Nearshore Americas

Q&A: Questioning the Economic Basis for G-AI Use Cases

There’s a lot of noise surrounding generative AI, and few are the voices out there attempting to bring nuance to the many conversations about the technology’s role in human activity. 

One of those few voices belongs to Joe Procopio, a self-described Natural Language Generation (NLG) pioneer. Serving currently as Chief Product Officer at on-demand car care company Spiffy, Procopio built part of his fame and expertise as co-developer of Automated Insights, which holds the claim of being the first commercially available NLG platform.

Procopio is aware of the potential for AI in business, but he’s also painfully familiar with the mistakes a company can make when trying to sell AI as a product itself. 

In our chat with Procopio, he expresses (without mincing words) his skepticism towards the hype surrounding ChatGPT, retraces his journey with Automated Insights and leverages his experience with AI to throw some advice at business leaders who might find themselves confused or/and overly excited about this shiny, new toy being publicized everywhere.  

NSAM: There’s a lot of hype behind ChatGPT and generative AI in general. What’s your take on all of this hype? Will these tools live up to it?

Joe Procopio: There are a lot of reasons why the current hype around generative AI is overblown. Don’t get me wrong, the recent advancements in generative AI are impressive, game-changing, almost magical. But those advancements are only advancements in a specific niche of AI, and I think they’re just a natural evolution that has finally hit the mainstream.

Joe Procopio, CPO at Spiffy

When people call this evolution the dawn of true artificial intelligence, that’s where I start to wince.

I’ve been diving deep into NLG since 2010, so I’ve got unique insight into how far generative AI and NLG have come over the last decade and, more importantly, what the mainstream implications are for the current wave of AI advancements. At the end of the day, these are basically just advancements in GPTs, which are basically just automated content producers across text, audio, video, code, etc.

You can still make a boatload of money with AI, but it’s not snap-your-fingers money, it’s more try-and-fail-a-dozen-times money. There will be a lot of attempts at making the easy money, and my concern is it goes off the rails like it did with NFTs and becomes rife with fraud and scams.

NSAM: In an article for Built In, you wrote that products such as Bard and ChatGPT should be used as tools, not as tech to build products on. What did you mean by that? 

Joe Procopio: Well, the example is already out there. OpenAI already has the product, much like Zoom had the video chat product. And like the case with Zoom, Google and Microsoft are already rushing onto that bandwagon with their billions of dollars. The end of the pandemic lockdowns were the end of the gold rush for video chat. There will be an end of the gold rush at the end of this generative AI hype cycle. 

Is that a game of musical chairs that a startup wants to play without the billions-of-dollars-deep pocketbooks of Google, Microsoft and now OpenAI after their latest fundraise? I don’t think so. The generative AI train has left the station. 

There will be a lot of attempts at making the easy money [with AI], and my concern is it goes off the rails like it did with NFTs and becomes rife with fraud and scams.

That said, there are plenty of use-cases for GPT-based content creation that could be built upon in a novel way by an innovative entrepreneur. But it’s not the same thing. You’re not building an AI product in that case, you’re building your product on someone else’s AI product. 

NSAM: With Automated Insights, you successfully turned an NLG-based tool into a product that the market could use. Please, describe the process that led from the technology itself to a developed product.

Joe Procopio: In that same article you mentioned earlier I kind of laid it out and I’ll summarize it here. 

Back in 2010, we built a platform and an engine to create human-sounding content out of large sets of data, mostly sports data at that time.

We needed to show off our tech. So we spun up over 800 websites, one for every pro and college football, basketball and baseball team in the US. Then we populated those websites with automated content, up to five times a day. People thought it was amazing, but every pro team and almost every college team had at least one human covering that team, and they could write about stuff we could never catch.

Great tech. Lousy use case. 

But then we started noticing that smaller college conferences started promoting our content on social media, because no one was writing about these teams, at these smaller schools. We realized we were unearthing a new audience and fulfilling an unmet need, creating content where humans could not.

Same tech. Optimum use case.

[Generative AI] is a tech that can seemingly do anything. But that doesn’t mean those use cases are all economically viable or even materially useful. 

This led to our project for Yahoo Fantasy Football, producing fantasy matchup recap content where it couldn’t be produced by a human; then we moved on to large sets of financial data and signed the Associated Press, then marketing data, traffic data, population data. Our pitch was the ability to surface important insights in seconds that would take a human hours or even days or weeks to produce, if a human was even working with that data.

NSAM: You seem to envision the current business landscape of generative AI as one filled with pitfalls. Which are the biggest, most dangerous pitfalls companies can fall into when trying to adapt this technology to their businesses?

Joe Procopio: I don’t think there are a ton of individual pitfalls, necessarily. What I do believe is that a lot of companies are going to adopt generative AI without a plan for ROI. 

Again, it feels like the same bandwagon everyone hopped on with crypto and NFTs. Even something as mainstream as mobile apps; there were so many companies that created a mobile app that didn’t need to exist. 

The potential list of use cases for generative AI is infinite. It’s a tech that can seemingly do anything. But that doesn’t mean those use cases are all economically viable or even materially useful. 

NSAM: There seems to be a lot of hope for generative AI among business leadership, which can lead to misconceptions of the technology’s actual potential. With that in mind, who should oversee the implementation of this technology into the company’s business? 

Joe Procopio: At this point, I think you should look to the CPO to lead implementation, although I don’t feel super-strongly about that. 

I think the business case and the ROI are the most important factors in terms of adoption. If your CPO can’t figure out how to create viable revenue streams out of automated content, they’re probably the ones to tell you whether or not you should build around it and what that implementation should look like. 

NSAM: What’s your top recommendation for business leaders who are seeking to incorporate NLG/GAI into their products and operations?

Joe Procopio: Ask whether or not your customers are screaming for whatever solution you want to implement that only NLG/GAI technology can enable. 

Sign up for our Nearshore Americas newsletter:

I honestly don’t hear anyone saying “Wow, I wish I had MORE content.” If there isn’t a marked improvement in the customer experience, one that customers would pay more for, then you might be just playing with a cool new toy. That can be fun, just don’t bet the company’s future on it.

Cesar Cantu

Cesar is the Managing Editor of Nearshore Americas. He's a journalist based in Mexico City, with experience covering foreign trade policy, agribusiness and the food industry in Mexico and Latin America.

Add comment