Introduction

What do people expect? Because deep down, it's the subject that is difficult for me.

Because three days ago, I was doing an event here, we were 50.

I also think that to have, this is a parenthesis, but to have animated a conference yesterday in Bern, so there were 250 people, to give you an idea, a whole day full of speakers, of Zalando, all that, I discovered that the audience themselves were afraid to interact in the event. The people between them, even the interviewees, the questions they will receive.

I think it's a very different type of conference from a standard tech conference or a standard startup where people are maybe more comfortable. But I think everyone is a little on their guard on the subject.

Setting the Stage: Today’s Talk and a Simple Case

But anyway, today, as it is marked there, I will be a new speaker. Regina, my good friend, it's me. And I'm going to show you a case that is very simple.

Background: From Marketing Strategy to ChatGPT Curiosity

It's a case that is even personal, it's related to one of my other jobs where I wanted to try to understand. In fact, I am a marketing consultant at the base and then I had, I heard a lot lately, people who were telling me I was building a project and then my marketing strategy, I do it with ChatGPT.

My job is to do marketing strategy. I took some time to understand, I said, but how do people do marketing strategy, which is a tool that, on this subject in any case, is very theoretical.

The Initial Prompt and Motivation

It will show you what you can do, but I said, but ... And there was a case in my job where I said to myself, it was 8.50 p.m., I said to myself, well, I'm going to ask ChatGPT. I said to myself, I want to see what ChatGPT gives me to see if, I almost want to say, in a few lines, the tool can already understand my job, understand the context, understand the stakes, the variables that make them... Because deep down, I thought to myself, if I explain it to a colleague, apart from my colleagues, my bosses, I said to myself that I should give a whole series of contexts, I wanted to see.

So I was curious. And the result was quite striking.

Besides, the result, I had shown it, it seems to me, three or four months ago here.

And in the meantime, I implemented what I showed. So it's a return.

We hadn't filmed at the time. Well, I hadn't filmed myself.

But suddenly, you're going to see the genesis of the subject, to see the result.

Working With AI Without Losing Your Craft

And all of this, indeed, started from a prompt on ChatGPT. While ChatGPT does not intervene, it is very analogous as a case, but it is to show you how, depending on where you are, depending on what you know how to do or not, or even not do at all. In the end, it is at this point that I really begin to realize how we can work with certain NLMs, certain tools, without denaturing one's own work.

So that's kind of the general idea.

About We Make Change

So to understand the subject, I have to explain to you what I do.

Mission and Impact Startups

So my job, I work for this company, We Make Change. We Make Change has as its mission to help what we call an impact startup. So it's a startup that has a goal, a goal, let's say, linked to either one of the UN SDGs,

or who aim to have a positive environmental or societal impact, so it can often be an NGO, it can be a social enterprise, but also start-ups, after we talked about it. We aim to support their development by providing them with volunteers, so people who identify themselves as changemakers, so who want to have a positive impact on society, but using their know-how.

Let's imagine that you are an accountant. Basically, one of these start-ups may need an accountant.

And there is a whole selection of how we choose who is eligible or not.

Basically, you can help, let's say, a startup that is in Tanzania, for example, who needs to be an accountant for their budgets or others, remotely, without ever leaving home, keep your job, you don't need to give up everything, but you want to help causes that hold your heart. Causes like poverty, education, the climate or others.

The Platform: Changemakers and Organizations

So we have a platform where we have 50,000 changemakers and on the other side, about 1,300 start-ups or NGOs or others who are looking for volunteers but with precise skills in all areas. This is the basis of what we do. We started this activity 5 years ago, 5-6 years ago, and all those who are in this platform know us for that.

They come, it's a bit like if it was a great LinkedIn Jobs but voluntary, they read it, they connect to the platform, they find roles that correspond to what they want to do, they say the time they want to give per week, They give it to you, the company does their maintenance, they work together, but if afterwards they want to continue their professional relationship or get involved, it's their problem.

Our goal is, let's say, to demultiple the positive impacts of companies that have a goal, let's say, how to say, I will almost say noble. So that, when I arrived in this company, that's the core business.

First of all, it's an interesting core business because the volunteer does free volunteering, The company doesn't pay for the service, the start-up doesn't pay for the service, and we don't pay for the service. You might say, how did they get me engaged? I don't want to say it, but I'm more expensive than my parents.

Corporate Workshops: The Exchange/Change Day Format

And so we have a service that was created in the meantime, two years ago, which is that we discovered that in fact the multinationals, had the same need, well, they also had a need for volunteers internally. Companies with more than 1,000 employees, especially for social responsibility reasons, need to offer their employees opportunities to volunteer.

So, they don't necessarily... People can, for example, take unpaid leave, people can have projects that they can carry out, we can leave for 6 months, 1 year, there is no problem. in the idea that they want to deliver, so everyone has their own strategy, but they want to deliver inventory at the same time, in the absolute, without necessarily having to move, whether it's in the example of startups in Latin America, in Australia, in Asia, even the cost and the carbon footprint of doing that, it's complicated to put in place.

So we arrived with a version that was in the form of a workshop, with our own startups, but there targeting companies, where basically, we call, it's a three-hour workshop, where everything is remote, the company provides volunteers with specific skills. So, for example, the marketing department wants to volunteer next Friday, they come and so they sign up by saying, well, I have a digital marketing, branding, marketing skills.

Startups sign up knowing which is the company that is the patron of the workshop and what they have available and say, for 3 hours I will need support from marketing experts for blah blah blah. So basically they formulate a brief.

So rather than the long version which is the volunteer who comes to work and who has the job at your place, there is the version, I have free consultants. who come to share their knowledge, to help me advance a strategic problem within my company. So that's the corporate version. The company sponsors the event, so they pay us for that, and we coordinate the event.

Global Coordination Challenges

It's where it's all a mechanic too, it's that you have to imagine that if we take a company like Coca-Cola or Standard Chartered Bank or even we have consulting companies, our employees are all over the world, our startups are all over the world and our facilitators are all over the world. I am based in Geneva, so already everyone is in a different time zone, everyone must be at the same time on a Zoom link, we also have other platforms that we use, so there is a whole coordination that comes with it.

Until then, is that clear?

Selecting the Right Startups

Now, in my role, I am as much the one who accepts startups in the platform as the one who accepts startups to participate in a given workshop. So my role is as much to source startups that correspond to the profiles of those we want and can help, but also to start having, because we started with NGOs, socially responsible social enterprises, that's good,

Maturity Mismatch and Workshop Readiness

But the problem we had was that, depending on the size of the corporate client, their level of maturity on certain subjects was not adapted to do a three-hour workshop. There are people, in fact, if you give three hours with specialists, for example, from Coca-Cola, who are, for example, from the marketing department, which is perhaps one of the best marketing departments in the world, Well, according to the organization, there are some who are just at the level of trying to understand marketing. That is to say that the three hours they will spend saying to themselves, I don't know what I'm doing.

While a slightly more advanced company or with, let's say, a perhaps more mature management or a broader understanding of a subject in RH or other, you can get to something that is much closer to the advice or really the reflection, the discussion and also the method that we give them is much more more constructive over three hours, let's say even two and a half hours, than someone whose basic problem was not to know the subject.

Defining Minimum Maturity

So basically, we need a certain degree of maturity.

Why Turn to ChatGPT?

So, so far, I still don't talk about IA. It's deliberate. But then why did I go to see ChatGPT?

So ChatGPT, I don't know if it's still on, but there you go.

Framing the Problem

So, first of all, I started, I asked a question. It was late, it was late at night. I told him, I explained exactly what I was explaining, but now imagine, I spent almost 10 minutes explaining to you what we were doing.

There, I said to myself, first of all, do you see the platform? He said to me, I can't access it, but if you show me what I see, I can help you. Eventually, I can see the answers, finally see what you need.

And then I said to myself, I'm going to have to make a real prompt. Now we're more on the prompt of two or three sentences.

The Core Challenge: Registrations and Friction

I was tired and especially I was tired because the problem that I was trying to identify, finally, is the problem, the biggest problem I have with the workshops I'm talking about, the Change Days, is that people have always been coming to our platform to find volunteers. In fact, I had a problem with registration. And I thought to myself, how can I save time to increase the number of people who sign up? to not have to run after them to tell them, have you seen this email to sign up?

They don't know the product and until a few months later, to register for a Change Day, you had to receive an email. So you'll tell me, well, that's it, you get an email, people sign up, no, no, they have to go on the platform, they have to fill out a form, it's very analogous as a process. And according to the degree of maturity of the management, I discovered that there are some who are not very comfortable with their computer, there are some who have not yet understood what the product is, there are some who are not reliable.

And in fact, there are things that I can predict, now with time, according to the activity, according to the country where someone comes from, according to the size of the organization. The flow is very, very, very, very, very simple.

Education Gap on a New Product

But in a context where the product came, this product came three years after the existence of what we do. That is to say, in fact, I realized that on 1300 startups, maybe, I almost want to say more than 60%, The startups which were the closest were the ones who participated, but we never thought that out of the hundreds of others who are there, in fact, we never educated them to understand what it was and how it worked.

They don't need to know the workshop to participate, to benefit from the platform and look for volunteers. They are there for the volunteers. So there is also a lack of education. That is to say that we made this product to survive economically.

Audience Question: Value Proposition

Can I ask you a question? What is the value proposition on your product? It does not seem to me ultra clear and what are the benefits?

Because in the end, the start-up does not have a lot of clients, it will look for volunteers, why would it be the value proposition or not? I make a big parenthesis, but we have time, so it does not matter.

If you see a Dietschy motorcycle, you have to buy it because it will get stuck there. Profession of value.

Clarifying the Value for NGOs and Startups

Let's take an NGO. Let's stay on NGOs for a minute to define what a startup is. They need volunteers, they lack resources. That's why they come on the platform.

The version that we propose is that at the strategic level, the management sometimes needs coaching, mentoring on a given subject for free. They don't want to pay it, they can't pay it. That's what we offer.

We offer it with, unlike our volunteers on our platform, who can be you, anyone, Mr, Mrs, everyone. There, it's, for example, professionals, so the top management, sometimes, who participate, the leadership, who participate in these workshops, who give their time.

This time, whether we value it or not, but the start-ups, the management of the start-ups, they have access. to be able to, let's admit, it's like when we have an advisory board in a start-up. It's the same principle, it's just that it's accessible to you whenever you want, as soon as we organize it, according to the skills that we have available. You have questions about it that you have always asked or a problem, you can address it for three hours.

The proportion of value is 100%. So it's on a limited time versus when you take a... Someone who is there for longer time and who limits it, on the other hand, it becomes his role in the organization. So it's the version, let's say... It's a bit like the advice version,

with volunteers who have very little chance to be on the platform, lack of time, lack of whatever, compared to others who have more time to give and who will be there longer. So it's a bit like it was the simplified version.

Operational Bottlenecks: Emails, Process, and Reliability

Now, even up to that point, once the startups have understood what it is, the problem is that even when they have understood, my problem is that they don't read their emails, When they read it, they do not go to the end of the process, so they do not start the process. In short, I find myself in a problem when I have several at the same time.

I read what I wrote there. So here I explain what it is.

We have a premium service called Exchange, which lasts three hours.

To invite, they receive an email. Some do not see it. They do not read it. Others do not know what it is.

And suddenly, I can not match them with the companies or the volunteers as soon as the companies submit to us.

Lack of answers. I try to find the best strategy. to find the right startup linked to a change center. It's like if I had to do a real matching.

Ideal State: Smart Matching and Criteria

Ideally, I would have liked to be able to say, I put the criteria, let's say, of a given workshop, which company, which type of domain, and they tell me, contact this, this, this, this of your startup. In the absolute. Which would involve having as much information as possible on them, but it's to tell them,

For example, for us, it is important that the founder is on time, just because if I tell him, in a month, at 10 a.m., you can be free for 3 hours, because they have a job next door, they can't. These are our internal criteria, but I'll explain that to him later.

Current Outreach Tactics

There I explain my options, I say my options are to invite them and wait, to go find them on LinkedIn to understand why they did not start the process, by email, by LinkedIn, and finally write them a WhatsApp message when I have less time or when I am caught up in time to make sure that they sign up so that we have enough startups. Because we commit to the company to provide a number of data of startups that are presented for their participants.

And above all, knowing that I do not necessarily have, so for a skill where I do not need to know. So, for the startup, its hidden benefit, because for me, it's very good to have a volunteer, but is it behind saying to myself, I'm going to make a moment to be able to be bought by a new company, or I'm going to be recognized, or I'm going to be able to be at QBI, or things like that? It depends.

For the moment, at this level, at this stage, for the types of companies that we take, in fact, it's not because you're a startup that you're going to be in a platform. We reject it every day, even if it's because the purpose of what they do doesn't interest us. The answer is yes. The benefit in this case is not the problem.

Why the Workshop Isn’t the Issue

To advance on the subject, if you want, I'll say it like that. It's not the benefit of the workshop that is the problem. Basically, how they use our platform is much simpler to find someone in the long term than to do the workshop.

For example, you couldn't know which workshop was available when you log in, because they didn't tell you. In fact, it was only by email. That was four months ago.

No, but it's true that I... No, but I understood, but don't worry, I know my case, that's not the problem. The problem is that I'm going to show you the consequences of that.

Because it's logical, yes, the value prop, yes, yes, yes, yes, but no, no, on the case yes and no, as we describe it, yes, yes, that, but at the end of the end, once they participate and they arrive, when they receive the help they receive and they put us or estimate it, they are quite aware that these people will never work for them for free. They would not give them advice, nor time, nor that, that's not at all a problem.

It's really what I'm describing here, it's that I have to run after startups to sign up for something where once they are in it, they say to themselves, in fact, it's great, but you don't verify your emails. Or the people who are our contact in the platform leave the company, we don't know. Or, you see, it's a whole series of technical problems, but the problem to explain, if I don't explain all that, as I did there, it's going to be very complicated.

From Problem to Predictive Scoring

So, in fact, I want to say, how do I optimize, how could I immediately know who I have to talk to? So I went back and said to myself, I don't know if he'll know or not, but I'll see what he says.

ChatGPT’s Strategic Suggestions

So he said, Thank you for explaining it to me so clearly. It's an interesting position.

You have to manage the pressure of time, the reactivity of the start-up and also the skills that are available to companies. The challenge is to identify quickly and to activate the start-ups for whom it would really be a benefit in view of the skills available.

Here are some strategic and predictive ideas that I propose to you improve the conversion and save time. I say to myself, OK, I say that, I do predictive, I say to myself, OK, why not?

Because in addition, I'm going to explain, I'm going to give zero context apart from factual contexts of what blocks me. I'm not going to explain the criteria of how to choose.

And that's where suddenly, so after I suspect that also by doing certain research related to my job, I think he's been learning about me for a while. But in fact, over time, he tells me, OK, so

Key Criteria Identified

To understand, if we want to go towards the prediction, first there is the notion of urgency. There is the notion of mapping, of matching the right roles with the right skills. The size of the team and the level of development of the startup, for example, if they are very early stage and the founder is alone, he can have less resources and have the time to give three hours, which I explained before.

So I didn't tell him that, he understood it. These are things that are true, but that I did not explain to him. I just explained to him the methods that I have of contact, the limitations that I have to make them participate.

Action Model: Prioritizing Outreach by Score

And that's where he got into this idea of ​​saying, in fact you could, since you have certain actions that you have at your disposal, you could create a notion of calculation that tells you, according to a certain number of criteria that are aggregated, the skills available, the box in question, maybe the moment of the day or other, instead of doing the same actions for everyone, so the LinkedIn call or in any case the WhatsApp message, the personalized email or X or Y, you can have a way to calculate, so you would enter data on all the startups that you have to calculate and say on a challenge of data, it's useless to go any further than to receive a call to action email.

If they do it, they don't do it, you will have the same result. Which is a bit logical, because in the end you know a bit, because you build relationships with the participants, you build relationships with the startups, you know who could possibly be more reactive than others, in addition to the criteria that you set.

So from there, That's where he enters into a whole notion. And what's crazy is that I didn't expect all this.

and to have already participated in a Change Day, maybe they have already done it, maybe if they had a good experience, they will want to come back.

At one point, he starts to tell me, okay, is it also how the mail is formulated? In the meantime, indeed, we reviewed the copy of the mail, according to how it is brought, according to how it is written, according to what is put in the title. Will it have an impact on the understanding, on whether, X, can you anticipate before?

In short, all this to say that at one point, I told him, well, I asked him if he wanted more information on how to define a scoring. I said yes, of course.

At this stage, I don't do anything anymore. It's him who lives his best life.

My computer crashed. And so, it was something like that.

Designing the Scoring System

So in the idea, there was a notion of points. So two points according to the criteria, let's say if someone is active on our platform frequently.

One of the criteria, for example, is that sometimes we approach, if someone has a role in digital marketing and we have a challenge on digital marketing, there is a good chance that they want to participate. If these skills match, if the size of the team, in fact, it will give a certain number of points and it is actually reflected as below, right there.

So, by integrating this data on a box, they say, well, typically this box, in this given thing, with what is available, it has a score of 8, so is there a direct LinkedIn message? Or is there a WhatsApp message?

So I said to myself, well, at this stage, I had not yet, necessarily it was night, I had not yet spoken to my boss, but I said to myself, it's true that on the idea, if we could know as quickly as possible what was possible to have with, companies, if we could gather this information, we could go.

So here I spare you a little bit this part there because I will show you just the result, but the idea He gave me a lot of angles.

Operationalizing in Excel

He said to me, for example, would you like me to make you an Excel file? An Excel file that puts in place what I showed you on the screen.

I said to myself, well, it's simple enough. I'm not stupid either, it should work. But what he asked me, without me needing to specify anything, that's when I said to myself, it all depends on how we clarify the facts or how we prompt the problem we have.

So it led to some internal changes.

I don't know if I opened it, but I think I have it here. I'll see if I have it here. I think I have it here.

What is that?

No, it's not him. It led to that.

Implementation and Data Collection

So, in fact, we had done a project that was a kind of listing of 100 impact startups to monitor in 2025. And then we had more than 300 registered and then of those who were kept, we took them for specific reasons. So there was a reason for like really giving a list, it was more mature, they raised funds, they said it was Y. And suddenly I had integrated this notion.

Which is therefore from this, let's say, from this, from this research, from this strategy. It was that, so where it took me time and I couldn't hack it, it's that basically we had, it's a list of the top 100.

Top 100 Impact Startups Project

So I had to talk to all the winners. So I started mimicking it. For the moment, I talked to 75 of them.

Onboarding Calls and Data Capture

It was a call that allowed me to have their information. So basic information on what we need to know, their needs, the skills, for example, that they are looking for as a priority.

Knowing that for them, these were the types of startups that we wanted to put forward with possible bodies. That's also why we did this campaign.

Subjective Scoring Dimensions

And also, there was this scoring.

This scoring is subjective, it came from my conversation with them. It was to say, let's say, is the person open to feedback, open to to present themselves, to give information, to see if they were prepared for the call, because we would give them a little... What's it called? A little program, saying, well, you present for two minutes, you explain where you are, we explain what we do, and we explain our services.

Findings: Scores Predict Workshop Behavior

And in the end, it ended up on a ranking. So, very subjectively, there are people with whom it went very well, but what was funny to observe,

It's that people who have the least bad grades. So that, we're going to say it's a bad grade in the idea, particularly these two columns which are the personality for the challenge. 1Basically, a challenge is a coaching that is offered where you have professionals who talk to you about your business, who have just discovered it. So the personality is knowing how to be vulnerable, knowing how to accept feedback or that kind of thing.

Typically, some could have demonstrated in this call to onboarding already their limits, their incapacity to just already accept that it was not them who were in control of the conversation, that everything was not going well, that in fact they needed help. Well, what has been proven is that some of these participants

we managed to prove that when they participated in a change day, they were at exactly the same risk as what we calculated. And again, nothing was automated.

We could go further, because this is version 1, but in the score, we realized that we could correlate those who made a good score in the onboarding call of 20-30 minutes and those who, either are not reliable when we try to invite them to a change day, or are not, let's say, docile in terms of what we ask them to prepare, and above all, are not very good, let's say, between people, with our corporate partners during the workshop in itself.

Limitations and Nuance

All that... Totally. But the problem is that it works.

Where it works, first of all, there is a part of subjectivity in everything, but what was interesting is that someone who applies, because in fact the object that we try to avoid is that some of these startups, for example, I don't remember which ones, there are some who have raised 25, 30 million, there are some who have 65 employees, we could say they are advanced companies, but in fact it is not because they are that in the model of what we do they know, for example, to prove vulnerability, to know how to question oneself, to know how to receive feedback, which is very subjective, but to know how to do it or not after a while,

But the state of mind that makes it work, that the conversation works and that they can draw benefits, well, the ranking had an impact. So when there is a zero, it's that sometimes I forgot to note, that's also a human error. But it's to tell you how, at the base, so when I presented this for the first time, it was in...

It's... It's... It's... It's... It's... It's... It's... It's... It's... It's... It's... It's... It's... It's... It's... It's...

Rolling It Out Company-Wide

I was in the middle of it, it was in May, I had shown only the return of GPT and then in the meantime we had started the campaign to talk to the start-ups and others and suddenly we implemented it, so I presented it to my bosses, we implemented it and suddenly we were able to test then to see how we wanted to extend it to the whole company by even making an onboarding code that would be available to new companies, if we only have for roles, to be able to evaluate them and know if they are ready to participate also in these workshops with the corporates.

From Late-Night Idea to Process Change

So, all this on a problem that I released at 23h.

A Practical, Not Theoretical, AI Use Case

So, this example is an example that we call, so in the three types of TOC that I mentioned, I mentioned the theoretical, this is a practical example.

What This Is—and Isn’t

Me, in my job, it was useful to me, it's like I also made an example of automation, there was something that I wanted to test, I tested it for myself, did it impact the company or not? No. Did it make a tool, something that we built? No.

The majority of people we invite are people who build something, who come to present and do a demo.

Why Personal Use Cases Matter

But deep down, what we realized is that a lot of people have cases that are quite personal, who think they are not interesting, but according to what you do, in fact, it's just that sometimes we never thought that the tool that makes the link with the rest of what you want to do for your hobbies or for your work is missing.

And so that's exactly the kind of talk that makes us not always need a technical profile or an entrepreneur. Deep down, we rather need to But no matter how specific it can be, we need cases.

Q&A and Open Discussion

That's what I wanted to explain to you. If you have any questions, I'm a taker.

Otherwise, I'm chasing the pizzas that should arrive.

Session Logistics

So, normally, I'm supposed to have three stakeholders.

It's 7 p.m., which means that, theoretically, we end up 19h30 and then it happens.

Call for Live Demos and Tools

But suddenly, the talks that I prefer are very practical. I find something that requires slides or something else, it's good, but sometimes it's even too technical.

What I was wondering is, at this stage, there is no preparation, I admit it, there are people who need preparation, but would one of you have either a tool, that they have done or not, either an agent or a case, something that they have done on a private or professional level, that they would feel, I don't have to introduce myself, or talk about it.

Yes, I'm thinking about the fact that... Is it related to AI? Always. An IA tool or something that is built... For example, GPTs? For example.

So, for example, but it's this thing, it's my computer, I don't know. But it would be something that you can show in two or something that is a case or something, if you have a demo that you wanted to do, what you... That's where I take the fact, I'm talking about now, I'm not talking about next week or next month. It's if you had something in particular that you could show now, if you want.

Is it an experience that we lived with AI that we can share? For example. It's not exactly that. In any case, yes, but ideally something that would be to be shown.

An experience, something you've done, tested or... With AI? It's the subject, it's the spice of the... You don't have to, it's more of an idea if there is a tool that you use every day and you say it can be nice for others to discover or something you do every day with AI. That's the question. If there is not, there is not.

If there is not, I go to the pizza, it's not very good. I'm asking if it's a part of the project itself? It can bring me the tool. It can be anything. For example, it's like normally we show, look, I'm going to show you something.

Audience Discussion: Email Writing and Automation

Me, apart from using it for everything, in terms of information, projects, I have other ideas for the environment, for example. Knowing that I am in finance, it is very difficult to ask people, to ask them if they need to pay, if they need to pay. I think it's a great experience.

But if not, are there any tools that we can use to write emails? In addition to ChatGPT, for example? Listen, I think they all have... First, if I take the students in general, they all have a bit of their... It depends if you want to write precisely, like, a technical email. What's the particular editing challenge that you're trying to... What I like is that I use ChatGPT to talk to him, like I want to talk to a client.

Yeah. And after him, he will predict it to me in a direct way, without the insults. It's very good. And on the other hand, I think it's very good, it's a good experience because it also teaches me to have things differently and it teaches me more and more to talk to customers, even face to face.

If one day we can't do it, it's related to the GPT, to the other one, of course. Automate that with each one we received. We have an automatic response. But rest of the machine, would we have the control?

You can take steps. You can take steps to check. You're not going directly. I did all the automation directly. It's too dangerous. Okay. It's good to be able to utilize it.

It's good to discuss it too, but in fact, it's funny, you want to tend towards automation, it's the writing on it. Is it the writing or the automation?

Concerns: Sensitive Data and Compliance

What scares me is that we work with very delicate and confidential data and information. For example, there are numbers that we can't give to the customer, for example, and if we follow the logic, we have to give it to him. Today we can have the human side, the machine side, and give them limits, because I believe in it.

Can we completely automate, but we can't really help the thing? But is the automation okay? That it's not going to change anything?

Can you answer me? It's a discussion, I don't care. Automation?

If there is hypersensitive data, even on the content,

For me, it's a total automation, so you start with your trigger, you do your gpt chat, with your programs that you have automated, the optimization on all your platforms. And instead of, for example, a content, you will have an article, I will have it separated afterwards, I will ask him to make me a link, a trigger, a thing, everything is programmed for, okay?

And instead, it's too dangerous to send it directly on LinkedIn, to update directly on LinkedIn, you can have a SaaS, that is to say your SharePoint, your Google Drive, etc., where you will have the data and you will still be able to look at the eye what will come out. I prefer to do that because sometimes I do automations and I'm like,

If you don't have your AI in your company, in your own server, the data comes out. For example, for the OpenAI, the server hours are in the United States.

Do you think you can share the data with someone in the United States? Currently, I don't think it's allowed in the United States. It's much more complicated because if you do, like here, a request with OpenAI, you send the data

For the third party? When you're on Office 365 too. It's not true.

It's true, we stigmatize ChatGPT, but when you put your data on Google Mail, on Gmail, on Microsoft, it goes down too. You have to have servers in Europe.

With all the interfaces like this, you can buy the installation on your premise. For example, you can be in the local, for example. You can have OpenAI on your server, no problem, but it's not like that.

That's what we call the rags today, that protect you. But we don't really know. In fact, they exist.

Governance Before Automation

When I take the issue of sensitive data that we give to a client, the subject exists even before the IA. I'll give you an example.

When I was working at RTS for a long time, there were data that I couldn't give. For example, I could never give the amount of TV rights for a sports competition. So when I was asked the question, it was as if I said it was a red flag and from there I knew I had to send it to Zurich.

Basically, in what a client answers, when you want to automate something, especially when you arrive, basically you have to, whether it's a prompt or whether it's your agent or whatever you want, you have to ask, understand your internal process of what is the basis, you could not even say if you wanted.

It's like I said, you have to educate your… It was this idea that you will save time because it will answer everything.

Practical Automation Steps

For example, in a platform, for example, when someone signed up, I had an email template that I put every time, so I did the check-in manually. Then I had a check-in, an email template, always the same, where I put the name of the person and I sent it. Until the moment we said to ourselves, well, we always send the same email, so we can add an automation when we click.

confirmed, well there they receive it. But because the mail does not move. No matter what they say, we know what we answer, it's even logical.

Well, as long as you know that you are in particular cases, either you must be able to see the answer before, or it's you who have to send it. Here I'm talking to you in 100% just procedural mode, it's not even a question of IA, it's a question of if you know that there is a degree where they will ask you a point where it must come back to you and you must be able to intervene or not say, the case does not necessarily apply. We are not yet at this stage where you can close your eyes and say, I'm running my business.

Because what is sensitive and what has become sensitive, you are the one who knows it. You should constantly be able to say, in fact, in passing, this, if we tell you this thing, this acronym, this thing. It's like having an assistant, actually.

An assistant, I have to tell him, no, no, answer only this type of email. There is a part of automation, I think, which is really like if I told you you had to apply for ISO standards. They have a whole way of explaining how you have to work. It's almost, if you have to tell yourself how we work for each thing we do, and then what is it that, in all this, deep down, is it worth it to not intervene manually?

Human-in-the-Loop as a Principle

You know, it's almost a reflection of project management, you see.

As long as there are complex cases, I draft an answer, for example, this is something I sent yesterday, you draft an answer, and then from there, I changed what I wanted to change, or the bulk of what I wanted to say, the fluff of what has been said, I just want to save time on it, but I never said to myself, answer in my place, and I sent it as it is, because already I would call myself, your full name, and then, well, I mean, for example.

But that's a simple example, because deep down, for the moment, I have a very simple use of ChatGPT, because deep down, what's crazy is that the time I spend writing this, well, that's a copy-paste, it's that it annoys me, the answer was so long, I say, answer me to that. But what I mean is, the time I spend answering that, there are people, no, I have something else to do.

Closing Prompt: How Do You Use AI?

In fact, what interests us is to know, if you come back to the meetup one day, how you use IA or how you dream of using IA.

Invitation to Share and Return

In fact, that's my question. Because if you're there to a certain point, it's something you want to do or you're doing. And that's what I don't know if I don't talk to you.