One of the best parts of being Forrester’s CEO is that I have hundreds of analysts who help me understand new technology. So I thought I would share the wealth and pass on some of that knowledge to my fellow CEOs. I will be putting out a post about every quarter focusing on a technology that has three characteristics: 1) in the news; 2) frequently misunderstood; and 3) CEOs should understand it. First stop … artificial intelligence.

What Is AI?

It’s using computers to detect patterns and make predictions. Example: using a software program to identify which of your customers are not going to renew.

Today’s AI is decidedly unsexy — it’s not about killer robots, computers curing cancer, or runaway autonomous vehicles. If the AI application your company is building might show up in a Hollywood movie or amaze a layperson, it is likely doomed to failure and will be a titanic waste of time and money. True applications are quietly valuable, modest, and they won’t impress your mother.

How Does AI Work?

AI is just another computer program. Step one: Obtain a relevant, clean, and well-formatted data set. Step two: Feed the data to machine-learning algorithms to build and train a model (formulas). Step three: Use the model to predict and find patterns.

Confused? Here’s an example …

Like any good CEO, you don’t want to lose customers. So you go to your CIO and say, “Build an AI system that will predict which of our customers are going to churn.” Here’s what they will do:

Step one: They will gather up data on customers from last year and put that data in a consistent form. Think an Excel spreadsheet with the first row as “Customer 1” with columns showing information that will be relevant to whether they reupped with you or not — things like “tenure,” “engagement,” or “dollars spent.”

Step two: All of that data (imagine that it is 100,000 customers) is then fed into algorithms that statistically sort and identify relationships in the data. This step is called “machine learning” or “training.” The algorithms create a model of your customer base (or at least last year’s customer base) that can be used to predict whether a customer will churn or not churn. This is called an “AI model.”

Step three: You can now ask the model a simple question: “Given the behavior of our customers last year, what are the chances that a specific customer this year is going to stay with us?”

What do you get? The ability to lower your churn rate by identifying at-risk customers early and taking actions to retain them. And retention equals higher revenue.

This is an example of using AI for classification — putting customers into a high-risk category. But it can also be used to construct continuous ranges (enabling you to predict a customer’s lifetime value), clusters (identifying customer segments), anomaly detection (to pinpoint fraud), or association rules (if a customer bought product A, then they will probably buy product B).

What You Should Do

I really wanted to say “Do nothing,” but Forrester analysts reported to me that they have seen too many good AI applications to advocate no action. So the better answer is, “Get more AI in your company.” Why? It will enable you to subtly but powerfully optimize marketing, sales, operations, and customer support. If you don’t, be assured that your competitors will — and they will open up a gap that customers will begin to recognize. Here’s how to proceed:

  1. Go to your executive team and announce that you want your organization to not be on the bleeding edge of AI but that it should be good at AI. To do this, you want them to go on a three-year AI initiative sponsored by you, with a clearly stated vision and mission.
  2. Put out the word that you want the zombie AI projects in the organization (yes, they are out there) to be eliminated. As the company gets smarter in AI, the doomed projects will be more easily identified.
  3. It’s all about the data. As part of the AI initiative, assign your chief data officer (or CIO if you don’t have a CDO) to begin the multiyear project of standardizing data across divisions and functions. Start with improving customer data hygiene.
  4. Centralize governance of AI into a center of excellence that is focused on: 1) AI bias and transparency — making sure that your models are explainable and do not discriminate; 2) ModelOps — operationalizing models at scale and monitoring for performance decay; and 3) AI governance — ensuring that each business unit’s use of AI adheres to strict governance standards across the AI lifecycle.
  5. Train yourself and your executive team on AI and analytics. You don’t need to be able to code — but exposure to the technology will give you a better sense of where AI can be used to improve your business and, most importantly, how to make better decisions based on the output of AI. First Tech Federal Credit Union sent its executive team (including the CEO) to an executive-level class on AI and analytics at Northwestern — it’s improving the data literacy of the credit union’s culture and leading to the proliferation of game-changing analytics.

Vocabulary

Here’s enough terminology to make you slightly dangerous: DSMLAI — the overall name for the discipline — data science, machine learning, and artificial intelligence; AI-washing — projects that claim to be AI but aren’t; ML — machine learning; deep learning — also called neural networks — good at classifying (labeling) text, voice, images, and other unstructured data.

And if you’re a Forrester client, here are links to two great AI reports that will give you more: