I’ve lived outside of my home country for 5 years.
That means that for 5 years I’ve bought groceries, asked for directions, made friends, argued, dated and a bunch of other stuff in a language that is not my mother tongue.
Not speaking your mother tongue can be hard and scary. Try chopping vegetables with your non-dominant hand. It feels weird, right? That’s more or less how it can feel to deal with some aspects of life in a language that is not yours.
I saw the same fears and uncomfort with the people I coached, the first time they were trying to define AI projects. A new, complex technology with a new vocabulary would scare off anyone.
I got so used to the language of people not fluent in AI, that it takes a single sentence to gauge one’s confidence and understanding of AI. I can cluster the types of conversations I had in four different groups, with four different levels of AI understanding.
You probably know that Netflix uses AI to recommend movies to its users. Let’s now imagine the Netflix recommender system project would be pitched by four managers with four different levels of AI understanding.
We’ll use AI to give users the best movies.
What’s wrong with this explanation? I’d say the general problem is that it’s too generic. What does “Best Movie” mean? Is movie A better than movie B based on its production costs? Or maybe “best movie” means the longest one, to keep users on the screen for more time? How are you going to do it?
This definition is so vague that it’s probably hiding foggy ideas and poor understanding of what AI is and what it can do.
We’ll use a recommender system to suggest users the movies they may like.
Getting better! Now this manager is using AI-specific terminology (good she mentioned “recommender systems”), but most of all she specified an action: the goal of the AI is to provide movie recommendations based on the user’s taste. Quite an improvement from “best movie”.
We’ll use data from users’ interactions with our platform, demographics and view history to recommend movies that they’re most likely to start watching when they log in.
Here’s another big step forward. Now this manager has been both much more specific on the data she intends using, and on the specific action that needs to be taken by that AI system.
The Data Scientists will take care of translating all these bits of information into more specific features (for instance turning “demographics” into “age, gender,…”) , but this manager has shown that she has already thought about what data can be relevant, and probably also made sure that’s available.
We’ll use data from users’ interactions with our platform, demographics and view history to recommend movies that they’re most likely to start watching when they log in. We’ll measure the effectiveness of the project with KPIs like the time the user spends deciding what to watch, the satisfaction on the choice, and long term effects on retention.
Perfection of AI-project fluency. Being aware of how to measure success in an AI project is key to achieve measurable success, and the icing on the cake is the differentiation between short term, easy to measure KPIs (time to pick, satisfaction) and long-term, more strategic ones (churn).
A manager capable of expressing her thoughts in such a way doesn’t just sound fancier. She also has much higher chances of getting her AI projects off the ground. They have a stronger business case, they’re ready to be processed by a Data Science team, and her career in AI can’t do anything but grow.
Going from one to four may seem a lot. But learning how to do it is actually fairly straight-forward: the keys are a good understanding of AI principles and what makes a good AI project.
This is why we designed three courses that will turn anyone into the AI expert that companies need nowadays. Learn more about them here: https://learn.ai-academy.com.