WHERE SHOULD YOU INTEGRATE A.I INTO YOUR BUSINESS?

The decision to integrate any form of artificial intelligence into your existing business is not one to be made lightly. 

I don't need to explain why. There are countless dystopian science fiction stories serving as cautionary tales of A.I. in the wrong place (the Matrix series) at the wrong time (the Terminator series). 

And these days, we're even starting to see a few real-life science stories offering a non-fiction glimpse of improperly integrated A.I. making a mess of things. Seemingly on purpose. 

In all seriousness, I just wrote an article on whether your business should hop on the A.I. money train. In that article, I gave you a decision framework so you can decide for yourself if A.I. fits. 

Now let's talk about where it fits, using today's A.I. variant: GPTs. 

Should you use GPTs to create production code?

Oh, hell no. And I have a one-word argument: exploitation.

Hackers. 

Miscreants. Malfeasance. You only have to look at crypto to see how one little leak in a script can drain millions of dollars from thousands of people in a few seconds. 

And chicanery is just my top reason. I could come up with a hundred more, but that's a listicle and I don't do those. Anyway, what all those reasons boil down to is that all code, human-written or machine-written, fails at a breaking point. 

That breaking point is usually found at one of several outliers -- a series of "what-ifs" that the developer has to take into account when telling the code what to do and what not to do. Most of these outliers are unknown at the time of development, usually only found when "that one idiot does that one stupid thing for God knows what reason."

That's an actual coding term.

However ...

Should you use GPTs to assist your coding effort?

Absolutely.

I'm a former developer and a huge proponent of no-code and low-code platforms, which are essentially code chunks with a logical human interface to integrate low-level automation. 

Reducing that process down to its most basic form, using GPTs to assist in your coding effort is like the autonomous self-driving function for coders. You can take your hands off the wheel and let the machine drive for a while. 

But you want to limit how far you let the machine drive and in what kinds of conditions. The more serious the condition, the more you want human hands on the wheel. 

That's a choice that's up to you. Is a demo or a mockup serious? Probably not. Is anything that faces the customer serious? What about privacy and security? How about any transaction where money is exchanged? That's pretty serious. See the bit about crypto.

It's the same question I ask about no-code and low-code integration, but with less human input and more trusting the machine on the GPT side to think about where the loopholes are inevitably going to be created. 

Should you use GPTs to create assets?

So what about creating stuff that isn't code but is sold to the customer as part of the value proposition of your offering?

I say go nuts, but only if you expect those assets to be low value, and certainly not intellectual property.

That opinion might be a little controversial and might even get me dinged for being backward thinking in the machine overlord era. But I've been doing generative A.I. for almost 15 years, and I can't help but think that the more you lean on machines to create your value proposition, at some point you'll eventually get burned.

Here's why. 

Remember, all GPT output is based on the original work of other people. Yeah. Computers don't create, they borrow and blend. So you're counting on whoever programmed those GPTs to obfuscate the difference. Obfuscate is not a nice word in this instance. It means masking something untoward, like stealing someone else's original ideas for profit. 

Should you use GPTs for customer support? 

Yeah, probably. This seems like the most obvious way to get started.

You might irritate a couple of customers every once in a while, but the practice of automated support is already widely accepted, though barely tolerated, today. In my mind, we're entering an era where focus on the customer is going to be a bigger differentiator than it's ever been.

So if you want to be safe, plan on 80 percent of your support being automated and 20 percent being human, which in this case means always having a way to speak to a human. That costs money, but not as much money as 100 percent of your support being human or your chat and email support being really shitty.

Should you use GPTs for marketing?

Depends. You might already be dabbling in marketing A.I. if you're using a third-party marketing tool like HubSpot or Mailchimp or even Hootsuite.

The real question around GPTs for marketing is about content marketing. And my recommendation is to proceed with caution, and maybe not for reasons you might think, what with my being a content creator and all. 

Marketing is about getting people's attention and drawing them in. This is hard to do with generative A.I., as the primary focus of generative A.I. is to deliver information as efficiently as possible. 

And this is where I have a real-world example. 

The Associated Press uses our NLG platform from Automated Insights to write its quarterly earnings reports for public companies. How AP uses that 80/20 machine-to-human rule is that for smaller companies, it uses our platform and just generates an information-rich, zero-analysis recap. For big companies like Apple or Walmart, AP starts with our article and then a human adds analysis and quotes and such. Now AP can do thousands of reports every quarter, where it used to be lucky to be able to crank out 100. 

That's how I'd use GPTs for content marketing. Generate the facts and enhance the end product with your voice, the voice of your brand.

Should you use GPTs for sales?

Oh, hell yes (sorry, sales folk). But only after a certain point.

With every new company or new product, sales begin as more art than science. This is just another in a series of reasons why startups and innovation are so difficult to make successful in the short run. Scientists trying to paint. 

But once the sales flywheel starts ... flying, GPTs can automate not only the mechanics of the transaction, but even some of the personal touch and handholding required to get a prospect from first contact to close.

Should you use GPTs to create educational material?

I'd say no, but ... 

Maybe for simple education and maybe when you're not charging for education. 

If the education is for something serious, like medicine or aircraft maintenance, then I'd refer back to the cost of being wrong being larger than the benefit of the automation. 

If you're charging for your educational materials, then I'd refer back to a few sections ago in this article, when I talked about the nearly-nonexistent-until-lawsuit-time attribution of the original ideas that GPTs are aggregating.

Now, with that said, it's usually the case that the best and richest education is the most interactive education. GPTs are great at taking input and giving a response. In this sense, learning can be tailored to each individual in a way that's much more user-friendly than a class or a webinar.

Now, in this article, I'm only considering ChatGPT, one form of A.I. There are hundreds of these derivative use cases out there, but they almost all come down to predictive decision-making.

So when you're deciding which parts of your business you should integrate, your own personal (human) decision matrix should come down to benefit versus cost, using the 80/20 rule as your guide.

BY JOE PROCOPIO, FOUNDER, TEACHINGSTARTUP.COM@JPROCO