The most known, felt problem? The people who most need practice are rarely the people who receive the most of it.

 Executive coaching can cost thousands of dollars per engagement, leadership programmes take people away from their work, and roleplay exercises often depend on finding a coach, facilitator or colleague willing to play the other side of a difficult conversation.

That is the gap AI is beginning to attack.

In February 2025, CodeSignal, originally known for technical skills assessment, expanded into leadership and communication training with an AI-powered conversation simulator. Its system uses voice AI to recreate workplace situations such as delivering feedback and resolving conflict, with an AI mentor providing feedback afterwards. The company's CEO, Tigran Sloyan, put the economics bluntly: “Leadership training often costs $20,000 to $40,000 per person for a one-month program.” For a 10-person executive team, he said, that could mean a multi-million-dollar bill, effectively keeping this kind of development at the top of the organisational hierarchy.

The more interesting question, however, is not whether AI makes coaching cheaper.

It is whether AI changes what organisations mean by practice.

The problem with learning leadership only when it counts

Leadership has traditionally been learned through exposure.

You manage a difficult employee, eventually learn how to handle the conversation, and carry that experience into the next one. You present to the board, discover where your argument fell apart, and try to perform better the next time. You negotiate a difficult deal, make mistakes, reflect afterwards and hope the next negotiation arrives before you forget what you learned.

The problem is that the highest-value situations are also the least forgiving places to practise.

CodeSignal's early product was built around a simple alternative: give people a realistic scenario, let them practise it with an AI counterpart and allow them to repeat the interaction without the social pressure of performing in front of colleagues. The platform gave users five minutes to work through difficult management conversations, with its AI mentor, Cosmo, providing guidance afterwards.

Sloyan argued that the feedback from AI roleplay was actually stronger on average than traditional human roleplay, partly because the user could practise repeatedly without feeling judged. “The fact that you have no judgment with AI and can practice as many times as needed takes the pressure off.”

That matters because repetition is the part most leadership programmes struggle to provide.

And this is where Oraton enters the picture.

The fundamental idea behind Oraton is that executive communication should not be something leaders encounter only when the stakes are real. Difficult conversations, high-pressure questions, presentations and leadership moments can be rehearsed before they happen.

Rather than treating communication as a knowledge problem, something solved by giving leaders another framework or another workshop, Oraton approaches it as a performance problem.

• You can know what a strong executive answer sounds like and still struggle to deliver one under pressure.

• You can understand that a leader should be concise and still ramble when challenged.

• You can know that difficult feedback should be direct and empathetic and still become defensive when the other person pushes back.

The difference between knowing and doing is where practice matters.

The economics change when repetition becomes cheap

CodeSignal's approach was aimed particularly at middle managers, a group that often sits in the uncomfortable space between executive-level development budgets and self-directed learning. The reported pricing was $25 per month for individuals and $39 per user per month for enterprise licences.

That difference is not merely financial.

When practice becomes inexpensive, it can become routine.

An organisation no longer has to reserve development for annual leadership programmes. A manager preparing for a difficult performance conversation can practise beforehand. A salesperson can rehearse a customer objection without testing the response on an actual customer. A new manager can run through a conflict scenario before encountering one for the first time.

That is already how the technology was being positioned by CodeSignal's customers, which were using the simulator for management training, sales training and interviewer preparation. As Sloyan explained, “We don't want sales people practicing on actual customers.”

The principle is straightforward:

Move the learning moment before the consequence.

This is also the opportunity for Oraton.

The value of an AI communication coach is not simply that it can tell someone what they should have said. Its greater potential is to give the person an opportunity to say it differently before the real conversation happens.

A leader can rehearse an answer, receive feedback, try again, change the framing, handle a harder follow-up, and then try AGAIN.

That creates a fundamentally different relationship with communication development. Instead of waiting for real-world situations to provide experience, leaders can deliberately create more opportunities to build that experience.

But AI coaching has its own evaluation problem

There is a useful irony in the second part of this story.

As companies become more dependent on AI systems, they are discovering that evaluating one AI interaction in isolation can be misleading. At VB Transform 2026, leaders from LangChain, Conviva and CoreWeave argued that an interaction can look perfectly successful on its own while revealing a serious problem when compared with a larger population.

Hui Zhang of Conviva described the problem clearly. An individual interaction might look successful, but cohort-level analysis could reveal that an AI system was asking three times as many clarification questions as normal for a particular category, while users were five times more likely to complete their purchase outside the conversation. Neither problem would necessarily be visible by examining a single transcript.

That insight is directly relevant to AI-powered coaching. A leader having one excellent practice session proves very little. The real question is whether the pattern changes.

• Does the leader become less verbose over 20 sessions?

• Does their response become clearer under pressure?

• Do they stop hedging when making decisions?

• Do they become better at asking questions instead of immediately offering solutions?

• Do they maintain composure when the simulated counterpart becomes more challenging?

• Do the same weaknesses continue to appear, even when the subject of the conversation changes?

In other words:

The unit of progress should not be the conversation. It should be the trajectory.

This is an important distinction for Oraton.

The objective of communication coaching should not be to manufacture one impressive response. It should be to help a person develop communication behaviours that hold up repeatedly, across contexts and under pressure.

Evaluation becomes part of the training itself

The AI industry is increasingly moving towards continuous evaluation rather than assuming that a system is good because it passed a predefined test.

Harrison Chase of LangChain described evaluation criteria as a living specification rather than a one-time hurdle: “Evals are like the new PRD.” Emmanuel Turlay of CoreWeave made the limitation of exhaustive testing equally clear: “I was trying to reach 100% coverage for my tests, and I still had bugs in production.”

The implication for leadership development is important. A coaching platform should not simply tell someone, “Good job.” It should create a feedback loop.

• Practise.

• Measure.

• Identify a weakness.

• Practise again.

• Increase the difficulty.

• Measure again.

Over time, the system should reveal whether the behaviour is actually changing.

That is much closer to how high-performance disciplines have traditionally developed expertise.

And it is the more interesting proposition for a platform like Oraton.

The goal is not to turn executive communication into a test that someone passes once.

It is to make communication improvement observable over time.

The AI coach is not the replacement for judgment

There is also an important limit.

The most sophisticated AI evaluation systems still face the problem of knowing whether their judgement is grounded in the real world. At VB Transform, the discussion repeatedly returned to the role of human oversight. Hui Zhang described automated evaluation as scalable but potentially ungrounded, while Emmanuel Turlay argued that consequential systems still require someone who can ultimately say, “I endorse this and I take responsibility legally for it.”

Leadership development has the same boundary.

AI can simulate the board member who challenges your assumptions.

• It can play the employee who becomes defensive.

• It can create the investor who wants an answer you do not yet have.

• It can analyse patterns across repeated practice.

But it cannot take responsibility for the decision you eventually make. That remains human.

This distinction is central to how Oraton should be understood.

Oraton is not a replacement for leadership judgment. It is a practice environment around it.

The technology can help a leader prepare for the conversation. It can expose weaknesses, create difficult scenarios and provide a structured opportunity to improve.

But the leader still has to decide what to say.

And, more importantly, what to do.

The bigger shift is from training to rehearsal

This is where the two stories come together.

1. The first is about making practice accessible.

2. The second is about making evaluation continuous and meaningful.

Together, they point towards a different model of leadership development.

Traditional training asks:

What should this leader know?

Simulation asks: What will this leader actually do when the situation becomes difficult?

Continuous evaluation then asks: Is that behaviour getting better?

That is a considerably more useful sequence. It also explains why AI is potentially more significant for leadership development than simply becoming a cheaper version of an executive coach.

Its real advantage is not that it can dispense advice. It is that it can create repeatable exposure to situations that leaders previously encountered only occasionally.

That is precisely the space Oraton is designed to occupy.

• The difficult question doesn't have to wait for the board meeting.

• The challenging stakeholder doesn't have to appear for the first time in the investor call.

• The uncomfortable employee conversation doesn't have to become the first rehearsal.

• The leader can encounter versions of those situations beforehand.

And that changes the economics of experience itself.

What this means for the future of executive coaching

The strongest model is unlikely to be humans versus AI. It will be humans using AI to increase the amount of deliberate practice surrounding human coaching.

A coach can help a leader understand why a pattern exists, challenge assumptions, interpret organisational politics and make sense of a difficult personal situation. AI can then provide the repetitions: the difficult questions, the simulated resistance, the alternative scenarios and the immediate feedback. The coach becomes less of the person who provides every repetition and more of the person who helps make those repetitions meaningful.

Oraton fits into that emerging model as the layer where communication can be repeatedly rehearsed.

That distinction matters. Because a human coach's time is scarce. A leader's need for practice isn't.

The more accessible rehearsal becomes, the more valuable human coaching can become for the things that actually require human judgment: context, nuance, organisational dynamics, personal blind spots and consequential decisions. That is a profound change.

Because leadership has never suffered from a shortage of books, frameworks or advice. It has suffered from a shortage of safe opportunities to practise before the stakes are real.

AI is beginning to make those opportunities scalable.

A place to get it wrong before it matters

Perhaps that is the real promise of AI-powered communication coaching.

• Not replacing the human coach.

• Not automating leadership.

• Not producing a perfect score after one conversation.

But giving leaders something they have rarely had in sufficient quantity:

repetition without consequence.

That is what makes the shift from traditional leadership training to AI-powered rehearsal so significant.

The question is no longer simply whether a leader knows how to communicate.

It is whether they have had enough opportunities to practise communicating when someone disagrees, when the question is unexpected, when the stakes are high and when the easy answer is no longer available.

Oraton's role in that future is straightforward.

• Create the space to practise.

• Make the scenarios harder.

• Surface the patterns.

• Give the leader another attempt.

And another. Until the conversation that matters is no longer the first time they have had it.

Because the best time to discover how you communicate under pressure is before someone important is sitting across the table. Download the Oraton app now.

Available on Play Store & iOS