To use AI or not to use AI? That is the question

According to both pundits and many actual builders of AI, this is a dangerous technology that’s going to destroy jobs. 

Here’s a simple example of how technology supposedly obliterates jobs. Let’s think back to a time when we didn't have calculators or computers (Still well within living memory). As you can imagine, there were army of people in accounting firms and departments armed with pencils doing all the calculations on paper.

You would think that by introducing more modern technology, the number of accountants would be reduced to minuscule numbers. There would be accountants throwing this new technology from high-rise office windows onto the heads of innocent pedestrians on the sidewalk to protect their jobs. 

And yet, we’re still waiting for it. A quick search for accounting jobs on Indeed.com returned over 1,800 open jobs within 25 km from Vancouver. Where is the AI when you need it!

Yes, where is the AI?

When a business-minded person asks the question of what AI is and how it can help with their business, the answers can become a little bit nebulous. It’s not obvious which ones will really be practical.

Beware of jargon. One example I came across: enquire.ai (yes, they used the .ai domain to make sure that you know they mean business). What is it? 

'The Enquire AI Difference - Our patented artificial intelligence technology combines contextual and lexical NLP AI matching, 3D-skills scoring, and performance ratings to deliver the most relevant experts to your query..'.

This goes nowhere fast.

Let's forget the term AI for a moment and talk about the actual technology behind the scenes. Let me illustrate how simple methods can actually deliver valuable results, as they have done in the past, before talk about AI started blowing up.

You’ll see that the main issue of deploying these algorithms is not in the algorithms, but in the data.

One of the most basic models you can implement is Logistic Regression. You use it to discover which outcome is the most probable. Will something happen, or not? Practical use? Yes/No answers. Will customers buy your product? Will customers switch to your competition?

Another one is Decision-making Tree. Through a series of questions, you can identify which product is the right fit for your customer (like buying the right computer) or if you are going to lend that person money (if you are a bank).

These are two very basic examples and if you do a Google search for 'Examples of machine learning algorithms' you get pages and pages to read through. The point of this? You don't need to search for the latest technology. You have to first identify what exactly you want to do with it. That’s what the question is.

But then you run into a much bigger problem. This is the problem which none of the companies selling you the dream of AI ever talks about. The issue is the quality of your data. You can have the most sophisticated algorithms at your disposal. You can have the fastest computers. If your data is not good, nothing good will come out of it.

Let’s take the example of 'Will customers switch to your competition?' Imagine that every year some of your customers leave and every year. You are asking 'Why?'

Do you know which data (or combination of data) can predict that? Is this happening because of pricing or features that don’t quite match what the customers are looking for? Is it because of bad customer service? 

I can give you an example from my past when our company was providing service to school districts. At one point one of the school districts decided to migrate from one email system to another. Why? The school district hired a new superintendent who came from another school district and liked the other email system there more. Tell me which data point in your CRM system can predict this. And if you think that this was an outlier, I have many more ... sadly.

The noise around AI is deafening and it is drowning the practical questions about the actual technology, its practical use, how to derive the value and all the pitfalls when trying to implement even the basic algorithms. The recurrent pattern? Sound business questions and smart people will always be in demand. Just ask the accountants.