Introduction

Thanks everyone for showing up for this talk. I was just wondering how many data scientists, data analysts, machine learning engineers are here. Maybe you could show me by a show of hands.

Okay, interesting. And are there any managers of data teams? Okay, good to know.

Well, maybe the first thing to mention might be a bit of a disappointment. This is not the technical talk. This is, I think, the practical talk.

And also, Alphabet is not an AI tool that we developed. And also, I'm not going to mention all the letters in the Alphabet. That's not going to be this talk. this presentation.

The Alphabet and AI

But the alphabet is something that we use at Remain Solution that we believe drives AI to true potential.

Background

First, a little bit about me. I have a background in econometrics. I am very passionate about AI and about what it can do for our society, especially in terms of sustainability.

And since this year, I'm also an entrepreneur. I co-founded Remain Solution with Reinier Juistra, who is also here. And well, we have something to share with you.

We founded Remain Solution because we strongly believe that AI has the power to have positive impact in society. And we feel that we have almost an obligation to do this by bridging the gap that there is between the potential of AI and the actual value and impact that you're now seeing.

AI Implementation Journey

With our clients, we use an AI implementation and we tell them, if you work with us, we will implement from A to Z. What you see here is sort of the journey towards tangible value, so real impact with AI.

The road to come to this tangible value looks like a straight road. You just have an idea, you put it into a pilot, then you have a proof of concept, so it's actually a working model, and then you should expect that there is real impact.

In real life, this is not how it works. It looks more like this. And to come from this POC, this pilot, to tangible value, there are a lot of sort of hiccups, bumps in the road to get there.

Use Case Example

And I would like to explain to you using a use case. This is an old use case from one of our current customers.

I would like to explain to you which yeah, bumps in the road there are and how you could actually solve them to create a tangible value.

And the use case that I would like to use is pretty, yeah, I think it's an example that a lot of companies have that we want to predict customer contact.

And let's start with the first one.

Introduction to the Alphabet

Or maybe we should first introduce the concept of the alphabet. So as you can see, it's just six letters of the alphabet. So it's not going to be 26.

And this is, I think, a very sort of nerdy way to explain the six factors that we believe are essential to create a real impact. So it's about ambition, boundaries, culture, data, ecosystem, and focus.

And these six things, you don't need... like 10 or 20 people to achieve these six things. You can actually do them if you're just a single data scientist within a company.

You just should take all these things into account to make sure that what you're working on doesn't end up on a shelf. And for me, it's super simple to remember because it's A, B, C, D, E, F. So it's also very important to remember. So it's simple but very important.

Key Factors for AI Impact

But why is it important?

Ambition

So for this use case, the first thing in most companies, what is the AI ambition for this specific company? That's the first question.

And where I was working, there was a clear absence of any strategic goals. There wasn't a clue of which way the company or the specific team was going.

So my first goal as a data scientist was to actually find out what are the most important problems that we have within the company that I want to contribute to. Because if you're not looking for the strategic goals of the company, you're working on something that's suboptimal. You want to be working on something that you're sure that will help the company create a significant value with AI.

What we do within Remain to make sure that you're clear on the ambition of the company is it's mainly stakeholder management. And actually all the letters that we are going to discuss is about stakeholder management, making sure that the people that you're working for, that you're creating this AI tool for, know what you are doing.

And to understand what the ambition is, you can, of course, host a brainstorm. And it isn't that difficult to do it. You just have to do it.

You can just use post-its. It's very simple and you don't have to do anything digital or complex to just understand what is happening in the minds of this company, of the people of the minds of this company. You can do interviews, for instance, you can interview the board of your company.

It may seem like it's very far from your position, but that is, I think, the responsibility that you have as a data scientist because not a lot of people can do what data scientists or machine learning engineers can do. So your skill is very valuable for the company and that's why it's also very important to know what you're working on adds value to the company.

And in the end, it's very useful to have a roadmap. So know what you're heading for, because you might need other teams in the company, for instance, the IT team. And if you have a roadmap ahead of you, you know which solutions you need to actually create value for the strategy of the company. You also know which teams you need.

Ecosystem

Where we ended up was the ecosystem. But the ecosystem for this use case, we developed a working AI model.

We knew it had significant value for the company, 1The only problem was that the priority at the IT department was not the same as we had. So there was no possibility to actually deploy the model. We only had a sort of experimental environment where we built a model and there was no way to actually deploy it and put it into production.

So to make sure that doesn't happen to you, you can do an impact scan before you start modeling. So before you even start coding, look at all the sort of things that you would need to get it into production, to actually create the impact that you want.

Because it's not only about creating a successful model. A successful model is something that is implemented on a robust way that is sustainable also in the future.

And also the roadmap here is very important so that you're talking to the right people within your company. And it doesn't take a lot of time to do this. You just need one conversation with the IT department and then you know, are they willing to help you?

Where is their priority? Or do you need to go higher up in the company to help them make the right choices?

Or are you not working on the right AI model? Is it not aligned with the strategy of the whole company? Which might also be the case.

Endurance

So endurance is the one that I want to highlight here. The use case that I'm talking about was a use case within one specific part of the company and it was usable for a lot of parts within the company. The part that we developed it wasn't interested when we actually built the model, they said, well, it's not for us. It's not something that we want to use. But we were sure it was a significantly good working model that it could help the company.

So we were already working on it for a year. And I thought, well, let's not waste all this time. Let's just go to a different part of the company and try to sort of sell it within the company.

So you have to have a little bit of an endurance to hold on to your model if it's successful and you think it adds value to the company. You can also go to other stakeholders and try to sell it there.

I think the reviews from a scrum perspective are very important in this point of view. Because with the reviews, you can always talk to your stakeholders, get the feedback and make sure that you're doing something that they want. So you have the right focus and what you're building is also what they want you to focus on.

And an impact effort matrix, if you make this beforehand, you're sure that you're working on the right project, which makes sense for the company in terms of impact and effort. So if you, for instance, had a brainstorm session with a lot of stakeholders, you can, together with them, create this impact effort. And then you know when there's a moment in time that you and your team think, I'm not sure, should we continue? Or maybe a lot of variables are not working out the way that you want. You can always have in the back of your mind, yeah, this is the use case that the company wants. So we should put in a little bit more effort.

Data

data, of course, a very logical one. So you should be aware of what is the quality and quantity of the data and how is the governance organized and what is the data availability.

So also the impact scan is very important here. If you look before you start, you look at the data and what the quality and availability of the data is, it can well, help you select the right use case. Because if there's no data availability, this is not the right use case for you.

Actually, in this case, because we sort of sold the idea to a different part of the company, we already had a successful model, but the data was less easily available there than it was where we made it. So it actually became a problem later on in the implementation of this model. And one of the main reasons was the data warehousing.

There was a new data warehouse and it was just differently than the one that we used previously. So data warehousing is of course really important in terms of how you manage your data and is everything that you need there in the way that you need it.

Boundaries

Boundaries. And yeah, resources wasted on an illegal project.

This is something, this use case took a lot of time to get to the end state in this company. So there were also other issues in terms of data, infrastructure and everything.

And we, from the start we've, spoken to risk and compliance officers and there were no problems. And in the end, there was one privacy officer that woke up and said, no, this is not something that you can do in this company.

So it was quite a surprise for us that there was another privacy officer that we didn't know. And it almost was the case that it was a project that we could not continue with.

So this for us was a big lesson that you need to always have risk and compliance from the start of the project, directly talk to them and invite them to all your reviews, make sure that they are aware of all the things that you're doing. And especially in large companies, there are a lot of different people who have an opinion about this. Also people from ethics or legal.

And don't forget to include these people. Always invite them to your reviews. Make sure that you're available for their feedback.

And especially don't be afraid. to talk to them because it might be a little bit sort of daunting because they might say that's not going to work and then you're just disappointed. But it's better to know it upfront than the other way around of course.

Culture

culture. So C is for culture. You could also use the C for change. Of course, this is a really big thing.

I think the chat GPT has helped at least me a lot in terms of the culture and how people look at AI. At the same time, now everyone knows what it is, but that's the only thing that they see.

So there are things about that you could maybe want to change. But the fair and limited understanding of AI, I think, is still a problem.

And you can do so many things to help people within the company also think of their own use cases. A lot of people don't know yeah, I see sort of the value, but I don't see what it can do for me.

And just, again, the brainstorm is so easy. Just ask people to write down some ideas. What would you want to change in your role or in your function? Or what is your biggest frustration?

It doesn't have to be about AI or... specifically about data, but the technology that we now have is so strong that you can literally almost solve any problem. So that is I think the beauty now of AI that you can try to link it to so many problems, but you just have to do it and tell people that there is actually some way you could help them and solve their problem with AI.

And user sessions, of course, are also very important also for adoption. So when you created a tool, make sure that you test this tool with users, not at the end, but as soon as you have something that you can test.

Conclusion

So this whole road leads to this tangible value. And this is actually also the concluding slide.

So what we tried to... sort of convey here is that these six things which are easily to remember I hope everyone leaves here with these six letters and now sort of remembers them also is if you're if you want to do something with AI especially if you want to create impact you should just look at these six different things and you can do it in all different ways we use this workshop brainstorms and

There are many different ways in which you could do this, but just keep it in the back of your mind that you are not only programming something or using an AI tool to make impact.

The impact is only there if it's used, if it's integrated in the current infrastructure, and if it's still there within one year. So then you have the actual impact that you're looking for.

Final Thoughts

Thank you for your attention.

If there are any questions, I would be happy to...