Why generative AI just hits different – and why organizations need to embrace it now

Generative AI isn’t just a new craze or a shiny toy; it’s set to bring about major macroeconomic effects — including adding $7 trillion to the global GDP and lifting productivity growth by 1.5%. In other words, it’s not going to fall out of popularity — it’s going to become an essential tool for companies across industries.

Sav Khetan, senior director of product strategy at Tealium, spoke at Transform 2023 about why gen AI it’s important, how it’s making a difference, and how business leaders should be considering it for their own organizations.

“The magic we’re all feeling and experiencing right now is because AI is suddenly able to communicate directly in our language, both in and out,” Khetan said. “It can understand what we say with full context and it can respond with language and images that we can understand. That’s what flipped.”

How did this happen, and why now?

AI has been integrated into our lives for a long time, but generative AI just hits different. Previously it was very specific and purpose-driven, but large language models (LLMs), the backbone of generative AI, have rewritten the script. The “4” in GPT-4 is a way of describing how much complexity and scale the model can handle. GPT-3 was 175 billion parameters, and the newly available GPT-4 is 170 trillion parameters.

“What this shows you is the scale at which these models are operating,” he said. “This is the reason why the 80-year effort suddenly clicked into place. These models were able to access the internet at large in the last couple of years. It turns out that they needed that much data to figure this out.”

They can consume both structured and unstructured data, which is where the game especially changed, since there’s more than 80 percent unstructured data in the world, primarily video.

The impact on UX and how we interact with technology is profound, he adds. Up until now, UX design has been about translating human commands into action, requiring structured, organized, tagged data and operational code. LLMs can consume data in its current form, without any translation necessary.

“If you play this out, if you look ahead, this has the opportunity to completely change how we interact with data directly,” he said. “What if there was no software in the middle? I realize that LLMs in some way are machines themselves, but our relationship with data that we consume for ourselves – for research, for analysis, for insights – now has the power to completely change.”

Where generative AI is headed

Back in February, ChatGPT became the fastest growing platform of all time. The pace of generative AI innovation and adoption isn’t slowing down any. Gartner predicts that by 2026, 50% of all sales and marketing providers will incorporate assistants, and 60% of design process by new websites will be by generative AI. At the front end, 30% of HR software will use the assistants. And by 2025, 75% percent of digital marketing communications will have avatars.

But those use cases are almost certainly going to evolve into even more powerful applications as stage two, or wave two, as venture capital firm Andreessen Horowitz (a16z) calls it. We’re still in what Khetan called a “pull world,” in which we’re asking AI for responses. Synthesis or synth AI is when the AI automatically looks at the data and tells us what it sees, and can be set up at any cadence we want.

“Now imagine the power of decisions you could make,” he said. “If you could genuinely consume the data that you want from the research that you want to do, for the analysis and insights that you want, your decisions just get that much better and that much more powerful. That’s why we’re so excited about this future. So yes, please take this seriously. Please do your work.”

Getting started with gen AI now

“Everybody is asking, what do we do to get started?” Khetan said. “It’s simple. Learn. Explore. Play with it. Spend time with it. Nobody can explain this to you. You have to go experience it for yourself.”

Right now all these tools, such as DALL-E, Midjourney, ChatGPT and so on are free, but there’s no guarantee how long that will last. Now’s the time to experiment.

You also need to start preparing your data, your APIs, and your systems so that they can connect with these tools, and access the data they need. And you cannot embrace AI in your workplace without as much first party data as possible, because it’s the main input for your AI workflows, if you’re going to make it customer facing.

“If you’re not already doing it, prioritize it,” he said. “If you’re not already planning it, start having those meetings. This is the time.”