The most useful moments in leadership development feel one-to-one: someone responds to what you actually say and lets you try again. That level of attention has always been hard to scale. AI does not prove Bloom's two-sigma promise. It changes the economics of practice.
It is Tuesday afternoon. A manager has just told their team how AI will change a familiar part of their work. The message was clear in the manager's head. In the room, it landed as a threat.
The workshop on leading change was six weeks ago. Their next coaching session is in three weeks. What they need is not another model. They need ten quiet minutes to replay the moment, see what they missed, and try the conversation again.
That small gap between knowing and doing is where most leadership development is won or lost.
Why Bloom called it a problem
In 1984, educational psychologist Benjamin Bloom described what became known as the 2 sigma problem.
He reported two doctoral studies comparing conventional classes, mastery learning and tutoring. Across four small samples, taught in two subjects over three weeks, the average tutored student performed about two standard deviations above the conventional-class average—roughly the 98th percentile.
That was not a universal law of tutoring. The tutorial condition sometimes placed two or three students with one tutor, and it included formative checks, corrective feedback and another chance to reach mastery.
Tutoring still helps. But two sigma is better read as a possibility under unusually favorable conditions than as a promise.
Bloom called it a problem because one-to-one tutoring was “too costly for most societies to bear on a large scale.” The question was never simply, “Does personal attention work?” It was, “How can we make responsive, individual learning practical for everyone?”
The durable idea is not two sigma. It is feedback, correction and another attempt.
Leadership has the same personalization gap
A workshop can give a group shared language. A facilitator can create a moment that changes how a team sees itself. A human coach can notice the hesitation behind an answer.
Each matters. But they do different jobs.
The workshop sees the cohort. The coach sees the person. The workplace supplies the moment that actually counts.
That moment is highly specific: the high performer who shuts others down; the AI change the team quietly distrusts; the decision that is right on paper and wrong in the room. Human coaching is valuable because it can work at that level of detail. It has also usually been reserved for a relatively small group. Human attention is finite. When a program includes fifty or five hundred leaders, the personal part gets thinner.
What AI changes
AI does not make leadership simple. It makes practice more available.
A leader can now rehearse a difficult conversation before walking into it. They can ask the system to play the skeptical colleague, the defensive expert, or the board member who wants the answer in one sentence. They can pause, ask what made their response unclear, and try again.
After the real conversation, they can debrief: What did I intend? What did the other person seem to need? Where did I become defensive? What will I do differently next time?
None of that requires waiting three weeks for the next module.
At Relevance, this is not one generic chatbot. Relevance Capable® adapts the wider learning journey to the role and the individual. Persona Pro provides realistic conversations with more than 60 AI personas and immediate feedback. Around both sits a carefully accredited Relevance® faculty pool of trainers and coaches who bring context, challenge and human judgment. Together, the three make personal practice part of a genuinely blended program.
Leadership-development research already points in this direction. A meta-analysis of 335 independent samples identified needs analysis, feedback, multiple delivery methods—especially practice—and spaced sessions among the features associated with stronger leadership-training outcomes. AI can make several of those ingredients easier to provide between human sessions.
That is an inference, not proof that any AI practice platform creates a particular effect size. But it is a useful design question: what changes when individual practice stops being scarce?
A small experiment
Build your leadership learning mix
Move the sliders and watch the old one-to-one cost problem change. The Relevance blend combines Capable® pathways, Persona Pro rehearsal and protected faculty time.
A steady practice rhythm
Practice becomes frequent; human time stays focused on judgment and context.
Illustrative design math only. One Persona Pro rehearsal is set at ten minutes and one month at 4.3 weeks. “Same minutes” does not mean equal quality or equal learning impact, and excludes platform and program-design costs.
Build a Persona Pro practice brief
Pick a moment. The curveball changes each time.
Curveball
The design matters more than the chatbot
Recent AI-tutoring studies show both sides of the opportunity.
In a 2025 Harvard physics experiment, 194 students learned more in less time with a purpose-built AI tutor than with the same material in an active-learning class. But the experiment covered two lessons, and the system relied on expert-written, step-by-step answers and a platform that controlled the sequence. It was not a blank chat window.
A separate field experiment with nearly 1,000 high-school math students in Turkey found the opposite risk. A standard GPT-4 chat interface improved practice performance, but users scored 17% below the no-AI group on an unassisted exam. A teacher-guided version that offered hints instead of answers largely removed the loss, but did not produce an exam advantage.
The leadership lesson is straightforward: doing the work with AI is not the same as developing the judgment to do it yourself.
An AI practice partner should not rush to write the perfect feedback message. It should make the leader think. Ask for a first attempt. Challenge an assumption. Play the other person's perspective. Request evidence. Let the leader decide, speak, and try again.
That is why Persona Pro is built around a conversation with a persona, not an answer written for the learner. The leader still has to speak, listen and adjust.
Keep the human work human
AI can support repetition, role-play and reflection. Humans remain essential for trust, ethical judgment, organizational context and accountability. A model cannot know what was left unsaid or carry responsibility for a decision. Leaders should never paste sensitive employee details or confidential strategy into a tool not approved for that use.
A strong blend is:
Adapt the pathway
Let Relevance Capable® target the competencies each leader needs most.
Practice personally
Use Persona Pro between sessions for short, realistic conversations and immediate feedback.
Make sense with faculty
Bring patterns into live work with the accredited Relevance® faculty pool.
Prove it in the work
Apply the behavior in a real moment and return with evidence.
AI makes practice less scarce. That leaves human attention for the work that needs a human.
Where to start
Choose one real conversation from the week ahead. Remove names and confidential details. In Persona Pro, choose the closest persona and scenario, then add three things: the outcome you want, what makes the conversation difficult, and the response you are most worried about hearing.
Practice the conversation for five minutes. Do not ask for an answer to repeat. Respond in your own words.
Then ask three questions:
- Where was I clear, and where did I hide behind vague language?
- What might the other person need that I did not acknowledge?
- What is one sentence I should try differently?
Run the conversation again.
That is not a two-sigma transformation. It is ten minutes of deliberate practice that is easy to skip without a practice partner. Repeated across a program, that is where the economics start to change.
