SHOULD YOU USE A.I IN YOUR BUSINESS?

Should you use A.I. in your business? 

And if so, how?

Maybe you've quickly become a generative A.I. expert. Or maybe you've been playing around with GPTs and know enough to be dangerous. Or maybe you have no idea what all the fuss is about but it seems dystopian. 

Either way, we're all acutely aware that the A.I. bandwagon is running out of room. It's time to hop on. 

Right?

Look, no one can give you the exact right answer, but I'll give you a framework to answer it for yourself. 

First, Let's Ask the Right Question

The part of A.I. that makes all the money is not so much about getting the right answer as it is about asking the right question.

So let's make sure we do that here too. 

This new flavor of A.I. (and it is indeed just a flavor, is not the kind that's going to kill us all, yet) isn't all that new. 

In 2010 and 2011, I co-invented the first commercially available natural language generation (NLG) engine and platform at Automated Insights, which is a fancy way to say that we taught computers how to write articles based on data. 

While we used both A.I. and machine learning (ML) to enhance the engine and the platform, our product was neither pure A.I. nor pure ML. Since those early days, NLG has been combined with natural language processing (NLP), a science that started going mainstream with Alexa and Siri, and has now evolved to become generative A.I. -- what we think of as OpenAI and ChatGPT and the like. 

But back in 2010, the term NLG hadn't been coined yet, or at least it wasn't mainstream enough to get on into our consciousness, so we referred to what we were doing as automated content, because automation is like 90 percent of what makes A.I. seem like magic and money.

So the real question you should be asking is, "How much automation should I use in my business?"

And to get to that answer, we have to understand the difference.

Thinking Versus Acting

Machine learning is "thinking" and automation is "acting." As technology continues to blur the lines between machine learning (thinking) and automation (acting), we roll it all into one smart technology called A.I. 

Think of a self-driving vehicle. The ML tells it where to go, the automation executes those decisions. Those technologies are unrelated and usually self-contained, but they have to work in complete harmony. 

This is what ChatGPT is. It has an engine that takes in a prompt and uses ML to turn that into a well-formed question and answer -- the logic that decides what and how to write, draw, or code. Then it uses automation to generate the words or pixels or syntax and make the results available somewhere for someone to see. 

This is where today's A.I. seems magic enough to be human because we do these tasks every day. Our brains take in the question and provide the answer, and then our mouths or hands act to deliver that answer. Thinking and acting.

Develop Use Cases for A.I.

There are a ton of derivative use cases for machine learning. But in a business context, they all pretty much boil down to one thing. 

Predictive decision-making. 

That's it. That's the use case. So let's talk about what that means.

Predictive decision-making is about using the machine to line up all the data needed to make a decision, and then spitting out that decision. Very simple. When this is done really fast, like in microseconds, it can power automation to act on those decisions as well. 

The predictive nature of those decisions is based on both prior data (knowledge) and perpetual data collection (learning). 

In one of my actual work examples, if we have years of data telling us that, every Thursday in a particular geographic location, we end up needing six trucks on average, we can determine that next Thursday, we should schedule six trucks. 

Then we can also take into account data about growth, seasonality, weather, holidays, traffic, etc., and we can improve the accuracy of that decision. The result is we roll only five trucks on the Thursdays we need only five trucks, thereby saving money, and we roll seven trucks on the Thursdays we need seven, thereby capturing more revenue.

Boom. Benefit. If that benefit is worth more than creating and implementing that tech (and provided we have the data and the means to use it), we adopt A.I. for that use case. If our calculations are accurate, we can do all the scheduling automatically as well. 

That's just scratching the surface. Automation and ML together can do some crazy things, like make those trucks self-driving or spit out emails to our customers to generate more demand. So this is where you bring in ...

The 80/20 rule

I'm a huge proponent of the 80/20 rule for the automation (acting) side. What this means is, outside of the simplest tasks, automation should always be 80 percent machine and 20 percent human. 

Those percentages should shift on a scale of "what happens when you're wrong" being more expensive than the benefit derived from the automation. In other words, the more expensive the mistake, in terms of both money and time, the more human input you need.

Once the machine learning (thinking) is experienced enough and robust enough to make the decision to act and devise the instructions for how to act -- at close to 100 percent accuracy -- the task becomes repetitive enough to be able to 100 percent automate. 

So the reason why the question "Should I use A.I. in my business?" is so difficult to answer is because there is no single broad answer. That means it's the wrong question. You have to take into account all the decisions and tasks that make your business work, then ask the question "Should this be automated?" to get a 100 percent accurate answer.

In a future post, I'll run through some broader use cases and start answering the A.I. question based on where the tech is most accessible today. Now would be a good time to subscribe to my newsletter so you don't miss that or any other posts. See the signup link below.

BY JOE PROCOPIO, FOUNDER, TEACHINGSTARTUP.COM@JPROCO