What If You Could Practise The Difficult Conversation First?

There are some conversations you would probably like to have more than once. An investor pitch, a difficult sales meeting or a presentation to the board might all be easier if you could try them, get challenged, rethink your answers and then start again before having to do it for real.

Harvard Business School is using AI to give entrepreneurs exactly that opportunity. Its Foundry programme includes simulated investor pitches, sales calls and board meetings where participants can practise against AI versions of Harvard faculty and experts, repeating the experience as many times as they need before facing the real conversation.

It sounds slightly strange, particularly when the person looking back at you on screen is an AI-generated version of someone who actually exists. But underneath the novelty of talking to a virtual professor is a much more practical use of the technology: creating somewhere to get things wrong when getting them wrong doesn’t really matter.

Making Practice Easier To Repeat

Good rehearsal needs someone who can react to what you’re saying, ask difficult questions and make you think about your answers. Finding someone willing to do that repeatedly is harder.

You might persuade a colleague to listen to your pitch a couple of times. Asking an experienced investor to sit through 20 versions while you work out what lands is rather less realistic.

That’s where an AI simulation becomes useful. Foundry allows participants to test different approaches and repeat their pitches before the real meeting. The AI agents are designed to challenge their ideas rather than simply agree with them.

Participants can therefore try an answer, see how it holds up and change their approach. If it doesn’t work, they can have another go without needing someone else to sit through the whole thing again.

Getting The Bad Version Out Of The Way

There’s something particularly useful about being able to fail privately.

In a real investor meeting, getting flustered by an unexpected question or discovering halfway through an answer that your explanation doesn’t make much sense has consequences. In a simulation, it’s simply another attempt.

That could make AI particularly useful for practising the parts of work that are difficult to learn from instructions alone. You can read advice about handling objections, presenting an idea or responding to a challenging question, but actually doing those things requires a different kind of preparation.

The simulation doesn’t need to reproduce a real person perfectly to be useful. What matters is whether it creates enough pressure, unpredictability and challenge to expose the bits you haven’t properly thought through yet.

Before The Real Conversation

An AI investor isn’t going to react exactly like the person sitting across from you next week, and a simulated board meeting can’t recreate everything that happens in a real one. The personalities are different, the relationships matter and there’s considerably more at stake when you’re actually in the room.

But the simulation doesn’t have to perfectly recreate the real conversation to be useful. Harvard’s approach gives people somewhere to practise how they might respond, discover the questions they struggle with and try a different approach before those mistakes actually matter.

That’s where this use of AI gets interesting. The difficult conversation still happens between people, but some of the preparation for it no longer has to. You can get the awkward answer, muddled explanation or terrible first pitch out of the way before you walk into the room.