Rafael
Rafael CRO @ FetchBug, basketball fan. Wine sorcerer and hype beast, official Jordan athlete (and customer).

AI from 0 to 100

AI from 0 to 100

Yes, this post is about AI, and rather you think of it as “Artificial Intelligence” or the fancier “Augmented Intelligence” you have heard lots of webinars, use cases, moral debates, etc.. You have heard so much about AI, that having AI initiatives has become an urgent need for all your enterprise. But there is something I must say to almost all of you: you are not ready for AI. And trust me on this, I don’t wanna be a wet blanket, what I want you to understand is that you are not ready today, but there is a chance you can be, tomorrow.

Big data [such as AI, Data science, etc.] is like teenage sex, everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it… — Dan Ariely

There is a lot of hype, FOMO, and disinformation that has caused fear, uncertainty, and doubt; this, plus the lack of evidence that AI can actually have an impact (or at least a short term impact) on the day by day operations is making it harder than ever to even consider small AI initiatives. The truth is there is no doubt all industries will benefit from data sooner or later (you are now in a position to lead or follow), some including Financial Services, Insurance, and Retail have already used data to have an impact on their business, and the main reason is that they already generate big lots of data, and as you may know, that is the base of all AI initiatives.

Hierarchy of needs

So this post is about how to go from 0 to 100, so first of all, I need to show you the structure that you should follow in order to have AI, deep learning, and data science initiatives running smoothly at your company. Here is an explanation by Monica Rogati that will help us:

Think of AI as the top of a pyramid of needs. Yes, self-actualization (AI) is great, but first you need food, water and shelter (data literacy, collection and infrastructure). — Monica Rogati

Let’s take a look at the image above, at the bottom of the pyramid we have data collection, we all want to get to the top of the pyramid but we don’t even know what data to use, and regardless of whether you are a startup or a large company you should first know that everything that happens in your company or with your customers can be usable data, log all interactions and get data from sensors, to put it simply: collect it all.

Now that you have decided to collect all the data you can, you need to know where and how to store it in a way that is easy to be accessed, analyzed and worked with. As Jay Kreps says “Reliable data flow is key to doing anything with data.”

The next step is key in the process of taking advantage of your data, this step is data cleaning, and even though it is mistakenly one of the most underrated activities in data science, it is in this process in which we find out we are missing data, our sensors are unreliable, we are misinterpreting a flag, we are full of duplicate data, and it may even be the case in which most data is trash. On the bright side, this process will help you become better at collecting better data, developing a discipline of data generation.

Once you finally got useful, clean data from different sources, then you can start doing what is traditionally thought of as BI or analytics, you can now define metrics to track, observe and analyze seasonality and sensitivity to external factors; you can start to segment your customers based on behavior, demographics, etc.. At this stage, you will start to think of what would you like to predict or learn, and you can start preparing your data by generating labels, you can also start using quality external data to get the full picture.

You are now at the promised land, the sky’s the limit , you are instrumented, your data is organized and cleaned, you have dashboards, labels and good features, you are ready for all the use cases you have been talked about, you are ready to start experimenting, deploy ML algorithms with your own and external data, keep adding new signals to improve performance and finally…bring on the AI. 🎉

Final thoughts

Now the only advice I can give you is: start by processes, products, or services verticals, generate your pyramids for each, measure success, ROI and set a data-led culture. Eventually, you will grow your pyramid horizontally and become a company comfortable with data, clean, useful, accessible data. Remember to be strategic about the use of technology, focus on the results you are trying to achieve instead of focusing on the technologies you will use, and there is a big chance you will make a huge difference to your users, clients, and your company.