Introduction

Thank you everyone for coming and I just wanted to do a little disclaimer before we start and it's that before coming today I just didn't know I had to present a slideshow. I thought it was just a talk and it was literally five minutes before that I created this using AI.

I did have the outline and everything that I want to say with all the bullet points in a document thankfully for organization but it kind of like demonstrates one of like the biggest like wonderful things of a generative AI and how can be the pretty beautiful things that we could do it's kind of charming and it's what most people implement in their day -to -day when we talk about AI and so before we start I

Understanding the Audience and Context

wanted to understand who you are so then I can add that a little bit also the talk to this because the reasons why it failed and the recommendations will depend also on the type of business or profiles that we have in the audience.

So I would love to hear a little bit more about you, like maybe two or three people, so like which industry do you come from? Is it you're a business owner or you work for a corporate? So if you want to participate, that would be great. If not, I'll just pick random people.

I just tried to go for the extroverts first you know business owners what kind of business okay interesting cool and can I ask how how big is your team it's just you oh this is wonderful okay yeah oh good oh yes it's for that ah it's a

the microphone yes I'm a business owner aha 3d printing business okay and I am the only employee so I definitely need help this is fantastic is is one of the biggest opportunities for AI purchase for small business owners so that is

yeah we're gonna go into that fantastic yes okay hello business owner as well I I have a little product agency, and we're using AI for leverage. Okay.

How big is the team? Maybe we are seven people. Okay. Cool.

Cool. Perfect. We have another person there.

I'm a medical doctor, and I use AI for my private clinic and to help with the patients and send active reminders. I love that.

and it's wonderful because it demonstrates that AI is not just for like traditional startups or businesses it is for every professional and it's changing our day -to -day lives thank you maybe one

more yeah yeah we are an AI based company and we are applying to real estate okay yeah okay cool interesting okay so more or less having an idea and

Many of the use cases that we see in the reports of like AI implementation failing or like how success would look like, it comes from enterprise. And so this is why I wanted to understand this, because when we talk about startups or small businesses or even just professionals, the way that we learn and implement AI can change a little bit. So what do we prioritize kind of like, yeah, it depends on who we are and what are our objectives.

About the Speaker

So, before going into the next presentation, lovely to meet you. I'm Stephanie Oliveros. I am the CEO and co -founder of SheAI.

We are an edtech platform, a partner of United Nations, where we have like a global community to close the gender gap in AI literacy specifically.

I'm a psychology researcher. My research area is the impact that AI has on human behavior specifically. specifically.

So I think it's like, there is like some intertwined things that we could do.

The AI Bubble: Hype vs. Return on Investment

So first, we have a huge problem here, the AI bubble. 1So we know that AI has potential, or we've been told that it has.

And we see the fancy stuff like yes, I can create a presentation in five minutes, that is quite handy if you are a solopreneur or a small business owner. but what happened when you truly want to measure that return of investment like are you really cutting time are you really getting the money back that you

are investing and the numbers is like 252 billion was invested in AI in 2024 and seventy to ninety five percent failed in implementing AI effectively I'm talking about bigger corporations and enterprise, no? And the thing is that it is kind of like, well, it obviously depends a little bit on

the projects, but the reason why this is happening, and I see that also in small business owners because they get caught into the chat GPT loop, like, oh, if I get this prompt and I'm going to buy this, that is going to solve all my problems, it's not going to solve all your problems, okay?

Why AI Implementations Fail: Five Common Patterns

1) Tool-First Thinking (and the Paradox of Choice)

Okay, so what happened here is that I've identified five patterns that we need to tackle. And the first one is tool -first thinking. So this means that we say like, oh, I want to add AI to my product or I want to add AI to my day -to -day life. Instead of thinking, what processes should I improve and then how AI would look like into that as well.

So it's not about testing random AI tools. actually there is this thing in psychology that's called the paradox of choice you have many like so many like thousands of AI tools and they evolve

and they change all the time if the amount of time waste for the particular small businesses is insane like in in terms of money but in terms of like testing and buying credits so for for things and to experiment and time that you cannot spend okay so that is like one of the the biggest thing and then we

have things like uh big budgets normally goes towards generative ai is specifically marketing and sales because that's the fancy stuff where it should be to go to operations where is the

2) Data and Workflows: Turning Unstructured Information into Systems

action the actual value that you're gonna see okay so then the second thing is data so 20 percent of the data that you see in your company structure the invoices the the people the contracts coming in, things that are on structures like emails, meetings, like okay,

I would assume that most of you would have already a meeting assistant AI. What do you do with that? You transcribe it, and then what happened with that transcription?

So if you are just using the tool for the sake of using a tool, and then you just say, yes, I'm just going to go manually and and check this document, and it's just, that is not a process that is effective.

But if we create a workflow that takes your meeting and then extracts the key points and then assign already in your calendar, based on priorities, what you need to do,

and send the email to the person that you discussed during the meeting that you should be, that is really cutting the time.

And kind of like all of that builds into a database of conversations with clients so that you can pull out of that database automatically and you don't have to go again and check the meeting

that you don't even remember where is the transcription so it's not about ai tools it's about how we implement it in the systems and also like how we organize this data so if we just start like particularly with with with things that are unstructured like the the potential the ai has to to kind of like grab all of this and help us build specific systems and structures to help us is huge

so we should pay attention to that when we talk about enterprise there are all the issues uh that that like you have to clean a lot of data and they spend years doing that. They have big data engineers specifically for that.

Kind of like the strategy would be like, instead of like cleaning all, you just select specifically like experiment. Like, okay, we just want to create a system

that is super efficient at invoicing. So then we're gonna focus only on cleaning and creating a data set that's specifically only for that.

I know that we wanna implement AI everywhere, but Volkswagen just lost 7 .3 billion because they want to implement AI everywhere and it didn't went well so we just select one area experiment document the process and double down on that and

3) Build vs. Buy: Where Custom AI Makes Sense

that will go ten times faster okay then when do we build and when do we buy and that is kind of a big constraint particularly for startups for big businesses of course it's just faster buying like I don't have to think about

it like these things about AI innovation hubs I support them and I think they are super cool but you have to prioritize sometimes 33 % is the rate of success with internal builds versus 67 % of buying AI tools if somebody else is doing it better it's just more efficient to outsource

The thing, me also, as a co -founder of a startup, is that you can't just buy 20 tools and paste descriptions all the time, like that is going, the cost is going to be insane, and then what you prioritize.

So then here is what is valuable to spend a specific amount of time in your day to day, for example, like a 15 % of your time with your team, kind of like brainstorming and learning how to be a little bit more tech savvy.

so then you could implement that time into building your own solutions that specifically would help you I think I have a slide out of that later but basically specifically help you either be more competitive or have something special for your core product that that others won't offer so if but if it's

about kind of like a process that is is very well known and it's complex to build then it just doesn't really make sense to spend three months trying to reinvent the wheel.

4) AI Isn’t Great at “New”: Document Processes and Provide Context

Okay, then the other issue that happens is that AI is not very good at new stuff. So probably our little niece of six years old is gonna be way better at understanding how to they choose or made a bet from the first time that AI, because AI just needs, it's a predictive machine based on a lot of data points, so it kind of like it needs context in order to create something that predicts the next thing, this is why you write the sky is probably will say blue because it just learned that from many times before.

so what happens here is that if you kind of like expect AI to understand your systems and then create something new without any references in the back it's just not it's not gonna work it's gonna

break at some point so instead of like looking a way of automating what we're currently doing or just like using a I said like do it better what we want is to check the processes that we have, so document your day -to -day, what are the bottlenecks, how does it look like, what are the outcomes that you want to get, and then we figure out how AI goes inside that strategy.

1So we get rid of that AI tool mentality, then with this documentation process we create knowledge bases that agents and models can use to feed and understand our workflow, not from zero.

5) Organizational Unreadiness: Skills, Adoption, and Change Management

Okay, the last pattern is organizational unreadiness.

So we think, I ran a survey to understand like how people think they know about AI. Most of them, this survey went to 100 people from different countries, and the mean, the average was, I am confident around six to seven.

And when I asked them, okay, what are the AI skills, I know how to build automations and prompt very well it's just like this is insane like it's kind of like this paradox like people just don't know what they don't know uh so the the what is happening is that right now people

are using it 64 percent are of workers are already using ai but only 36 percent report that they feel satisfy with the learning that their companies are giving and also there is a in my area of expertise I see that adoption for women is three times more slow than for men that creates a huge gap in opportunities as well so the

thing is that you're not gonna get anywhere just by knowing how to prom very well AI tools evolve so we need to create also structure plans of of implementation in companies or in our day -to -days to keep up to date.

And I know that it can be challenging, especially for small business owners, but there are ways of doing so.

So you just know that you can create your own automations to feed you news, to like block specific times in your calendar to put these things, like to test new things as well.

But you will prioritize with, with, I hope that come out later in the Claude tutorial, but you could use your agents to kind of like give you guidance as in understanding.

So if we have a 3D printing or if you are a doctor, it's just like specifically how can you use AI and what is new in the AI world

that specifically affects me. So then you cut out the noise. Like all of these AI influencer, there's so many.

This is the prompts that you need. This is the tool, like forget about that. Like it's just noise. you need to focus on your outcomes what do you want to get not the tools okay so

A Practical Framework for AI Adoption

yes this is what we want we define the business outcome we map our structure we identify them where the value goes we experiment with a specific data set and then we build a replicable process okay we define where should we build when when should we build and what we should buy. And kind of like this is the framework, no?

So if it makes it harder for competitors to copy, then you build it. Like your core business is something that we need to identify.

So imagine that you are a baker, like a pastry chef who is very, very famous for your cakes. So you don't need to kind of like know about the chickens that lay the eggs.

You don't need to know about where the sugar comes from. You don't need to make the ovens yourself, right? But you need to perfect your recipe because that is what you're buying.

That is your core business and your core value that you're offering. So if this AI opportunity helps to enhance that core offer, that is something that we could consider building in -house.

If it doesn't, it's kind of like losing time. so we try to outsource it. Okay, we don't redesign workflows that are broken, sorry,

we do redesign workflows that are broken so we identify where are these bottlenecks happening and then we create new workflows around potential AI implementation. And the budget for AI is not, shouldn't go on tools should go on people and structures so you go about

organizational change about ourself our own education because then you'll be more selective and and the return of investment is like five to one more effective than just testing random tools that a random guy said that is gonna make you rich on the internet.

Implementation Roadmap: Three Phases and a Pilot Mindset

Okay, how do you go specifically with that? Three phases.

One, you like clean your objectives, select vendors if you need those outsourcing decisions, the decisions tools, sorry, to outsource, and we build these knowledge bases that we're going to need for AI implementation. So we need this little backend before implementing any fancy AI stuff.

Okay, now we're on a pilot. We're not going to automate everything in our business. We just go with one.

I do recommend a lot of finance though. That is a headache. So particularly for startups, just get rid of the invoicing. So if you can do that, then you put specific APIs of just say, okay, Okay, I don't want to see my accountant saying, like, where is this other invoice that you missed?

Again, it's just, okay, so how do I go towards that? One thing. And then when I've identified that process, then I move to the next one. So then we scale systematically.

Okay, we don't expect results in two, three months. We expect results in a longer period of time.

Conclusion

Yeah, so whoever wins, to conclude, is not those who implement just random tools or because they are lucky or because they have the best consultants in the world. It's because they build systems and workflows.

So thinking first in your product, in your offer, and then in how everything else fit that need. Thank you.