(TNND) — Goldman Sachs’ chief information officer offered five predictions about how artificial intelligence could evolve and interact with businesses and society this year, from AI “workers” assigned to hybrid teams of human employees to robotic breakthroughs powered by AI.

It’s been over two years now since OpenAI's ChatGPT ushered in a breakout period for generative AI.

And Goldman Sachs Chief Information Officer Marco Argenti says in the new article that the potential for generative AI is coming into focus.

AI systems are becoming more like humans that could increasingly be employed like humans, he said.

Humans and AI could work together as members of hybrid teams at companies, he said.

Molly Kinder, a fellow at Brookings Metro and an expert on how innovations can impact the labor market, said she agrees that “AI agents” will become more capable and sophisticated over the next few years, and it’s likely that workplaces will deploy AI alongside humans not simply as supportive co-pilots or tools, but rather with more agency to act on their own.

“As a result, roles over the medium term will evolve to entail humans managing AI agents,” Kinder said via email Wednesday. “What this looks like – the scale, the scope, the nature – is still abstract, which is why 2025 is a key year as workplaces pilot AI agents to start to grapple with what a workplace with AI agents might look like. Coding and software development is an early mover industry to watch these trends.”

Last year, Kinder told The National News Desk that AI will disrupt jobs. It will change roles. And sometimes positions will be eliminated.

But she said AI, specifically generative AI, wasn’t yet an imminent threat to most jobs.

Does she feel the same way now?

“I have the same assessment for 2025. However, as expected, the capabilities of AI have already evolved and advanced, even over the last year, so the scope of what jobs are at risk of elimination has widened – albeit still a relatively modest percent of all jobs,” she said.

Argenti also predicted that companies will increasingly integrate AI with their proprietary data.

One way is through something called retrieval-augmented generation, or RAG, which is an architecture that can connect large language models to external, specialized datasets.

AI expert Anton Dahbura said he agrees with Argenti’s prediction there.

“I'm already working on a couple of projects with companies on these,” said Dahbura, the co-director of the Johns Hopkins Institute for Assured Autonomy. “There's a tremendous hunger for companies to incorporate their proprietary data, their internal data, for special purposes.”

Companies want to use generative AI for more than crafting emails, he said.

RELATED STORY: Artificial intelligence engineers compute their way to the top of LinkedIn's job list

Dahbura said 2025 should see ongoing improvements to the AI models, though there could be legal challenges with the data used to train the models.

“But so far, the large companies have been able to get away with just harvesting data wherever they can grab it,” he said.

Content providers might want to cash in, too.

The Goldman Sachs chief information officer predicted robotic breakthroughs powered by AI that could allow the systems to learn from the physical world in a way that can improve their reasoning capabilities.

He predicted consolidation in the number of companies developing large AI engines.

And he said responsible AI will become an even bigger priority for CEOs and boards of major companies.

Dahbura said he’s been “underwhelmed” by corporate efforts to rein in AI so far.

Vague policies, however well-intentioned, won’t be enough, he said.

“The devil's in the details,” Dahbura said. “And it's very, very difficult to anticipate all of the instances that can arise when you're trying to implement AI. And there are many, many pitfalls.”

Self-regulation also only goes so far, he added.

Kinder said the U.S. is in a preregulatory moment for AI, with little reason to anticipate federal regulation in the near term.

And she said there is not yet any good data on how companies are managing AI implementation.

“There has been little conversation to date about what ‘good’ employer and corporate practice should look like, especially on questions around responsibly deploying this technology vis-à-vis employees, and few industry leaders showing positive use cases,” she said. “My hope is that this changes in 2025.”