What can you do with AI?

That is the question that so many companies and individuals have been asking over the past year. Companies are directing their employees to find every way possible to use AI. Annual goals and strategies are focused on AI implementations.

Don’t get me wrong, we have seen some really great examples of using AI to solve problems. The problem is that Ai isn’t the only solution. Maybe a better question should be:

What should you do with AI?

There are many things that we can do with AI, but that doesn’t mean that we should use AI. I have seen a lot things that people are doing with AI and my first thought is that we were doing that with a simple script 30 years ago.

I guess it sounds fancy to say AI is doing the job, but why us AI if simple automation can perform the task? The argument may be that prompting AI to do the task is easier and doesn’t require scripting skills like writing a script may take. It also may be much faster to create the prompt than write a script. In that case, maybe the Ai should be used to help create the script to then perform the task.

Why not use AI for the task?

Cost.

We have seen the cost of AI going up and I don’t see it coming down anytime soon. New models require more tokens, which costs more money. This cost is incurred every time the AI needs to run. Scripted automation, on the other hand, has an upfront cost to write the script, but is then free to run after that.

Let me give you a real life example of the difference between a script and AI. I have a pool and years ago we decided to get a heater to extend our swimming season. There are different options available:

• Gas/electric
• Solar

I broke these into 2 categories because the operating cost is different. However, they do have one thing in common, they all have an upfront purchase and installation cost. The difference is in operating cost.

For gas and electric, there is an additional cost that is incurred to run the heating element. This is similar to AI. You want to use it, you have to pay more for it every time. How much depends on your models, which is similar to the difference you may pay between gas and electric utility prices.

For solar, there is no additional cost to heat the pool. Like an automated script, you just paid for the initial install, after that it runs when the pool filter runs. Using tiny tubes in the panels on the roof, the water just flows through and the sun heats it on its way through. The pool runs 12 hours a day anyway, so there is no additional costs.

You could argue that AI may provide some added value, similar to how gas or electric heaters may provide more capability than the solar heating. The question then becomes is the value received worth the extra cost of the utility.

What’s Next?

I think that we will see a wave of new consultancies popping up that focus on helping companies right their ships after the all-in AI approach. These experts will be able to review a companies use of AI and be able to identify what should be converted from AI to regular automation or what models make sense for specific tasks.

The cost of AI is going to sneak up on a lot of places because they just went all-in with “AI-ify Everything and Anything” with no guard rails or direction. Employees will have looked for anything they can to show thier loyalty to the cause or move up on the leaderboard. Then comes the realization that maybe not everything needed to be AI’ed.

it is going to take a good understanding of AI technologies paired with knowledge of the old ways of working to get good results from this.

How is your company thinking about “could” vs. “should” when it comes to AI? Have you already hit the point of starting to right the ship, or are you still full steam ahead?