Introduction: Assessing Your AI Readiness

So my main goal today is that all of you take from today how to take all the power of AI or at least know where you are at what point you are maybe you are in a 3 out of 10 or maybe you are in a 6 out of 10 so before beginning I want to

ask you a few questions I need you to raise your hand if it's true are you paying any AI tool? Okay, you are the only one you are not paying, you should do it.

Do you have created, well, what AI's? We are a few people, we can serve it.

GPT? Okay, okay, yeah, the main one, the main one. GPT?

Cloud and GPT, you know it. PrePIC? Perplexit? Perplexit and GPT.

The Rise and Reach of ChatGPT

Okay, the truth is that ChatGPT is the the most used tool right now.

In fact, we are 11 days to be their third birthday of ChatGPT.

It was, you already know, the fastest growing platform in the whole history of humanity. And that's because it has a lot of power.

And what I find usually is that the people don't get all the potential they can.

Productivity, Projects, and Real-World Adoption

So how many of you have created a GPT or a project? okay and how many of you are or do you think are at least double their productivity okay okay that's

Getting Started: From Basics to Custom GPTs

the point I want to get so CGPT has a lot of functions I'm going to focus in here but for example slot in programming and in tech industry is one of the best right now and in big corps CoPilot and Gemini are leading because of compliance so first of all I want to

go into the most basic things and I'm going to start going up our we are

Using Community GPTs vs. Building Your Own

starting speaking okay so first of all here we have the GPT's we have two ways of using GPT's first of all is okay I want to use a GPT built by the community community, that I can imagine that some of you are 100 % doing it, that is the GPT store

and we could come here and say, okay, I want to use this professional writing coach. If I enter here, I can use it and it has its own prompt focused on helping me, coach me as a professional coach okay if you are using that that gpt you are like in at the very beginning

because where you need the ai is to have your own context so for example so you can see a difference you can use a gpt public from the gpt store or you can use your own gpt for writing in this case in emails what I have is my own GPT based on how I like to write emails okay

Instruction Prompts and Personal Context

so if I come to the instructions we can see it's in Spanish and we aren't going to focus in the prompt itself but I want you to see the length of the prompt okay

Attaching Reference Docs and Style Guides

and also it has attached one document about me so basically every single time I tell AI I want to write an email I'm 100 % confident that the email is going to work as I want because I have give the best emails I think I have great in

Live Email Demo and Voice Workflows

my life and a guide of a style so for example I'm going to do one example I want you to write an email to Cecilia saying that tomorrow we are running the event and we need food and drinks by the way what you have just

seen is SuperWhisperer. It's an AI tool for transcribing.

Whisperer is from ChagiPT, from OpenAI. It's open source and you can download it.

If you are not using AI, if you're not chatting by voice with your laptop or... How many of you are chatting with your AI? Okay. We have a really advanced group.

So, for example, when I find an email, I just tell what i want to answer and in this case i don't do it here i only do it with more complex emails i go to outlook and i do it just in outlook with super whisper also it has the same prompt

i'm going to see if it's charged so what it gives me is uh some things i can say as topic and the draft of the email and also it gives me okay before sending check if this is correct

Trust, Review Loops, and Safeguards

And this is one thing that is really a thing many people miss. You don't have to trust 100 % AI until you have trained for safeguard itself.

For example, what we were hearing before is that she doesn't upload directly the video, she edits it, because AI, generative AI, is not in the point of being 100 % autonomous.

So if I came here I have to answer one person and I tell him yes we see us tomorrow this email is in and it's really simple and it's in Spanish so I don't know in what language is going to be but the thing I want you to take from here is that is is in Spanish is taking all the context of the historical of the

email so you could do it manually copying and pasting and it what we recommend at the beginning take the time to copy and paste it because you are going to review it better but later when you have a use case validated you can just use it with the workflows and the

moves.

Beyond Basics: Data Sources, Projects, and Privacy

GPTs that Connect to External Knowledge Bases

ecosystem okay later there is another type of gpt this gpt is quite powerful because each gpt depends on how good or bad is the data it have so for example there are gpt's like

consensus that what it does is connect to a database so you are not just giving it information if you tell him it's not good for health it's going to connect to a scientific database from consensus in this case and it's going to give you the information based on that some ai's are doing

something similar for example perplexity notebook lm of google but the thing is that these ai's are mainly working with Google and a professional researcher is not going to Google and doing their research, they are going to a scientific database.

Okay, so someone knows what is that? It's in Spanish but it's basically saying you are going to go outside, are you okay with that? Do you want to allow it?

Permissions and Security Considerations

So, DGPTs you are connecting to an external database, to an external tool so don't trust all the tools and more if you are with a confidential or delicate things but in this case is i'm not i'm doing a question i'm not saying something dangerous okay

one concern about ai any of you is working in a big companies more than 50 employees maybe no okay in case you found it and for being aligned with the regulation in

Enterprise Data Controls and Compliance

Europe you have to work with that if you have a paid version you have to come to personalization and go all the way down to sorry to data control and here it says improve the model for everyone you have to disable because if not by default in all the chatGPT accounts, OpenAI is storing all the data for later, giving it to

chatGPT .1 .2 or chatGPT .6 for improving it. So the truth is that AI is working like a black box. We don't really know how it's working.

So they supposedly are doing a process for getting it anonymous but no one right now can assure that this is going to be 100 % anonymous.

Projects: Organizing Work, Instructions, and Docs

Okay, we have more things in ChatGPT. This is really similar to the GPT's. It's called projects.

The biggest difference is more like internal with documents and you can store it all your history of chats.

For example, here I have one that I have with many information about subbing, about the company and I I tell him general questions about chapping.

And you can do it all in one part, and you also have, like in GPTs, and if you don't have this you are missing something really important, the instructions of how it should behave and the documents.

It's not only one folder that you have in your desktop. For going a little bit deeper, in ChatGPT we have this part.

Connecting Files and Company Drives

When we hit plus, we can attach files. I think all of you know how to do it.

You can connect it to your company's database, SharePoint, Google Drive, and it will go and search information there.

But what I also

Limitations with Scans and Visual Slides

see that many people miss is that if you have a file that is not properly read by an AI, it's not going to work.

For example, if you have a scanned document, a picture, or something similar, presentation or Google slides and it has a visual concept it's only going to read the text so many people I see like it's not reading properly my document yeah

it's only going to go into text and if for example you have a PowerPoint because it has many visual things that make it impossible because for example

if it's really important to read it what I say okay download the document as image with a limit of 10 and upload it to chatGPT because if not it's going to read only the text.

Deep Research and Model Choice

What Deep Research Is and When to Use It

Deep research. Deep research is basically a way that I'm going to show you some deep research that I have done before.

It's a way of spending many time researching on one topic.

You said you had perplexity. Have you tried a deep research of perplexity? And have you compared it with chatGPTs?

Comparing Tools: ChatGPT vs. Claude vs. Perplexity

okay so the thing with this is that every single AI has their good things or their bad things so CGPT is the best if you want to pay only one tool that is quite good in everything but is not the best in nothing for example here

you can see that well in fact this deep research is a deep research that I did it with a more potential model that I am going to explain this later and it went to 180 place like searches but only 10 websites where it stopped to look.

So I'm going to show you also in Cloudy. Cloudy is another AI that we have here that we are using it and

And what we can see is that this deep research did 562 websites in one hour, one hour and 10 minutes, while Chagipiti, 10 websites in 21 minutes.

Why Switching Models and Tools Matters

1So one thing I want you to take from here also is if you want the best results, you cannot use only Chagipiti. ChagGPT is the best if you want versatility. If you want to use one and only use one because you cannot afford more.

But in the research you can see a big difference.

So how many of you are switching models proactively? Then that is going to be 100 % pay this effort to come here.

Minimizing Mistakes: Data, Prompts, and Model Switching

ChagGPT has different models. We can even come here and see the old models, even 04.

The thing is, as we have said before, AI gets a lot of mistakes. We can even see it here. If you go to your own chat GPT, it says the same. In Copilot it says the same. Chat GPT can do mistakes. They are saying to you.

So how we can minimize the amount of mistakes it has? One is with data.

Okay, I want you to go into this website that is a scientific reference in this country and check the information from there. With the prompt.

If I do a prompt like the one I show you in the GPT of emails, it's more likely to work than one prompt of one line, answer this to to these people and try to write like me. Okay, but how do you write? Give him some examples.

And the third way of avoiding hallucination, that is how it's called in AI, it's by the model switching.

Auto vs. Instant vs. Thinking Modes

So by default, you have auto. Auto, it decides when you are going to use a more thinking model or a low thinking model.

The thing is, the instant mode, that is when it answer quickly,

If you ask him, for example, it's, well, this was happening in CGPT 5, and now in CGPT 5 .1, it didn't happen, but it has other mistakes.

Examples of Reasoning Errors and Hallucinations

It's Malaga, 12 in the evening. We have lunch right now.

It's possible and it's sunny without clouds today. It's possible that in 12 hours it's still sunny. Yeah, because Malaga is a Mediterranean environment, and if it's not cloudy, probably it's going to continue without clouds in 12 hours.

But the reality, if you stop to thinking, it's night, so it's not sunny in this moment. And this is the kind of mistakes that GPT does.

For example, other example with GPT, GPT 5 .1 released really a few time ago, and I didn't try to find new things, but because they are updating it.

Okay, I want to start a company of melted ice cubes. I want you to help me to do the business idea.

Yeah, that's a great idea, let's go ahead, that's an awesome idea, let's build this plan.

While Chagipiti thinking, while thinking for a while, I'm not sure if melted ice is going to work.

We should try to reframe the proposal for making it work.

And when you are asking something technological, something really like more deep to understand, thinking is your model.

In fact, I'm going to do an experiment that we are testing right now with clients for helping you understand how it works.

I need you to be really focused because there is going to be people passing a ball and one of these balls switching you are not going to see it if you are not really really with the attention in this exam in this video so it's one minute count how many times the balls is passed and later you have

to tell me how many do you think it has the ball pass okay how many how many 31 anyone disagrees okay and 31 haven't you seen anything weird d okay okay so i'm going to replay the video really quickly because this is exactly what happened with if you are using an instant mode for something

hard and thank you because it has been amazing what you have seen because this is exactly what happened and this is funny to do it between 1500 people okay

so here the the ball they are switching and suddenly something appears haven't you seen it that's exactly what happened when you ask AI something and you use instant mode without realizing you should be using the thinking.

Okay this is a funny example if you go to your mates today this evening do it if you search count ball pass Daniel Kahneman is in their book.

Okay so this is the

Integrations and Workflows for Teams

Citations, CRMs, and Tool Integrations

example I showed you before it's basically okay it has connect to to this database consensus and it gives us the studies the research in the scientific part so for example if you

work in sales you have an integration with that you can come here and integrate with different tools and one of these you can configure is for example a hotspot that is a CRM

and it's the same logic you are not going to try to pay some documents now you are going to use a workflow a connection for making this more accessible for example here you can see some of what you can use you can connect

Connecting Knowledge Bases like Notion

it to a slack to Soho to notion and the high performance teams with AI what they are doing for example not it's not only this but connecting to notion because in notion you can access to all the meetings you have done all the database and playbooks of good practice in your

Agent Mode: Letting AI Take the Wheel

company okay and for finishing i want to show you a agent mode example have any of you tried agent mode okay agent mode basically what have you heard of manus maybe operator okay there are also tools that allow you to give the ai a virtual machine that is going to use and you can see the mouse moving and doing actions that you ask the ai okay this is not the i have here the in the historical

What Agent Mode Can Do Today

I don't know that, but maybe I have to send this to you. Basically, what I ask is an AA agent to invite people on my LinkedIn. So this is something that is going to work well. This is something that in the moment is not working 100 % properly.

For example, one mistake that it did is that it started typing the message to all the people, but it didn't hit send properly and I told him okay you are not hitting send please hit send and he did it and I don't know if we are going to have one of these examples of mistake in live because it's not working right now but you can try it you can you have to say

the AI I want you to in fact we are going to do an example because the the one I have saved it that doesn't work you have to connect it you can connect it to your outlook or your mail but in this case only internet I want you to enter my LinkedIn account and invite or ask Manuel Vila what we are going to do in the next event okay

the idea for that is doing it in different exact examples what you have should do here, I hope we can do it, see it in life, is taking the control of the laptop, enter your credentials that are a hundred percent, yeah, we, chatGPT is down. This kind of things remember me three years ago when we started to, now it's working,

when we started doing workshops and the only AI tool it was chatGPT, right now imagine it's a words of chatGPT, chatGPT only tool, chatGPT is down. I have lived that, it's kind of funny if we can say it like that.

Okay, so you should enter LinkedIn, put your credentials and later it will start moving through the screen and doing actions. But, okay, I don't know if it's going to work, this example, I'm going to try again, but it will start moving it,

When to Automate with Agents vs. Traditional Automation

but the thing I want you to take from here is, this is possible, yes, it's the best way of doing something that you are going to do every week, every month. No.

If you are going to do it every month, every week, you should probably connect to NA10, to MEC, to Zapier and do it with code.

Because here you are taking a lot of effort of AI for something that it could be done more easily.

But if it's something you are going to do once in a month, once in a semester, once in a quarter, in a year, it's better to use this.

Real Example: Booking Flights and Guardrails

For example, a fun example, these holidays of Christmas, I'm going on a trip and CGPT bought my flights. Bought my flights.

The only thing I had to do is when it went to a checkout, I had to take control and enter my credit card details.

They tell you that it's not going to happen, but just in case I use a temporal card, it's better to be saved.

Okay, so this example is not going to work, but here you can see what has happened.

I have to take the control of the laptop because there is something that doesn't want... let me do it, but basically I can see... yeah.

LinkedIn Outreach Demo and Approval Flow

Okay, what you can see here is that ChatGPT has... I don't know if you are seeing it... has entered

my LinkedIn account has searched to the people, to a person I told him, it has entered the profile and now it's going to send the message and now it's asking me okay before sending are you okay

with this draft message so the thing is if you are going to send it to 30, 50, 100 people you only have to approve one message and one draft and you don't have to do it with everyone yes you can send

Q&A and Closing

it okay so right now let me know any questions you have and if we miss the

time we have some lunch and some food and drinks that we can speak there