Introduction: A Non-Technical Path Into AI-Driven Workflows

To try to bring ourselves into the future by showing how a non -technical person like myself can do something similar to what Tino has built on a lower scale with the tool that we have built as a company, MindStone Rebel.

We We are using this tool ourselves. That's how we work these days. And this is how we have tested the tool.

So we have been the guinea pigs.

We have written 6 million lines of code in, I mean, I've not done that personally, but the team has, over four months.

And we have created this system, which basically allows every non -technical person to work like a developer would work.

so what I would like to do is to show you what's possible with a tool like this that we think it's going to be the standard one or two years for some organizations and three to five for others for how you can work with AI and in professional services and I will start so what I'm planning to do is to

Demo Goal: Generate Multiple Deliverables From One Prompt

to create a mapping, an HTML research report, so a research report, and a presentation just on the back of one prompt.

And while that will be running, I will walk you through some of the features of the system so that we don't waste any time.

Choosing the Research Question (Travel & Aviation)

So what I would like us to start with is let's pick a research question and let's have a very quick brainstorming exercise a research question of any kind on the travel and aviation industry a burning question that you have when are we gonna run out of gasoline okay so

where airports built okay now what do we want to look at what are some of the factors that could influence where airports are being built and why population density i need to write well because i'm going to take a picture of this and the poor other is going to struggle otherwise traffic density what else i heard something else so it's like demand destinations geopolitics or you want safety okay okay sorry okay let's stop

Live-Demo Reality: Embracing Imperfection and Iteration

up here and normally i mean we can go on forever but this is not meant to do something i want to show you the practical application i don't want to someone is calling me not now okay take a picture of this so i will airdrop it on my phone by the way this is a live demo i can

guarantee you that something will go wrong even if it's a very short time and i mean it's if you have not done it before it's a terrifying experience because you're basically standing in front of her many people and basically playing with fire in a way because these tools are still not entirely predictable but I think

that's part of the experience and that's where we want to bring everyone I mean just making a fool of yourself from time to time for the sake of learning something it's actually a good idea because it encourages that try and error which is really important with these AI models so I need to convert

this to JPEG I've done this mistake before and so I copy and paste it I could act I could tell rebel to look into my downloads I'm not going to do that because it takes a bit longer I want to save time so I will do it the the old -fashioned way, copy -paste it.

Prompting Rebel With the Audience Brainstorm

Hey, Rebel. I'm doing a live demo for the Lufthansa Innovation Hub in Berlin. You should have some context on this, because we had a conversation earlier today.

And on the back of this brainstorming that I've done with the audience, I wanted to put together a web HTML -like summary of the brainstorming exercise, a research report, and a gamma presentation.

And the question that we are looking at is, where is it most likely that airports will be built in the future? you should have a skill for this we have worked on this skill today so you

Why Voice Input Improves Speed and Context

notice that I've been speaking to the system and that's how I work these days I would rarely type anything I would only type if I need precision and the reason is twofold when we speak we share more information with the models that that the models can use to actually tailor their output.

And then I'm a lot faster at speaking than I am at typing. So my productivity improves quite a bit.

I can probably do 10 emails. If I just speak to the system and the system cleans them up and sends them, I can probably save already just by that simple fact

of not writing things down, not typing things down, i can probably already save half of the time okay this is cooking so as this is cooking

you see we have a little plan we have the summary research report gamma presentation great the instructions seem to have worked as this is cooking i'm going to show you the other functionalities of the system.

Core Product Capabilities: Connectors (MCPs) Across Your Tool Stack

And so, Tino mentioned earlier MCPs. So, Rebel has MCPs, or we call them connectors in non -technical terms, that basically allow it to interact with a suite of applications. applications, email, Slack, Google Docs, so on and so forth.

So basically, you can speak to the system. You can speak to Rebel. And Rebel would then produce, take information from other applications, and upload it to other applications.

So I can say, write an email. I will find the email in the inbox. And that is a massive unlock.

lock because you bring it basically augments it's your the ability of the user to take information on board and translate it into into other formats so

Example: Summarizing Event Logistics From Email

let's demonstrate that hey rebel I'm doing a live demo so please be very careful about the sort of information that you share but go through my email and just give me a summary of the logistics for the meetup at the Lufthansa innovation hub that I'm organizing with Anna and Benjamin first and foremost and

so this is going through my email and it's producing a summary for me so we let that go in the meantime I show you something else and you see how I'm

working I'm all I've always had three or four things active and I review them in parallel because that's how you get to the productivity benefits it's not by focusing

Knowledge Management: Local Memories, Skills, and No Vendor Lock-In

on one thing at a time yes thank you that would have been my next point so on top of skills we have memories so sorry on top of the connectors we have memories so these memories are local text files that you can ask rebel to record and all of this is and so you whenever you have an interesting conversation you can sort of sign post it and say save it as a memory because I

want this information later and share it with my team because also my team needs to know about this and it runs entirely locally it can plug into any LLM and he

will also be able six to twelve months from now to plug into local LLMs for the technical folks so it will be able to run locally with no vendor lock -in because you you own the text files you will choose which model to use and that's where ultimately the rebellion comes from

Models and Context Window: What Depends on the LLM

that's where the names the name comes from yes so if it's running locally what's the context window size for answers and everything that would depend on the model itself typically so i mean now So now, for this demo, I'm using Opus and Sonnet, the anthropic models, which are some

of the most powerful and precise on the market. So that would be 1 million, 1 or 2 million context window, depending on the model.

OK, so you see the logistic summary. I can tell you 96 registered participants, registration in the lobby, the program. you are getting pizzas from Big Nona so we are not saving on pizzas great this is still going

Example: Recalling a Saved Demo Plan Memory

so let me show you the memories because I think it could be interesting we were preparing earlier today a demo the demo that I'm delivering now at the Lufthansa Innovation Hub can you share the demo plan with the audience and so this is gonna give us the memory that is related to today's

demo you will see the plan that I've prepared when I was pretty much doing the same thing on the open floor like chatting with a laptop to just get the demo in order as that is going I want to show you I want to show you these luck

Example: Posting an Agenda to Slack via a Connector

integration so the father you can send messages directly hello rebel I am in the in a live demo at Lufthansa innovation hub I want to share with my teammates the file that I'm doing a live demo and I'm thinking about them can you send a message to the general channel summarizing the agenda directly from the

events page and so you see I've said go on the events page because I know that I know that I want the information from there so the system is going on the milestone event page where some of you have registered and is taking that

information and then it's going to share that information with with the teammates

Outputs: Research Summary, Report, and Sources

okay ah we have got the research results before we do that okay it's trying to do it we are going to go back to that later that doesn't look right this memory is

still going but we've got our research results research report the text is quite small let's go through it together quickly and remember this was on the back of a simple prompt and three minutes

airport locations interplay of five core factors destination demand safety requirements national budget capacity and our traffic density forecasts these are our factors we have set the agenda for the LLM the LLM is not hallucinated it.

Where are they being built?

1 ,031 active projects, total value, a few openings, so you can get safety, a few renovation projects, and there are the sources. And this was just on the back of one prompt.

Now, presentation.

Presentation Output: Gamma via a Connector

How many of you have heard about Gamma? Half of the audience.

So I'm using Gamma through a connector, through an MCP. I could also use it as an app but this is what we build together which is the same information just in a different format you know you might not like some of these elements you can effectively chat to gamma and change pretty much anything

that you don't like no the way I would do that is through an HTML dashboard that I would create through rebel itself because gamma is more of a visual simple visualization tool if you want to do more complex visualizations then it's more that Latino game that we have seen but we can make simple changes I don't really like this you know if you do pick anything but I mean this is not sure I

Interpreting Results and When Human Judgment Matters

like this so how do I do this again well I think it's it shows that tension between all right the geopolitical angle and the destination and geopolitical and the travel demand I think what's also also suggesting is it's an analysis based on the past ultimately and that's where our judgment becomes useful maybe we should not be looking at past data to assess this and we should do more of a foresight exercise but it's a very good

point so we have seen these two outputs now let's send this luck message to the team great the agenda is there he did not find the right link but we still have the information you see that's the very interesting thing that although i did not find the event page on mindstone for reasons that i can look up later it still found the agenda through the email that is the the powerful thing about a system like that that if it if it finds an obstacle it thinks about ways to overcome it that's great post it on the general channel

so we did our research this is now the slack message is being posted the demo screen I mean it's struggling a little bit to load because it's running hot but you see here the plan photo of the audience brainstorms research report gamma presentation six seven minutes then I did a few other things as we were waiting okay let's see if this posted let me see so you can see how this works

Governance and Access Control: Working in Separate Spaces

there you go questions very good question we set up separate spaces and And Rebel is only allowed to operate on a certain space for a certain conversation.

So we would have an executive space that where our CO, CO, and CTO work. And so I would not have access to that space. And they would be the only ones who

are allowed to access it and manipulate information from that. OK. So we could create our own spaces for that.

Differentiation: Cross-Space Work, DIY MCPs, and Ownership of Data

MARCO PAGLIA - Yeah. And besides the window looking, what is something rebel can do that so cloud co -work is a very vertical system so cloud co -work would do what rebel can do in a folder in a vertical you say this is your space

work here rebel can do the same thing and more across multiple spaces and you have no vendor lock -in you own the files you own the memories you own the skills so you can choose the model later down the line and I think that's right now a very

distinctive feature on the market because vendors are really are generally trying to capture people into their own system and leave very little leeway to to integration with other providers or other solutions, generally speaking.

It is through MCPs. You can set up your own MCPs, and it's quite easy. I've done it myself. Remember, I'm not technical.

I used to be a diplomat. I'm as non -technical as it gets. I was, and I still am, a little bit of a grandpa when it comes to this whole technology business.

but yeah you go in your settings and then you can set up your own connectors quite easily I was there because I it's a question that I often get asked so I'm spending quite a lot on this so the

Cost and Deployment Reality: Local Files, API Model Spend

last seven days I spent six hundred dollars on tokens which seems really high but when I think about the sort of and tokens by the way are it's basically the cost of using an LLM for people that know they don't know so I basically spent $600 on on this system and using this system in the past seven seven days but when I think about

the amount of work that I manage 20 meetups across the world more than 10 partners across three different geographies I don't think there would would be any way in which I would be able to do this job

without spending, in the way it is currently set up, without spending a similar amount.

But you said it's running locally. So it's running the file, all the files, the memories, the skills are saved locally. Then the LLM needs to be plugged in.

So I'm using currently Anthropic, and that's going through APIs. OK. That's where you pay the money for it. Exactly. Exactly.

But you are able to insert your own model. If you want to work with a local model, with an open source model, you can do that.

Conclusion: Q&A, Feedback, and Next Steps

Guys, we have the pizzas. So I'm really happy to take any questions that you have, but let's do it over the pizzas.

One last thing. This has been an absolute pleasure. I've really enjoyed the questions and the and the brainstorming.

You've been quite active, so thank you. We love feedback, so I will put a QR code on the screen. And this QR code also gives you access,

gives you the possibility of winning a free license to the program that Lufthansa Innovation Hub was talking about.

It's a bit of a laborious process because you might need to check in if you're not a Mindstorm member, but you should do it

and then we will announce the winner later during the pizzas thank you so much it has been fantastic

and looking forward to the networking the pizza and the drinks