The ultra-fast pace at which new artificial intelligence (AI) innovations are emerging and being used by traditional and digitally native organizations  is enough to make your head spin.

But there's good reason its adoption is growing by the day.

Rackspace Technology's 2023 AI and Machine Learning Research Report found "AI and machine learning is expanding as a key driver of competitive advantage" for companies.

The study found manufacturers, hospitality companies, government agencies, and healthcare providers — among organizations in many other sectors —are industries increasingly turning to AI innovations to solve business challenges old and new.

Despite 69% of surveyed IT pros deeming AI a high priority for their companies, though, only 41% said their businesses have seen notable benefits and ROI from AI tech implemented.

Acknowledging AI's growing impact on organizational success for new startups and established enterprises is one thing. Knowing where and how to use it effectively is another.

That topic was one a panel of AI experts discussed in detail at Alloy 2023.

‍Understanding the new "automation industrial revolution": How AI innovations are impacting companies old and new

"History can be really informative," SK Ventures Managing Partner Paul Kedrosky said at Alloy.

"In many ways, we're in the throes of a kind of automation industrial revolution analogous to ones we've gone through historically," Paul added. "But it's very different in very important ways many people gloss over in their Substacks."

Whereas previous automation "waves" were very capital-intensive, Paul noted the current one is the least capital-intensive from the buyer's standpoint.

"We should expect similarities [to past automation waves] in terms of capital replacing labor, which is sort of the economist's way of thinking about automation: that the labor becomes expensive and the capital becomes cheaper," Paul added.

But since this latest wave is so "structurally different" from prior ones, Paul expects to see marked impacts businesses and consumers — some positive, some negative, but all transformational in ways we haven't seen before.

From a legal standpoint, Cooley Partner Keith Berets detailed how recent lawsuits filed by and against large-scale organizations and ongoing litigation tied to those cases isn't really anything new, as they're mostly about intellectual property and scraping tied to AI innovations used by the companies.

Put another way? "There's not a novel AI lawsuit out there that doesn't tie into the traditional legal framework for bringing claims," Keith relayed, referencing suits against some of today's well-known businesses.

That said, Keith indicated it's on general counsel to stay up to speed on these types of legal battles to ensure they protect their businesses from getting sued or facing regulatory issues around perceived misuse of artificial intelligence solutions.

A prime example of something chief legal officers and GC should monitor closely: recruiting and hiring tactics that rely on emerging AI innovations.

"A lot of AI models are being used for hiring decisions," Paul noted at Alloy. "And, in the employment context, you use AI and it's biased, you've got a bias problem. It doesn't matter that it was AI versus a human making those decisions. You're going to be held accountable. You can extrapolate that to almost [any business case]."

‍Turning the right AI innovations into an accelerant for various business use cases

A big point Biogen Head of Technology & AI Dave Clifford drove home with the Alloy audience was the importance of identifying "hallucinations" often presented by foundational AI models and how those results can be (and often are) much different from large language model (LLM) results.

"Foundational models are typically models that are trained on large amounts of data ... that has been organized by the Web," said Dave. "The challenge is, when you're in an industry that has a regulatory context like I work in, the concept of trying to validate systems that hallucinate is very difficult."

An apt comparison Dave brought up was the use of servers for storing critical company data sets. When we think about systems like AWS that have a near-100% uptime, Dave noted, we don't have any qualms about using them.

But when we use 'safety-critical' applications, hallucinations can be very bad for business — and even dangerous to consumers.

"The standard for most machine learning algorithms ... you deploy in an industry context is, 'Have you validated that algorithm?'" said Dave. "There are real challenges with trying to figure out domain-specific ways to validate foundational models."

Dave added It's "difficult to think about integrating some of these emergent systems at regulated companies."

Using "thin wrappers" of AI innovations, like infusing ChatGPT's model into one's products to create a new AI-backed offering (think 'intelligent' tools that can write better website copy or streamline a back-office task) is something many companies should be weary about, according to Paul.

"We're in the mainframe age of AI, where we're relying on these monolithic, centralized large language models which is kind of analogous to the 1950s world, and things are going to move more and more to the edge," Paul explained.

Executives at Fortune 500 companies and CEOs of newly created startups share two problems, according to Dave:

• 1) Determining how to get their foot in the door with utilization of AI innovations
• 2) Ensuring their organizations won't run into any regulatory or compliance issues

While both are valid concerns, Dave said what these C-suites and founders should really be focused on initially is where data required to take advantage of AI tools resides within their businesses — an issue that's more common than you might think.

"Most companies have a reasonable idea of where their crown jewels [in terms of data] are," said Dave. "And after they know where [they] are, they tend to deprioritize investment in where everything else is. That's because they found the [data] for the business challenges they're facing today."

However, Dave noted it's not about data quantity. It's about locating the most relevant data that companies need to tackle the distinct problems they want and need to solve.

Artificial intelligence adoption at banking and telecommunications companies is working out well because their data is inherently highly structured data that is core to the business and easy to leverage in real time, Dave relayed.

Businesses like automakers and those working in molecular discovery, on the other hand, don't have that luxury.

"Trying to make sure that your technologically advanced company also has a centralized focus on being digitally native is I think the first step in [leveraging] AI at scale," said Dave. And that means making data management a top priority.