Introduction

What am I gonna do today?

I will use AI for coaching for coaching myself and

this will be I think I really care about this demo for

Why AI Coaching Matters

For two reasons first it shows

Augmentation over Automation

The way in which we think about AI at MindStone, which is as a way of empowering people and not replacing them augmentation and automation and

Learning Through Live Demos

And then it allows me to show some techniques that I think are useful.

Please come in. Please.

Two places.

That we think are useful and that we would cover in the program.

And I'm really putting myself out there in two ways. First of all, it's a live demo. Things can go wrong.

And at some point, something will go wrong. And that's perfectly fine, because these tools are not predictable and they are not entirely reliable.

The point is not, I wouldn't want to convince you today that they are a hundred percent reliable, the point is even when they are 70 -80 percent reliable they can still be very useful.

The second way in which I'm putting myself out there, I'm coaching myself so I will be exposing some of the things that I'm not doing well in my job and see with you how I can make them better.

Reverse Prompting: Flipping the Script

So to start things off, how many of you have heard about reverse prompting?

Show us. Yes, not many.

What Is Reverse Prompting?

So reverse prompting is when you flip the script.

So rather than asking questions to AI, you ask AI to ask you questions. Does that sound familiar?

Any more people?

Yeah, some heads there, some heads there, but okay, probably I would say three in ten. I've seen it before.

Using Reverse Prompting for Self-Coaching

So I'm gonna use this technique to coach myself, and then after we do that, I will show you how that technique could be used in a system that is a lot more powerful, that is the system that we use internally, than what you see in the public domain.

Live Coaching Walkthrough

Imagine you are an experienced executive coach.

I'm not doing as well as my CEO in sales meetings and I want to understand how I can do better.

Can you help me through through this problem asking me one question at a time.

Voice Input and Super Whisper

So I'm using an app. So I'm using voice to speak the model. I'm using an app that is called Super Whisper.

How many of you have heard of Super Whisper or dictation apps? Not many.

And this is useful for two reasons at least.

Why Dictation Helps

First, you speak a lot faster than you type, three times as fast, which is why throughout history, history, people have been dictating things.

Like Stendhal, the writer, would have dictated this entire novel in the 19th century, just as an example.

And then it changes the patterns of interactions with the model.

Because when you speak, you reveal a lot more detail, a lot more quickly, and that detail is useful for the model to be useful for you.

So, can you all see the screen? Is it too small? Can you see in the back? No?

I will zoom in. Okay.

Model Choice and Trade-offs

You might notice I'm using the thinking model, which means that it will be a bit slower, but it will be more accurate in the sort of responses that I get.

If it takes too long, then I will switch to the standard model. Okay.

Diagnosing the Sales Meeting Gap

Okay, thinking of your most recent sales meeting that fell flat, walk me through it from opener to close. What exact moment did the moment dip?

I think people get really excited about the product that we have, but I then struggle, compared to my boss, in getting some follow -ups. and getting some actions out of them.

And so you see, it's really following the script and asking me one question at a time.

When you reach the final five minutes, what exact call to action do you make? Word for word, and what concrete commitment, person plus date plus deliverable, do you ask them to confirm?

So it has gone straight to the point, and it has pointed out that it has a very specific question about something that I'm probably not doing or not doing right.

I would send a follow -up proposal with some information about the program, how they can register, and what steps they should be taking to discuss it internally with colleagues.

Zooming In on the Call to Action

welcome let's see any questions that this is progressing is this making sense let's see

1on the call itself you explicitly ask them to open their calendar and book a concrete concrete next meeting before you hang up no i do not should i do this you see that the model is saying thing okay lock the next step on the call keeps momentum reduce it reduces drop off and signals leadership.

Try one of these. Very interesting.

Does this look like something that could be useful for you in your job?

Or can you think about any other ways of using this technique?

Why This Technique Works

This flow, you can use it for pretty much any problem that you're facing. sync, this exact flow.

And it dramatically reduces the scope for hallucination.

And basically, AI, you're catching sort of two birds with one stone.

You're giving to the model a lot of information that can be useful for it to be useful for you, but you're also structuring your thinking at the same time.

Speed, Structure, and Reduced Hallucinations

And when you use voice, you're making it a lot faster as well.

A Playful Interlude: Turning Coaching into Song

Now, before I show you, let's say, the more sophisticated version of this that I would do on our company system, I want to break things up a little bit.

From Prompt to Lyrics

So imagine you are an experienced songwriter. can you put together the lyrics for a 1960s pop song that basically reiterate how I should be running my sales meetings, something about the importance of committing to a next step or booking a call.

I've not done this before so let's see how it goes and in the meantime how many of you have played with suno i saw one one hand in the back it's just to to make it a little bit more light -hearted

so that we just don't discuss my my failures publicly we start with sparkles in their eyes the demo is feeling right but promise fade after the call they vanish in the night it's not bad

so darling just don't send the note and hope they will circle back let's turn that glow into a plan get action on the track okay could be worse now all we do

Generating Music with Suno

is should take this sooner is a tool that I consider brilliant it's very user -friendly I use it sometimes to put music around my poems because I write write poems my spare time you put the lyrics we said 1960s pop let's see what

it comes up with so you see here these I actually created them out of a poem I had written now we stick to our book it before goodbye song so you don't like 60s okay let's see

Scaling Up with Company Systems

it's fairly good now back to us so this is the way we work with days these days inside MindStone.

So this can be interesting for the more the more technical people.

Cursor as an Operating System

How many of you have used or heard of Cursor? So the way we work with basically and Cursor is just a tool for developers but mainly but we're

using it for the entire company and so it just has an agent window and what we have done is to create some playbooks and skills and put a lot of company knowledge around cursor so that it becomes our operating system.

Playbooks, Data, and Drafting

And so it can access all my data, the data from my colleagues, their playbooks, the way they like things to be done.

It can prepare draft emails, it's linked to my entire working universe.

Applying It to Sales Coaching

So how can I use this for coaching let's see I'm not happy about the performance in my sales meetings you should have access to my anonymized transcripts so go through

those those transcripts do not mention by any means any names of people or companies because I am in front of people and doing a demo live but I want you to tell me how I'm doing on my sales meeting compared to what Josh, our CEO, would have done.

So this is not a very detailed prompt. It's not even a very good prompt by any means.

I'm just explaining a situation and giving some general guidance to the model. You see it's going through the transcript.

It's going through the sales playbook from my boss. So this is the way he likes things to be done.

It's going through the anonymized transcripts. It has understood that it needs to go through the anonymous version.

It is getting some information on Josh and how he likes things to be done. And it's giving me feedback.

So it's not too different from what we had before, but there is just a lot more background on everything that we had. And and we are getting some more information out of it.

Insights and Scoring

It would spend more time understanding the specific pain points of the client.

It's better at handling objections, at creating urgency. And then it's giving me, on the back of all of those, all of that information, some things that I can be doing differently.

And it's giving me a seven out of 10. It's not bad, could be worse.

Takeaways and Q&A

so I hope this gives you a sense of like the dramatic improvement once you start having it's what we had before with the reverse prompting was already useful but

once you're actually plugging all your knowledge into a company system you're creating a system around that knowledge then you can get a lot more feedback on on how you're working a lot more quickly. And ultimately, that should make you better.

Key Benefits Recap

I've run a bit over time. So let's take two or three questions. And then if you have more, we can do them over the pizza and drinks.

Audience Questions

Yes? Yeah.

So my question is, in the past, now, GABRIEL SANCHEZ - So my CEO has designed that for himself. Yes.

We still need to roll it out in the entire company. But the speed of execution is there. Because you have seen, it didn't take that much to do.

And you see this small thing here? It is very important.

Behind the Scenes: Token Throughput

Because what this tells you is that Cursor has just been through 75 ,000 tokens.

Think about them as words. So 70 ,000 words would be 30, 40 pages. pages?

So just to give me this answer, it has reviewed over, say, 30 seconds, 70 ,000 words worth of information.

Running It Live vs. In-Depth

So when you do it live, you would keep it a bit simpler. You feed less information. It would be, just remind me to do this at this particular point.

You would follow the transcript, and then it would tell you, do that, or do the other thing I would do it through the transcript it would record the

transcript live when you hit a certain mode it would keep that somebody up to date when you hit a certain moment then it would give you a certain feedback you would know that you need to do something but I would only link you to that transcript no I would base it on the actual transcript or the call that the

Why This Dictation App?

model is recording in real time no questions yes so you use that yes completely for two reasons uh the quality is higher the quality is higher than most and then there is also a habit thing look for the microphone of the specific app so quality is higher and it's faster

i can type this in whatever field and it's the same shortcut every time i don't need to go and option spacebar and it would just start typing in whatever field which could be

from an execution perspective it's also easier I think I need to do option could be sooner if I say heavy metal we just type there so these two reasons