

A subtler failure is mistaking psychological safety for low standards. Edmondson and Michaela Kerrissey of the Harvard T.H. Chan School warned in 2025 that many leaders read about psychological safety and conclude it means lowering the bar. It doesn’t. Safety without accountability produces complacency. The goal is both at once.

Then there’s avoidance, which is the most common failure mode of all. 82% of managers report they either try and fail at the accountability conversation or simply skip it. The conversation that doesn’t happen is the one that does the most damage.

The 2026 Wrinkle: AI in the Accountability Loop

AI tools are now being used to track productivity, generate performance summaries, score employee output, and even draft accountability conversations. The research is early, but a few patterns are already consistent.

The first is that AI raises the stakes on system design rather than lowering them. Wharton’s Kevin Werbach, who runs the school’s Accountable AI Lab, argues that accountability has to be designed into AI systems from the start, not bolted on later. The same principle applies when AI is evaluating people. If the underlying performance system is unfair, AI will scale that unfairness, not fix it.

The second is that biased historical data produces biased AI judgment. Legal and HR research published in early 2026 documents that AI performance tools trained on past evaluations inherit any existing bias, including biased reviews, unequal access to development, and historic underrepresentation, then apply that bias at scale unless leaders actively clean the data and validate the models.

The third is a visibility gap that’s widening fast. McKinsey’s 2025 workplace AI research found employees are three times more likely to use AI for 30% or more of their work than their leaders realize. Managers are increasingly evaluating output without understanding how it was produced, which breaks the accountability conversation in both directions.

The fourth is that monitoring without dialogue erodes trust. When AI systems make consequential employment decisions without transparency, employees report feeling powerless and alienated. The legal community has converged on a clear principle: AI can inform decisions, but it can’t replace human accountability for those decisions.

The practical guardrails follow from the research. Be transparent about which AI tools are being used in performance evaluation and what they measure, because silence creates suspicion. Keep a human in the loop. Audit for disparate impact. Train managers in AI literacy before asking them to act on AI-generated insights, including how to push back when the tool is wrong. And use AI to remove friction from the conversation, not to remove the conversation itself. Drafting talking points,
summarizing prior commitments, and surfacing patterns are legitimate uses. Auto-firing people based on a dashboard score is not.

The Bottom Line

Accountability isn’t a personality trait. It’s a system, and one that most organizations are designing badly. The leaders who get it right share a few things in common: they make expectations clear and public, they check in often enough to course-correct, they treat commitments as something the team owns rather than something delivered from above, they apply the standard consistently, and they coach more than they punish. They also pair all of that with psychological safety, because without it, the system collapses into fear and silence.

The research has been pointing in this direction for years. What’s changed in 2026 is that AI is now sitting inside the loop, and the leaders who don’t think carefully about how it’s being used will scale their existing problems rather than solve them.