Introduction: The Future of Work with AI Agents

So today I wanted to give you a little bit of a view of what we consider at MindStone to be the future of work. Basically, we really think that at this point it's extremely clear that in the very, very near term, every person in a company will have an agent in one way or another.

That agent will have to go and collaborate with every other agent in the company.

and the ability to both manage your own agent and enhance it, basically the extension of yourself.

From Personal Agents to a Chief of Staff

We talk about it as a chief of staff, which I know is not a very well -known concept for most people, but the idea of a really great chief of staff is they're your representation

wherever you can't be, and they end up extending whatever you want getting done in situations you can't personally be president.

isn't it?

Introducing MindStone Rebel

And so we built Rebel. We call it MindStone Rebel.

There's a whole reason behind it, one of them being that we're a European -based company and we're now somehow putting ourselves right in the middle of what every single AI company wants to go and build.

So how many of you have heard of Claude Cowork? Yeah. And well, at least all of you now will have heard about MindStone Rebel, too.

So there you go. So that's 100 % coverage versus only about 40%, so we're winning, right?

But we're very aware of the fact that we're a small company, but I think we have something that is actually dramatically more powerful than that today, and I want to try and run you through it.

Why Rebel Is Different: Local, Private, Yours

Now there's a few things that really make this special, which is, one, it runs on your machine. It doesn't run in the cloud.

It runs on your own files. files, so you can actually see if I go into library here, you have a little you can actually see the files that lay underneath, and then if I go here, you can actually see you can literally see the folder structure that I've got here.

So work, this is all our MindStone stuff, and then these files are literally what you would see here as well.

So everything runs locally. locally.

Skills: Codifying Processes for Consistent Execution

The other thing is that, and I'll go through the general interface afterwards, you have this concept called skills. Now skills is actually something that Anthropic was talking about as well.

We used to call them playbooks, we just decided to use the same terminology so it's easy for everyone else to go and talk about it, but basically a skill is a set of steps or process process that you decide to formulise in a way that AI can execute it all the time.

So I've been using this for about a year and a half at this point, and so I think at this point I'm at 150, 200 or so different skills that have been documented in one way or another, and because they are now documented, it means that any AI can go and execute on them.

And that's really powerful because it means that at any point, for example, I can go and say, use your proposal writing skill, and it will we'll go and build the best possible proposal based on a transcript that I've had with a client.

And it knows exactly what good looks like because I've taken the time once to document what does it mean to have a great proposal?

1One of the keys, by the way, in this, when we build skills is actually having examples of what good looks like.

So not only the process itself, but also like, here are three examples of great proposals that I've written in the past, and then suddenly we'll go and do that.

So that's another thing.

From Skills to Automation: Scheduling What Matters

that's the once you have skills you then also have another thing which is that you can start to tell the ai to automatically execute these skills on a regular basis so you can see here

uh i i'm a serial founder i'm computer science back of computer science background and one of One of the big things that I'm not great at is focusing on anything but the things that need to change and the things that are not working.

So I literally have a daily team shout -out identifier skill whose only job it is to go and crawl all of my Slack and Gmail and everything else and figure out,

wait, who's the person and company that I should talk to today and say, hey, you did a really great job.

I know it's, I guess, somewhat sad that, that I need a skill to remind me of the thing, I guess I'm at least aware of it, but mostly what this is about is I document at once what is the flow, and then it executes

every day, in this case, every day at 5 p .m., it will go and execute, and then it goes into my inbox, which tells me what are the things I should go and focus on.

So it's literally like a chief of staff going around and figuring out, hey, Josh, really, Really, you should pay attention to this thing that happened today. And shout out a particular person.

Connecting Rebel to Your Tools and Data

Now all of this works because you can, of course, connect Rebel to every single one of your systems.

So you can see here, I've got it linked to Gamma, to all of my email, to, well, in this case it's got Outlook. We're not actually a Microsoft shop, but I had to test Outlook, so I do have a Microsoft Microsoft email that allowed me to go and do that.

We've got browser automation, I've got a personal CRM, basically just all of my contacts that are nicely catalogued.

I've got Fireflies for our notes, I've got a HubSpot integration, so it means that everything I can do, I can just instantly, I can tell it, based on all the emails I had today, go and we were literally having this conversation earlier today, go and update HubSpot based on the activity that happened in my inbox today. day.

Why? Because I'm terrible at updating HubSpot and making sure that all deals are present in there. So now I can go and do that.

All of that to say, all of these again are executed locally. So this is not in the cloud.

Supercharging Knowledge: 20 Years of Notes, Indexed

And then the last, there's just a cool thing, which is if I go into library, you can see here, it's going to take a few seconds.

Every single one, so I had fun. I have about 20 years' worth of notes. notes, they were manual notes, and about two weeks ago I started cataloguing them, basically

I sent every one of my manually written notes over the last 20 years to Claude, and I just got it to do optical character recognition on all my manual notes, so now I have 20 years of digital notes that are all on my machine, and they're all now semantically indexed here,

and you can literally see all the links between the different notes that I had in the past, And you can go and zoom into a particular topic you can have no idea. Okay.

This is Jeremy Utley I'm probably not going to go and ask a question about this because I have no idea what's gonna come out if I do but

You can actually Start a chat based on this hub of notes now So literally what will happen at that point is it starts a conversation and it'll have

have all of the linked notes of 20 years of my life, and then I start a conversation with that as being the hub. All of that to say, it's really pretty cool.

You can even just go and you can get, as you can imagine, this is going to give me a quick summary of what that particular note was about.

If I click on zoom out, what it's doing is giving me a summary of the cluster, which basically means a summary of all of the files that are related in one thing, so it gives gives me a summary of that particular topic, and it actually tells me how they are related, so there we go.

Now I need to actually click out of this.

Live Demo: From Chat to Action

Now I want to go into, well, I guess this was already live demo, but let's go into live demo which is actual chat.

This interface everyone will know, why? Because everyone has the chat interface or whatever, the conversation on the left -hand side and then the thing here, the conversation.

Now, I'm going to start with go through everything you know about MindStone, and I've got an audience in front of me here for the AI community in London, and tell me the best way to go and pitch MindStone.

There you go. I'm going to do that.

The best way to pitch MindStone to this audience to give them an idea of what we do. do.

Crawl through all your members and everything you know about Mind Zone and then put the perfect pitch together.

Designed for Deep Work, Not Instant Replies

So the first thing to know is that this will be slower than ChatGPT or Cloud or anything else. But that is by design.

It's very interesting. As we started rolling this out to different customers, you can actually see what's happening here behind the scenes.

It's literally reading my file system because my file system is semantically indexed. Basically, it's indexed based on meaning, so Rebel is able to go and figure out what are all the files that are related to this particular query that I can go and read in.

If you're the type to sit behind your AI waiting for it to finish, interestingly enough, it's actually the wrong way of thinking about AI.

Just like you wouldn't get an employee that you would onboard to your company and sit behind them whilst they do work, the whole idea is that you give AI tasks that are big enough so that you don't have to worry for about about it for a bit then you

Give Chunky Tasks, Not Micro Prompts

can go and do other work and come back when the work is done so what you really want to do is you want to give it tasks that are chunky and that require a whole bunch of things that otherwise you would have done and that is really what rebel is built for I have tasks that sometimes take 45 minutes it'll go off and like

correlate they actually this morning so we're about to go and raise around around actually that wasn't this morning so this weekend I got it to crawl through six years of my emails and to prepare a briefing on all the investors I talked to over the last six years

tell me where the conversation was left at in order for me to be able to tell them now it's like hey three years ago we decided not to go and do it together as something together oh but now we have this thing and like contextualize that in a way that would be useful that took 45 minutes I think something like that.

But that's the type of task that you can give rebel. So let's figure out if it actually does something good. Where do we go?

The opening hook.

Okay. We've trained the C -suites, let's see here. We've trained the C -suites of five of the top 100 companies in the world on AI transformation. Here's what we've learned.

10 % of AI adoption is about the tools 90 % is about changing how people think about work Absolutely, it's actually one of the main lines that I use which is that I can probably give this tool to everyone in the world And very few will be able to understand how to actually use it Why because some of what I just said most people will go and sit behind it and wait for it to go and do work And they're going to be limited by what they even asked the AI to begin with

Generating Presentations with External Tools

so So let's go and go one step further. Now go and take all this content and build me a gamma presentation that really drives home the point.

So what I did there is I just double tapped the speaker icon here, which basically just sends it automatically versus populating the input area. And now let's see what it's going to go and do, because Because they've got a Gamma MCP that is directly related to Rebel. So let's see if it is able to go and execute on it.

Come on, live demo. Any questions in the meantime? Yes. 100%.

One of the other things that everyone is going to have to go through is getting comfortable paying a shit ton of money to different AI services.

I was doing the I was talking to one of my founders earlier today, and my co -founders. is we're currently spending about $1 ,000 per head in the company just on Rebel users per month, per month.

So let that sink in. That's without our development usage, which is another $1 ,000 per user. That's the reality.

But the other thing, Rebel didn't exist, there was not a single line of code written before the start of December.

We're 300 ,000 lines of code in and we're mid January. This normally would have taken three and a half years for a team of ten. That's also the reality of what's happening.

So let's see actually what's coming up here. I can see that it was it had found gamma, it made the first one didn't work, and then there's a mismatch between the schema and what the API accepts.

Let me try with PaceX. Somehow it's figuring out there was something going wrong with the gamma MCP. hopefully it will go and do a better job now. Yes.

Under the Hood: MCP, Models, and Local Options

Yeah, you can, absolutely. Anyone familiar with cloud code? So basically this uses the cloud agent SDK under the hood,

and all the bits underneath, it just follows the standard MCP protocol. There is an MCP file that Rebel uses.

Literally if you copy -paste the configuration of an MCP server into Rebel, it will self -configure,

here, we use the super MCP in the background in order to not exhaust context, it will relaunch that and it will configure the tool for you.

Sorry?

So the model at the moment, good question, so the model, indeed, we use Claude's models, so Opus, Sonnet and Haiku, depending on what goes on behind.

This actually works on local models too. So the other very interesting thing is we can go into companies that never were able able to use AI because they can't be sending it to some external cloud, this same thing works with a local model. And that is because it executes on your machine, there's a lot more that you can do with it.

Watching Agents Orchestrate Work

So it looks like it's actually building the presentation. It's just waiting for Gamma to have actually finished that. I think probably in a few seconds after this. So here, basically, it just determined I'm going to wait for a little bit until Gamma is ready.

There we go. Hopefully that's it. Okay. So now you can see here.

It's built a gamma presentation Let's see if it's any good Yes

AI for the many not the few most companies do AI backwards by chance to be the licenses run it to our workshop expects transformation to happen doesn't happen

One of the biggest things that people get stuck with is the moment like if you just roll out the tools everyone's like I'm using AI all the time it's like replace my Google usage so you know what Google wasn't bad Google did Google pretty well and if all you're using generative AI for is a slightly better Google no no surprise that everyone's talking about AI being hype if the only thing you see is a slightly better Google right this is

part of the problem is people don't realize training and and actually your Your imagination is literally the limit that gets you to understand how much you can actually do with AI.

Because we've all got years and years and years of doing things certain ways that now you have to do differently before you can get to the real value of AI.

Anyways, one of the, and here it looks at the 90 % mindset, but our credentials, top tier experience, yes, we have worked with the biggest companies in the world, like Delta Delta Airlines, Home Depot, Hyatt, Pearson, Fortnum Mason, kind of all across.

Three pillars of transformation, Enterprise AI Academy, executive sessions, rebel, that's new. First time I'm doing this. So there you go.

Okay. It's actually not bad. The pitch is landing.

I didn't write it. You saw I didn't write it. Okay.

Now let's figure out, okay, now it's the new thing.

Proactive Meeting Companion: Spark

So, I'm going to start a meeting. I'm going to wait for a second, or a few seconds. Here we go.

So, you might think, like, great, note taker. We all have those, right? Nothing particular. But this is a special note taker.

and so hey spark can you tell me what we were talking about in this meeting hey spark go ahead on the transcript so far you were introducing a special note taker to the meeting that's really all that's been discussed so far there wasn't much else i mean spark is

right oh shit now i use this i use the trigger word it's gonna go again um okay wait i'm gonna going to have to spark go again now the key is a rebel knows that it's in the meeting and I've

We've all had this. Okay? You don't need to see my face there.

I've got this toggle in Rebel which allows me to give it access to my local Rebel. So what I can go and do now, hey, Spark, can you tell me how I should pitch Mindstones to an audience of the London AI community?

Real-Time Reasoning on Your Local Knowledge

Now it's going to take a little bit longer. Why? Because in the background, it's running a full process, it's going through all my files, It's actually figuring out wait. How do I answer this question?

Now it waits, so it's circling and it will change its animation. It will actually change its state The moment it has something to contribute to the meeting so right now. It's it's indicating that it's thinking It's basically going through stuff, and you'll see it change state the moment that it is finished.

It's thinking hey spark go ahead use it four hours a day to run MindStone, and that 10x productivity became your product. Lead with the explosion.

Training business grew 7 to 8x to 400k monthly while staying profitable, and Revit hit 100k ARR in two weeks before it was even finished.

Close with a mission, bringing AI transformation to everyone, not just tech elites.

So, I mean, it disclosed a little bit more than I wanted to.

From Passive Note-Taker to Proactive Teammate

but you can see where this is all headed right I mean the next step and I'm talking like literally next week it just starts to go through the transcript and proactively start saying hey by the way I've got I've got something to say based on what is currently in the meeting We're taking these AI agents and going from passive to being genuine, proactive co -workers.

And everything here that is happening is shared across the organisation.

If I actually show you here in the chat, let's see, if I go here, okay, in this case it didn't save anything. I think I can go here.

Shared Memory Across the Organization

here so this was this was one of the big chance I had recently which was about updating the content of a program as I'm having a conversation with rebel about how should we update the content of a program it stores these as memories that are accessible to everybody else in the organization and so anybody who then has a question about like what is the content of a program will have that up

to date when I have a sales call it takes all the important information what What the rest of the company should know gets put into company memory.

And then when Neil needs to go and put a proposal together, can put a proposal together based on the examples that I've had and based on the skill that we built together on how proposals get built at MindStone.

And that's just on the sales front.

Automating Customer Insights and Testimonials

Custom success, we have the same thing. We were recently talking about, like, there's a new skill now,

which is about we get feedback from customers by email, by Slack, intercom, a whole bunch of different things.

Now there's an automated skill at the end of every day that just figures out. Okay. What are the customer testimonials that we have? What are the quotes that we can actually take?

That normally you probably all went through this stuff in different ways where it's like, okay I need to have a central repository where I can figure out the quotes about the the quotes we can use from a customer's Perspective who have had a good experience now.

You can just query it. It literally just plugs into everything and every day Decides to store this in a nice and central place

Building at AI Speed

The and the last bit so as I said a month and a half ago Literally nothing of this existed. I mean the thought process was there

I had my personal operating system, but like this whole thing 300 ,000 lines of code built in the last month and a half Entirely vibe coded for what it's worth

Q&A and Wrap-Up

Any questions

I'm probably going to keep it out of the meeting here. If I kick it out of the meeting, it does the...

Oh, the other thing is just a fun thing.

That's a wrap. Notes will be ready shortly.