Introduction

Okay, my name is Ricardo Galeano and I'm in the last so I will try to keep you awake or at least interested before the wine.

From Early AI Investing to a New Approach

So I have been investing in I have I have invested in several companies through my career and I started investing in AI in 2018 and I was trying to make an AI algorithm an AI service service able to predict images, predict the images, which Google did and I realized after some time and some money that I was not going to be able to make it.

So it came out this tagging of Facebook and then you had this little cat coming out of okay you can recognize there is a cat and the bicycle and all this. So this is from that time I've been working on this.

1and we have done we have created a team of agents and that those agents share the knowledge base so they can talk they don't talk to each other but they share the knowledge base so we started doing that and realizing that companies big companies are not adopting AI as we think as we expect as we all know so it was like okay so what we do so we pivoted and changed a bit on the approach to them the approach to the market and where are we going where we are going and i'm sure you guys are going the same way

Why “Who’s Installing It?” Matters

um so i put the title here we're building the future and my title was um we're building the future and who's freaking installing it but it wasn't so polite and tillman wouldn't allow me me to put this in and it's, you know, it's going to be recorded, so I put this. Who's installing it? So this is the context of the presentation, right?

The Core Question: Where to Scale AI?

So let's see. Okay, so this is the question and this is the question we'll try to answer together at the end, right? Where to scale AI? So we leave it in the parking lot for the moment and then we talk about

Agenda: Gaps, Talent Compression, and Scalability

it okay so first the agenda we're talking about the gaps the talent compression and the real scalability problem i'm going to focus on this this matters the other things we know about it and if not we can talk about you know around the networking about it but uh real quick

Adoption Gaps: Exponential Capability, Linear Uptake

on the adoption gaps the ai capability is exponential and we all know that and this is going very fast and it's going you know every day we have something new and the news are you know it's overwhelming because a lot of stuff is coming out uh so it's in it's improving a lot and uh Jorge showed us the uh LLM arena and this is you know everything stuff coming out every day

China very very strong in models capacities free i'm not free because the information they are eating our information but still very very challenging the status quo of the

US but the adoption is linear the adoption is not exponential so I have been or we have been in working with companies in sweet in Switzerland in the US in the UK trying you know doing the consultancy work doing the legwork let's

What Enterprises Ask For: Multi-Agent Setups

Let's create this agent's organization. The HR agent can talk with the marketing agent or the finance agent to share the knowledge base on the payroll, for example, and stuff like that.

So we built that all private and secure so that all self -hosted and everything inside the box.

Okay. Very cool. Good. Excellent. But we'll think about it.

Why Big Companies Move Slowly

We need to talk with the IT department. We need to see the guys. You know the guys in the IT department were saying like you know what we are the IT guys. We're not so happy

this is happening and you know this IT guy is thinking that We are going to take his job away and all these things that happen in the big companies, and I know a bit about it So no so fast adoption.

Workarounds and Early Wins

Let's do what let's buy 20 ,000 licenses of chat EPT It's 20 bucks, 20 dollars each, or 20 euro, and what about if I buy 20 ,000, give them to me at three dollars or three euro, it's a big advertisement for you, and I use it and I use AI.

So BBVA Bank is using AI all over and is saving a lot of money, and it's great, it's on the at the top and is using ChatGPT for that matter. And different ways, different, you know, I have a lot of stories about that. And so there is that gap.

Different Segments, Different Adoption Paths

Then the big ones, or mid -big ones have their own IT department and they are doing their own thing. Garrigues, which which is a very well -known law firm, Spanish law firm,

which is sold in multiple countries. They are doing their own thing. They are taking their time. I know them well. And, you know, everybody goes into a different piece.

The Missed Opportunity: Small and Medium Businesses

But the small ones, smaller ones, you know, they are not approached. They are like, you know, this is not for me. It's too high.

It's too low. It's too far. I can't, you know, not for me.

And then when I start talking, I always like to say that the deck of cards, the deck was dealt once again. So the cards were dealt.

So now small companies like us, we're here, have the opportunity to compete with the big ones and have the opportunity to tackle into the markets where we can help and where we can create a difference.

And the difference is in the small and medium business, which is the huge amount of business in the world, in Spain, in Europe, elsewhere. So that's what we are, it just seems very obvious.

To me it wasn't, and I spent a lot of time and resources figuring this out because I wanted the big customer, the big fish.

So, about the gaps, about the adoption gaps, the SMBs are in the, it's on the silent minority, majority, sorry, and they still work in Excel, WhatsApp, and in a manual scheduling pencil, WhatsApp, oh, I missed, you know, a lot of mistakes, you know, that's the way it is.

is too expensive, a consultant is going to charge me this and that, I don't know who can do it, oh, a guy, you know, the nephew, my nephew can do this, or my, and this is the way it works.

Talent Compression: The Shrinking Junior Pipeline

So and then there is this other big challenge of the compression of the, of the talent. And this is, this is interesting because we talk about it, I just talked prior to the a meeting with a friend of mine it's a it's a very big law firm international law firm and he talks about and say hey i'm going to talk about the ai and um and we talk a little bit about

the firms that are letting go the juniors mackenzie just uh let go 10 000 i mean it's like 10 20 you know numbers no people and okay we let go these people junior people consultants junior consultants but what's gonna happen when time passes

so who's going to become senior so this guy this friend of mine was telling me I don't know because we need people to learn it's not only AI people need to learn me need to learn to negotiate need to have the experience of what happened in the past in the past negotiations in with the contracts and that is you know

it's a big question so but this is happening the juniors or roles the junior roles are reduced and the first layer has been this cut and this is because what is important to any company that is in the in the market in the stock market is short term so here we're playing every quarter and then you know it's the

the results, long -term strategy, but results matter and the CEO needs to get the bonus as the senior leadership team as well and the proxies of everybody else that is investing is, you know, that's the game. The consulting pipelines are compressed.

A Concrete Example: Scaling ISO Implementations with Agents

we created a last week we have this company we were talking with and they do ISO ISO ISO 9000 ISO 27000 etc ISO implementations not certifications but implementations and they do with consultants that are you know people and they were they are trying to raise funds I said, well, I think that raising funds with people, with consultants, is not that easy. It has to be scalable. Okay, so let's think about how to use the AI.

So we took their basis of the document management, and in a short period of time, built an agent able to—well, three agents, actually—interview one, and the second one was building the context, And the third one was feeding the big agent, enabling it to build 130 documents, you know, structure the whole documentation for the 27001, the ISO, and then in each document generate with AI. And then there was a prompt, and it could do one assessment in a couple of hours. So it's like, man.

And she was like, this is a lady that went out of the company. She was like, no, no, no, but you need to do this with my people. I said, listen, I don't know. I know nothing about the ISO. I don't know about it. I have this prompt. People need to work with it. This is a tool.

And then there is the reluctancy of, oh, you know what? Do I keep my people? And I say, give your people super power to do more, faster, cheaper, we're there. But this is a real story from what the consultants and some of the jobs we know are being changed.

Two Worlds: Cutting-Edge AI vs. Operational Reality

So we have two worlds, the world of the advanced AI, multi -agents, everything, the good stuff stuff we hear about, what Iowa was showing us, and rack agents, and autonomous workflows, MCPs connect all together, stuff. Very nice.

But we have the reality, the operational reality, which is the car repair shops, veterinary clinics, dental clinics.

You know, we started receiving from Millennium just in July. You confirm, cancel, or modify the appointment,

because before July in the last summer they were calling hey do you have a dental appointment tomorrow at 12 Monday so they call them on Sunday and that was a cost so they started changing it but this happens but they are Millennium sanitize you know they have kid on even the they have the good IT department but

there are so many clinics dental clinics that don't have that and those in those services I mean we would you we were talking about about this or you guys were talking about restaurant service a lot of services a lot of businesses of

The Real Scalability Problem

services so now I'm going to the scalability problem Jorge just said that I have five minutes so I'm gonna run so scaling is the scalability problem so So I'm going to summarize and go quick through this.

Scaling Requires Systems Thinking (Not Just Automation)

To scale, it's necessary to build a system.

I call it with the pencil. With the pencil meaning, what is the system?

It's not automate what it is today and make it faster, but think, as consultants do, think think about the process with AI, all together, and then make the jump, make the change.

There is that one.

Document Chaos and Process Bottlenecks

And second is how the documents are organized, which is the other problem.

We have in our company, we have the portion of, everything is written, right? So everything AI uses is text. text. So we do all the tokenization, all the vectorization internal.

There are a lot of solutions today but before they weren't and they were too expensive.

So we do it on our own so we can take a big chunk of documents and tokenize, vectorize, have them into our own rack and with our own models and have it all together into this black box.

But the The documents are not organized. So you have a procedure. Oh, yeah, we have this procedure.

It's from the, I don't know, COVID. Okay, so it's still good. Well, you know what? Things have changed, of course. But they stay there without changes.

So the bottleneck is not the infrastructure. It's not the capabilities. It's not the APIs. It's not the connectivity. It's not the security.

It is the process. losses. So that is the gap for the issue for the scalability. So how to make it work?

Make It Simple: Deployability Beats Raw Intelligence

Make it easy. So the suggestion is make it simple. Make it simple, easy, and stupid. But make it simple.

How was it? I don't remember. Something like that.

Make it easy to use. Give it to the small companies. Something very easy.

Use this QR are to your customers to connect don't need to connect any CRM don't need to do anything and that is it that is what they care about they don't care about

technology they don't care at all they care about how this is gonna cost me which pain is gonna take away from me and once it's with you please stay there they're not they're not looking for it you know the last model of open AI or cloud or whatever they don't they don't care about that so if you have if you

have no I'm not giving any advice but I think that if I give them something very simple and easy and I explained it very simplistic don't need to talk about technology at all this is not about technology it's about what the problem is the what it's what the problem is how we solve their problem uh so clinics alone restaurants

Economics Matter: Tokens, Pricing, and Surprise Costs

logistics a lot of stuff uh you were talking about this guy that is has a service for e -commerce big e -commerce and the amount of tokens the amount of information that is used the amount of tokens that are spent in all the questions so you can have a solution you can have an ai that That helps you to answer, but depends on the tokens you use, the cost.

1So you can be charging 200, 300, 500, 1 ,000 euro a month, but that may cost you 5 ,000, and you may not know it because at the end of the month, you see, oh, man, it's 40 ,000 in OpenAI. Okay, not knowing where it's coming from.

And this is another not -advice -out -advice, and is deployment beats intelligence and this is the pencil way three things easy

A Practical Playbook: Onboarding, Installation, and Support

onboarding automatically second structure installation and third support discipline support customer follow -up this is going to be shared

these are another examples few more examples and to scale five things five five things to think about.

Payment, generic contract, for the customers to adhere to your contract so you don't need to do a contract every time you have a customer.

Automated deployment, onboarding automatically.

Control your KPIs and alerts and costs and the customer follow up.

So that's about 30 seconds.

Conclusion and Discussion

So this is the first question I put on the second slide so if you want to talk about it or you know or any questions you have if I don't know I will say I don't know I won't know I will I will hallucinate