Google executive says supply chain uses of generative AI flourishing

In the real world where solutions are being implemented, Barkley said, various estimates of savings and investment in particular industries are “shockingly huge” and open up “this whole new world.”

AI and machine learning models at Google that just a few years ago used millions of “parameters” to produce information or a solution now use trillions of records, “just magnitudes bigger than they were before,” he said.

The power of context

The models now have the ability to produce “context,” meaning that even if a specific piece of data is not in the model or is not provided by the user, the model can figure it out, Barkley said. He gave the example of asking a GenAI tool to produce a picture of an animal that is furry and has a tail and that humans view as pets. Even if the user of the model does not input the word “cat,” he said, today’s AI tools can produce that picture.

GenAI models increasingly have all the cognitive and sensory capabilities of humans except smell, Barkley said (though he did not mention taste and that doesn’t appear to be in GenAI’s toolkit yet).

He reviewed both theoretical uses for AI in supply chain management and real-world examples that are already are in place.

Among them:

• A British client that used AI to determine if the wording of its contracts conflicted with Brexit regulations, an action that otherwise would have taken a massive number of worker hours.
• A similar initiative to read California’s various governmental regulations “and try to figure out which one of them apply.”
• A food company using AI to generate a standard procurement contract that could be used internationally as well as in the U.S., because it could read and process data and text from multiple countries.
• “Last year was a year of proof of concepts,” Barkley said. “Lots of people were experimenting and this year we’re seeing people actually spending money.”

Manhattan Associates makes its own GenAI announcement

Barkley’s address came after Manhattan Associates had announced at the meeting that it was adding a GenAI solution for customer service to its flagship Manhattan Active product, specifically an AI-driven chatbot, Manhattan Active Maven. 

“Maven stands out as the first GenAI-powered customer service chatbot with native access to

orders, payments, store locations, and product availability,” Manhattan Associates said Wednesday in announcing the new offering that was previewed at Momentum by Brian Kinsella, a senior vice president at Manhattan.

The company also announced another new product, Manhattan Assist, which it described as “a GenAI-powered assistant that provides contextual responses to any questions regarding product functionality.”

In his address, Barkley said bots currently used for customer service can only generate a response that “someone had to sit down and write an answer for every single thing. With gen AI models, that doesn’t happen anymore. You don’t have to predict every single question because AI will generate an answer. I think the revolution around how bots can be used is enormous.”

He put the advantages of AI in business operations in several categories: optimization, automation, streamlining and transforming.

While these may overlap in many instances, it was Barkley’s discussion of the term “transforming” that was the most striking. The other areas mostly appeared to deal with efficiency, but his description of transforming was the most ambitious. 

He gave as an example pharmaceutical development. A generative AI model can analyze how molecules might work together far more efficiently and quicker than “the millions and millions of dollars of specialist work to actually do that now.”

“And so you can take these chemical compounds and create new chemical compounds, brand new things that were never thought of before, and you can do it at the speed of light,” Barkley said.

Generative AI also opens the door to personalized marketing, he said. He showed a picture of himself with a Texas background generated by AI. Similar images could be generated to target individual customers. “So marketing is a huge place when it comes to content generation and how that works,” he said.

Barkley presented a slide with guidelines for determining what a company should do to get started with AI solutions. Some of the key points sounded like they might have been recommendations for any technology project at any time in recent history: Prioritize values, build a team with expertise and experiment on applications, collect data, integrate with existing systems and expand gradually.

Another recommendation: “Doing all this while being trustworthy at all times.” He said AI can “hallucinate” in its early output, producing some wild preliminary material, and companies should guard against the damage that can do to their reputations.

Going it alone is inadvisable, according to Barkely. “Finding a partner who can work with you and to address challenges and align to your goals is critical,” he said. “You have to have those partners. We are all learning and are all working at blazingly fast speed.” He cited conference host Manhattan Associates as an integrator implementing AI solutions.

Whatever the cost is, Barkley said, it has come down sharply. In the past year, he said there were times when costs to implement an AI solution had dropped by a factor of 10 and then by 15. Now, he said, projects are 40 times cheaper than they were a year ago.

More articles by John Kingston

3 supply chain software providers tell their latest stories at NRF

‘Quiet quit’ product of disengaged supply chain workforce

E2open stock pops higher on announcement it is launching ‘strategic review’