AI Can Improve Your Business

Want happier customers, happier employers, less staff turnover, and greater productivity? Read on.

You know what Blockbuster, Sears, and Border's Books all had in common? They ignored the dotcom boom until it was too late and went bust. You may recall in the 90s when many people scoffed at the internet's potential. A generation later, it's indispensable.

This is happening with AI and you don't have to be the next Blockbuster.

There's a great study from MIT (pdf) where they investigated a call center which had a subset of workers adopt generative AI. These workers, in two months, were as productive as workers who had been there for five or six months. Customer and worker satisfaction was much higher, and worker retention substantially increased. AI took the grueling nature of their work and made it much easier, but how can you make it work for you?

I like to think of AI adoption strategies in three approaches: disruptive, strategic, and tactical.

Disruptive AI

OpenAI's Sora is disruptive. Immediately after it was first shown to the world, a Hollywood producer halted an $800 million studio expansion. That's disruption.

Magic, Lovable, Cognition, and many other companies are building AI systems which can autonomously write your software, instead of helping you write it (though they can do that, too). That's disruption.

You can probably only go disruptive if you have solid expertise in this area.

Strategic AI

Strategic approaches are where AI is deeply embedded in systems in a direct, customer-facing way. This has led to Google laying off hundreds of employees as their AI-powered marketing systems replace them. Whether that counts as a success or not may depend on which side of the desk you sit on. And while many companies are successfully integrating AI, others, such as McDonald's, are having hilarious struggles.

The disruptive and strategic approaches are high risk/high reward strategies and require much greater sophistication in your AI knowledge. So you probably want a tactical approach.

Tactical AI

Note: the following is dealing primarily with generative AI, not machine learning.

I view the tactical approach as low risk/medium reward. More importantly, it seems mandatory. The call center mentioned above is not an outlier; many companies are experiencing this.

Here's one way you can approach it.

Go around to your various departments see what kind of repetitive work they do on a regular basis. You know, the soul-destroying work of "click, click, click, undo, click, copy, paste, click", and so on. Those are the things you can consider using new generative AI technologies for.

Pick a department and roll up your sleeves. You'll need three things:

1. Employees who know their job.
2. Learning prompt engineering.
3. Not taking the AI output for granted.

Dave Birss, in one of his LinkedIn Learning courses, shows a great prompt that actually delivers interesting advertising campaigns, along with the necessary resources. You probably want to pick tasks where perfection is not necessary (programming is a topic I'll cover in another post). You can also get AI to sketch out decent project plans in minutes, and you spend a day or two refining it instead of a week or more to get to the first draft. In fact, when I've done this, I've spent longer writing a good prompt than it took ChatGPT to generate the output.

Currently, AI is like Bob. Bob is the name of Yuki's assistant (this story is true, but names changed to protect identities). Yuki's a close friend of mine who's also a high-powered executive at her company. She was traveling to Paris by train for a meeting, but arriving late the night before. She asked her assistant, Bob, to book her a room at a hotel close to the train station. He did.

When Yuki arrived at the hotel, she discovered, to her horror, that it was where sex workers rented rooms by the hour for their client work.

Bob did exactly what Yuki told him to do, and nothing more. Just like AI.

When you first learn prompt engineering, you need to be very specific or you will get bad results. This has led many people to falsely conclude that AI provides no value.

I won't do a deep dive in prompt engineering in this article, but note that because you have to be incredibly specific, it means you have to already know your job.

AI isn't replacing your skill, it's augmenting your skill.

Initially, AI works best where there's some flexibility in output. A great prompt can give you a decent draft for an article, and then you use your expertise to edit it, adding in your personal touches, correcting awkward wording, and so on (note: I didn't do that for this article).

AI is also fantastic for fast research (note: I did that for this article). Yes, AI hallucinates, so you need to double-check its work, but there are also prompt engineering techniques for reducing hallucinations. Oh, did I mention you need to learn prompt engineering? Yes, you need to learn prompt engineering.

AI isn't coming. It's already here.

It's making a big impact on businesses. Machine learning has been heavily used for the past decade, but generative AI gives you many more opportunities and doesn't require deep expertise. Unless you already have strong experience in this area, start with a tactical approach and be amazed at what you can accomplish.

And learn prompt engineering.

There is obviously far more I can say on this topic, so let me know if there's anything in particular you'd like me to cover.