Are you AI prepared?

Original article was published on Artificial Intelligence on Medium

Are you AI prepared?

From first principles

“Are you Artificial Intelligence (AI prepared)?” is one of the most hyped questions on an already highly hyped topic.

At the same time, it is a very pertinent and timely question. I know this because many people ask me how their organisations can be more “AI prepared”. Sadly, many also qualify the question with “whatever that means”.

Answering the “Are you AI prepared?” is big business. Consultancies take millions to “study your organisation” so that they might produce a report that evaluates your “AI Preparedness”. Oddly, a lot of the people who ask me the “Are we prepared?”-question, have already commissioned (and sometimes even read) such reports about their organisations. And yet they still ask the question.

The problem with the “Are you AI prepared?” question is that while many consider it important (and it is important), few have good answers. The goal of this article is to solve this problem. Provide a reasonably good answer.

“How can you succeed, where multi-national consultancies have failed?” you ask. Well, read on. If I provide you with a good answer, you will know for yourself.

Whether or not you are “AI Prepared” depends on the answer to two questions. We will consider each next.

Are you Data Prepared?

People say there is an AI revolution. There is no AI revolution because most of the ideas around AI are old ideas. Some over 50 years old.

The real revolution is not in AI but data. Data can be collected at rates and efficiencies never known before. The same applies to transporting data, storing data, and processing data. These revolutions in sensor, network, storage and processing technology have driven the “Data Revolution”. If there is an AI Revolution, it has merely piggy-backed on this data revolution. For example, the theory behind Deep Learning is over 50 years old. But the data revolution has made it practical. Hence, if you are not Data Prepared, you are not AI Prepared.

What does it mean to be Data Prepared? Let me explain with a simple example.

Suppose you run a small shop. Every day customers walk and buy whatever they want. And once a week or so you replenish your inventory, and dispose of old stock. Each time someone buys something, you might enter the transaction in a exercise book, and you do similar “book-keeping” with replenishing and disposing of inventory.

In this “pre-digital” example, there is still much data. Most of it recorded on paper. “How data prepared you are” depends on what proportion of your data is in a form AI systems can process. If all your data is in such a form, then you are “100% data prepared”.

Even today, many organisations are like this “pre-digital” shop. A lot of their data is on paper. Often, even data which is in some digital form cannot be easily processed by AI systems. Hence, “becoming data prepared” consists of transforming data collection and storage into an AI friendly format.

Are you Decision Prepared?

Suppose our shop-keeper becomes 100% data prepared. She replaces the “exercise book” with a “digital inventory system”. Now, every piece of data in the business is not only digital but “AI-systems friendly”.

But now what? What’s the point of all this data?

To answer this question, we must first understand how AI uses data. Most of AI is about using data to answer questions. If you have millions or billions of pairs of questions and answers, you can “train” in AI, to answer new questions. For example, if you have many pairs of human faces along with the name of the human, you can train an AI (a.k.a. a Face Recognition system), to answer the question “Who is this human?” given a picture of a face. Similarly, our shop-keeper could answer the question, “What inventory should I purchase?” given (inventory, “How much of that inventory sold?”) data.

Hence, the power of AI comes from using it to answer questions that your business cares about.

Let’s again consider our shop-keeper. She cares about maximising profit. She does this by selling goods that her customers want. And not selling products that her customers don’t want. Hence, she has to make two decisions:

  1. What products to sell, and how much?
  2. At what price to sell the products? (Note, even products that customers want must be sold at the right price. Too high, and no one buys. Too low, and she makes a loss.)

Both these questions can be answered by an AI system that trains on the data that our shop-keeper already has. And if she uses such an AI system to answer both questions, we could say that she uses “AI optimally to make decisions”. This is what I mean by “decision preparedness”. Whether you know what decisions you want AI to help you with, and whether you are using AI to help make those decisions?

If you are using AI to help you with all your decisions, you are 100% Decision Prepared. If your 100% Data Prepared and 100% Decision Prepared, you are 100% AI Prepared.

But I’m not a Shop-Keeper

“But I am not the small shop-keeper on the corner”, you say. “My business is worth billions.” Fair point. How do we go from small to big?

The principle is the same. If you run a large organisation with many parts, you need to break down the organisation and its decisions into smaller chunks. And ask the “Data Preparedness” and “Decision Preparedness” questions for these subsets. Then you can aggregate the results for an “aggregate preparedness” for the company as a whole.

We could even go beyond this. And aggregate multiple organisations or even a whole country.

What Next?

If you run (or are part of) an organisation, and want to know if it is AI Prepared, I’d recommend trying and answering the Data Preparedness and Decision Preparedness questions for at least a small subset of the business.

After you identify some subset with an untapped opportunity, you could try and solve the problem (at least partially) using AI. Once you realise some value out of the exercise, you could move to other parts of the organisation.

“But wait!”, you say. “We don’t have any AI expertise. We don’t have the people, and we don’t have the technology”.

That’s ok. The hardest step is the first step: The Data Preparedness and Decision Preparedness questions. And, happily, you don’t need any knowledge of AI to answer these questions. All you need is a good understanding of your business (which, hopefully, you have). Once, this first, difficult step is complete, hiring people and building technology is easy (relatively). I will share my thoughts on these later steps soon, but in the meantime, you should worry about the first step.

So, are you AI Prepared?