Introduction

I'd like to introduce myself. My name is Anastasia.

I am Enterprise Solution Engineer at Salesforce for the past few years, specializing in high-tech and gaming verticals, as well as data and AI for the regions of Mediterranean and Central Eastern Europe.

So I have a great experience working with many enterprise and mid-market customers, which is what inspired the talk today.

Preparing for the AI Revolution

So what we'll be talking is It's pretty much like a boot camp focus on how companies can prepare for the AI revolution.

So how can we introduce AI into our operations and unlock its full potential.

So whether you're working for a large corporation, have a business on your own, or working for a smaller company, the goal of today is I will be giving you a framework that I've introduced and talked to many of my customers about. And like any AI tool, you have to see how it fits within your business. So you have to customize it in your mind.

So as I said before, I've been working for an enterprise. This is some of the examples of, let's say, businesses that I will be talking about here today.

So when I ask, let's say, what business these companies are in, many of you think, well, Netflix is a streaming platform, Tesla, autonomous vehicles, Uber. taxi service, but what actually is interesting, those companies have been developing and becoming AI companies for the past few years.

So I'll be focusing a lot on talking about them and using them as an example of AI companies and AI development within different industries.

Utilizing Existing AI Tools

What I'd like you to focus on here now is that less than 5% of companies actually create their own AI tools and less than 15 fine tune it.

What does it mean? It means that we have enough AI tools right now to help your business to start up and running as an AI business.

Many companies think, okay, We need to develop our own AI tools to do this and this and this and that. To be fair, whether you're a small company or a large enterprise, you can find the tool that satisfies exactly the need of your company, or you can get another tool to work and fine-tune it within if you do have to do so.

Three Pillars of a Successful AI Project

Without further ado, here are the three pillars of a successful AI project.

First and foremost, the company has to be digital. To have AI working, you have to have your data within digital worlds. You have to exist in a digital.

Because what happens when you bring AI tool and you have your data living on paper? Well, something like this, to be fair. If we have AI tool in the middle of all the paperwork and sticky notes all over your laptop, it can't give you any value, can't do anything within.

Next step is AI maturity. So I tend to recommend most of the companies to start here.

This is an Gartner AI maturity graph. This is to pinpoint where your company currently is when it comes to AI readiness. So you can establish a strategy and measure your organization against it.

We have five levels, awareness, active, operational, systemic, and transformational. So in the beginning, you have some early interest in AI and you can kind of get a bit too excited about it. Then you have a little bit of experimentation, but mainly within data science context.

Then you have AI in production, which is starting to create some value. You have maybe some process optimization. Then AI is used for digital processes and chain transformation.

And you have some new digital business models. And then level five is AI is part of a business DNA. Now, only few companies are at level five.

That will be, let's say, Google or Microsoft or Meta or Salesforce. Simply because, let's say, myself working at Salesforce, I will be using AI tools. They'll be working in the background, and I would not even realize that it's happening.

We'll enter Slack. Slack now uses AI. We can go into search. We can ask a question. AI would work to give us an answer by summarizing all the conversations that happened, let's say, within that chat in particular.

1So no matter where your organization is on the map or how far it has to go, you need to ensure the strategies are highly adaptable and you need to leave some room for experimentation because we all have very different businesses with different needs and timeframes.

Keys to Unlocking AI Success

Now, step three is internal commitment and plan. And that will be my next focus.

So we're starting very kind of on a global scale and we're going to narrow it down as we go because we have a mix of people in a room today, some technical, some business. So...

So here are some keys to unlocking that AI success. Step one, well, data is king. So you need to ensure that data is clean and accurate so we can use this for an AI models. 1Now for that, usually some companies have data sciences working within or people who are cleaning this data or there are tools available in the market that can clean the data for you.

So let's use an example of Netflix, Uber, and Tesla and how they're tackling the data that's a problem and how important is the data for the AI tool that they're using. For example, Netflix's Apache Iceberg. So it's a tool that's optimizing large-scale data processing by leveraging machine learning models. So it can predict how the data will be used.

It can adjust how data pipelines work based on some real-time needs, for example. What it does in turn, it reduces costs and speeds up data handling. So Netflix has all over data every day, loading, updating, user needs, user wants.

Let's say right now we're going close to Halloween. For example, what we will see, we'll see some Halloween movies pop up in front of us. That's some data, that's prediction of what the user might need. Because you have that data from the past, Netflix can predict what we will want to watch now or how it can hook us to stay longer on their platform.

Uber have a hoodie. So hoodie, it helps AI. It's not an AI tool, but it helps AI by keeping data fresh and ready to use in real time. So it projects in an open source data platform designed for large scale data lakes.

In simpler words, let's say how Uber sets its prices. So for the past few days, it's been raining a lot. And what I've noticed when I wanted to call a taxi during the rain, it was really hard to get it. And the prices were climbing up a lot.

Why? Because the demand increased. higher. That is Uber's platform running on the data. More people requesting it based on the past data, people will be requesting it during the rain. So in real time it can update the data and match the price to its demand. So, yeah, this makes Uber's AI system faster and more accurate, being able to operate in those events.

Tesla uses a lot of AI tools. What I'll be focusing on today is more of a real-time learning. So this is where the AI learns from data that is collected during an actual driving. So it analyzes the situations, for example, where something might have gone wrong, let's say false braking, to improve the future performance of the vehicle.

Or also simulation training. So using simulations to create challenging driving scenarios that, for example, for that you don't have to actually be in a risk. So AI can understand it before actually going through the risky situation. That is used by analyzing data of what situations might have happened. what could happen and simulate that environment and of course regularly updates with the new software because you have more and more and more data learning and every new vehicle is enabled with more data now

Focusing on User Experience

Step two, focusing on the user experience. This is very essential. So many think, okay, AI should do this for me, or it will take my job, for example.

Well, AI is not at that stage right now. And what it should do, it should complement human expertise, not replace it. The best way you can use AI is to have this human-AI partnership.

So, this is some Salesforce data here, which is publicly available. So, 61% of customers say that they prefer self-service, like go on a website and just talk to a chatbot.

The second something, let's say, is more emotional or where they need to talk to someone, 65% say they need someone to immediately respond to them when they contact the customer. So let's say if your T-shirt doesn't arrive or it arrives false, you want to talk to a person straight away. You don't want to talk to a robot.

And having this partnership is what's essential when it comes to adapting AI to your company. How you can do that? So use AI as your partner.

Identify tasks where AI can automate repetitive processes so that can give you some more free time for strategic work. AI can act as your valuable assistant, for example, giving you insights and recommendations. The ultimate goal is to supercharge your capabilities but not to replace them.

Change Management

On to the next one, change management. So this is where we need to really address the employee concerns and provide the training. So I worked with this one large corporation and we went to see them and the CTO spoke to me and says, well, I bought all of those copilot licenses and no one is using it.

So I'm losing money. I don't understand why this is happening and this is just really unfortunate. And they've done no training to explain to their employees how to use them and what's the value of using that tool.

enabling your workers to actually know how to use those tools. So 40% of workers don't actually know how to effectively use GenAI tools. 43% don't know how to leverage them and use them within the trusted environment.

And 53% don't know how to get the most value out of it. So by providing regular trainings within your company, you will ensure that the workers can actually adapt the AI tools.

And for this, I can spend probably half an hour talking just about that, but it's very important to ensure the security. So your data is protected. The AI tools that you're using, let's say the trust is not carried out, the data stays within your company. When you're choosing an AI tool, make sure they have, let's say, a trust layer, so the data never leaves that.

And now quickly to run through whoever's technical here of course knows it's very simple explanation let's say of how it works. So step one, you train a model. So you get all of this data, then you go through the cleaning and preparing and manipulating the data.

Here you can use a specific tool that can do it for you or you can try to do it internally which is a very complex task. Then you train the model with the clean data. You test the data and you improve.

After that, you use that model. So you take the real world observation, put it into a model, and you get prediction out. All the data goes into model that you trained before.

Then you have a prediction or you engage with the model. That doesn't work, that works, give me this, give me that. You send it back into feedback loop and more learning goes into your model.

Starting Your AI Journey

How can we start? What do I do? What kind of models I need? I don't really know.

Here is the little step-by-step guide. Step one, you identify your need. Usually, you have three areas.

Repetitive tasks, so something you do too many times per day, data overload, or room for error reduction. Then you choose your play. You have efficiency and capability play.

Efficiency play is improving existing processes. You can make them faster, cheaper, or you can reduce the errors. It focuses on this automation or optimization of workflows, like data entry or report generation.

capability play, unlocking those new capabilities of functionality. So this is when you really leverage those AI tools to analyze data, identify patterns, make predictions like ChatGPT, for example, as well. That could be a sales forecasting or really advanced conversational AI.

Examples of AI Applications

Here are some examples of how AI can help. Of course, there's a lot more, but on the left, we have a conversational agent. like let's say ChatGPT or Gemini when you ask a question and you receive an answer.

Then you have something like a task automation. So I want to send an email follow up in two weeks, schedule it for me or send it in two weeks. You have AI doing an action.

So you have planning or next best action AI model like ClickUp, Zapier or let's say Slack when you can schedule certain workflows. Then you have a bit more of complex situations like image classification that is often used in medical fields. This is where you have image, visual understanding, and then an output, which is like a person or a child.

We have Amazon recognition or Google Vision API. And of course, prediction of more computational load, like time series. Let's say you have a graph, you have time series forecasting, and an output like a stock will grow by 10% next year. Here an example would be Tableau, Azure, or AWS.

Conclusion

And to wrap up, because I have to wrap up, let's look at some of those companies as examples of how AI tools can be leveraged in different companies. So after knowing all of that, you have to go back to a company and understand what is my need? What does the company's need?

What Amazon does with AI, it analyzes images and videos to improve this product recommendations that they have. They also make supply chain more efficient by forecasting the demand or optimizing inventory levels that they have.

Okay, Johnson & Johnson. Part of Johnson & Johnson is Neutrogena. Neutrogena has an app that analyzes your face. So you have an app that you can load on your phone. It uses your flash and your camera to scan your face. And based on that, It can upsell you on which products you might need. So to make sure your routine is good.

Then Boeing has an agreement with Shield AI. They're collaborating on autonomous capabilities. What they've been experimenting with lately is using AI in air traffic management system like speed recognition and computer vision. It sees an opportunity to use AI to take pictures after an aircraft has landed to see whether there has been any potential damage and avoid it in the future.

J.P. Morgan uses it to help with fraud detection and also has a GGBT-like model that analyzes speeches from the Federal Reserve over the last 25 years to try to predict or depict some signals to gain an advantage in the markets.

And ExxonMobil, of course, in an energy industry which is now Very different, what does AI has to do there? But it has ML algorithm to help avoid equipment failures. So equipment failures can cause major delays, like if some platforms are unmanaged. So it also uses to increase production and automate certain jobs.

So with that, I'm going to wrap up.

Connecting and Further Conversations

And what I want to leave you with is I would love to connect. And quite often what I do as well is I can have a conversation with some startups and scale ups.

So if any of you are interested to take this conversation a bit further than 15 minutes. I'll be more than happy to jump on a quick, free, 30-minute conversation call where I can talk to you about how you can try to start introducing AI into your company, how you can leverage that, how you can automate the processes.

Yeah.

Thank you so much.