Will our job be replaced by AI?

I recently debated this with a friend who claimed that in the long run, AI will replace most of our jobs. 

My take on it is different and more mainstream, but since we were debating AI, I decided to use an AI-related argument.

Having developed many AI products over the years, I saw the Pareto Principle (AKA 80-20 rule) in action. Specifically, I saw that in most cases, it was relatively easy to get to 80% accuracy or coverage, but covering the other 20% felt like an endless journey. The effort needed to crack the last 20% is infinitely larger than the effort needed to get the first 20%.

Accuracy or coverage here is just an example. You can replace it with simply how well the thing you are measuring works (I know it’s not a good measure, but it’s good enough for the point I’m trying to make here).

Let’s say that AI broke into our lives two years ago with a score of 60. It still keeps improving radically and covers new abilities at an amazing pace, but that’s still nowhere near a perfect score.

There are already specific things that AI does much better than we do. For example, identifying that two photos show the same person. There goes the “bring your childhood photo” office game. For example, summarizing long conversations or articles and even finding early indicators for various specific diseases. 

But the levels required to help a person do their job versus replace that person at the job are very different. Think about cruise control systems: they have existed for many years already and do a decent job. But what would it take to give up on the ability to disable it immediately by pressing the brake pedal? It would need to be flawless and nothing less.

So, my take is that AI will keep helping humans in many ways, and even ones we never thought of for many years to come. Some simple jobs might be replaced, but most jobs wouldn’t since the bar is too high and requires far more than the functionality itself. 

AI at 80% and AI at 100% are very different things.

Now, let’s say that you are a customer of ‘AI.’ And let’s say your goal is to entirely replace something with it. You would need it to be 100% good. But if your goal is merely to help people do their job better, the bar is much lower.

Now, replace ‘AI’ with your product. What your customers try to achieve with it will impact at which point your product will be valuable and how much value there is to get.

We often tend to think of value linearly: the better we are at what we do, the more value it brings. But not all products behave that way. Sometimes, you have to be perfect to bring the minimal value, and sometimes, most of the value comes at the beginning. Sometimes, you have a cap, and in other cases, you can keep bringing value almost infinitely.

Not all products behave the same, and not all values behave the same. Which one is yours? Here is a way to figure it out.

Choose Your X-Axis

To identify how your value behaves, we will draw an imaginary graph. The Y-axis represents how much value your customers get. The X-axis can represent multiple dimensions. Different X-axis definitions will lead to different questions this analysis can answer.

Broadly speaking, there are three possible X-axis dimension categories:

The first category is quality – a generic name for how good your product is at what it was meant to do. This category will help you think about a proper MVP or decide how much more to invest in certain activities.

The second category is quantity – generally talking about how much value your customers will get the more they use your product. More can mean more users, more frequently, in more categories, etc. 

The third category is time – how much more value your customers get when they use your product over time. 

The second and third categories will help you answer questions about your business model, onboarding challenges, lifetime value, and more.

You can run the analysis multiple times with different dimensions at the X-axis to get a holistic view of your product’s value curve.

Draw the Line

Now that we have our axes, we can draw our line. We are looking to understand the general behavior, not exact numbers.

To simplify, choose one of three options for your value (the names are here for clarity, ignore them if they confuse you):

Linear – a straight line rising steadily. This is almost never the case if you had chosen quality as your X-axis, but for quantity and time, it could be. For example, if your product provides coupons that save money on grocery shopping, the more they use it, the more value they get.

Exponential – growing slowly for the most part and then at some point rising very high. Examples for this case include my AI replacing people question, when the X-axis is quality, or products that need lots of training on the customer’s specific data until it provides insights when the X-axis is time.

Logarithmic – rising high quickly and then growing slowly. This can be the case if your X-axis is time and your product provides visibility into something your customers didn’t have visibility into, like many assessment tools.

There are many more examples, of course.

There can also be other forms for your line – that’s fine, just remember that we are looking for the general pattern that will give you insights, not a deep analysis. It’s a tool for discussion and thought.

Look Far Into the Right

Now, let’s take your graph further to the right. What happens as you go?

Does your graph have a cap, or will your value continue to grow forever?

For example, if your product helps with cost reduction for certain logistic operations, there is a limit to how much more you can optimize. At some point, it will stop, no matter how hard you work.

Does your value diminish over time? If you teach your customers to do something or create a habit, once they know how to do it themselves, they no longer need you, or at least not as much as they did initially.

Don’t be afraid if these graphs give you bad news. You are far better off knowing and acknowledging them than ignoring them altogether.

So, what does your value curve look like? And what are your main takeaways from this exercise? Help me draw my value curve 😉