AI skills aren’t the only ones that enterprises are demanding.

A survey published Tuesday of more than 750 senior technology leaders by Coursera and AWS found that three skills in particular – cloud, data and cybersecurity – ranked above AI skills in critical importance.

Some 95% of participants reported that cloud transformation is a key business goal over the next three years, while only 91% said the same for generative AI.

AI is a “tool in your arsenal,” but having skills in the other domains accelerates the way enterprises can leverage the trendy new tech, said Mustafa Furniturewala, CTO of Coursera. “Having that domain expertise, having that skill set, actually augments the AI skill set really well.”

Cloud, data and cybersecurity rank high for a few reasons, said Furniturewala:

• For one, the cloud data transformation isn’t yet complete, he noted. “We’re far from being in a mature state with that.”
• AI also accelerates the use of cloud technology, he noted. “A lot more people are able to build now. They’re able to use cloud and AI and data in ways that are a lot more accelerated.”
• Plus, while AI is taking up a good deal of enterprises’ brain space, other trends exist, said Furniturewala. “Security attacks have gotten more sophisticated, and so the need for cybersecurity has gotten more important, irrespective of AI.”

While all of the skills go hand-in-hand in the broader tech transformation, he said, enterprises still need to upskill their workforces to cope with rapid change.

Leaders are dealing with two transformations, he said: One involves the technology itself, and the other involves the workforce that’s leveraging it. About 88% of leaders surveyed agreed that AI investments wouldn’t succeed without training, and 77% said training is essential in realizing their tech goals over the next 12 to 18 months.

While 72% said they expect new hires to understand how to leverage AI, 74% recognized that new hires alone couldn’t fill all of the AI skill gaps. “Just doing the technology transformation without doing the upskilling transformation is not going to result in benefits to the organization,” said Furniturewala.

So how can enterprises keep up? One way is providing workers the tools to experiment with the tech hands-on so that they “learn by doing,” said Furniturewala. Enterprises can implement the tech directly into relevant work, do practical skill assessments and provide risk-free sandboxes that allow employees to feel safe experimenting – and making mistakes.

“Learning in the flow is extremely important,” he said. “The more you integrate AI in your actual workflows, the more you learn.”

Assuaging fears of AI replacement may also make workforces feel more comfortable strengthening their tech skills, he said. “Automation is augmentation,” he said. “Despite increasing levels of automation, leaders recognize that human contributions remain irreplaceable.”

When deadlines close in, security often slips through the cracks.

A recent survey of 250 IT leaders across industries by application security firm Cypress Data Defense found that 62% of organizations are knowingly releasing insecure code to meet delivery deadlines. While enterprises neglect tech safety to meet financial goals, security teams are finding themselves burned out and underresourced, said Aaron Cure, co-founder and director of cybersecurity at Cypress.

“They’re pushing code that they know has vulnerabilities,” Cure said. “They’re releasing because they have a deadline, because they have money goals to reach and they have sales goals to reach. So they prioritize those goals over security.”

The report found that only 36% of enterprises involve security in the planning phase of development:

• Often, the security focus of a business is on network security, or the protection of physical infrastructure like servers and routers, rather than application-layer security, or protection of software – despite the fact that application-layer attacks drive 43% of security breaches, according to the report.
• That’s because network security is “tangible,” said Cure. “It’s something that takes a lot of trust. As soon as I put that money into training (staff), that money walks out the door when they go to the next job. If I buy a piece of hardware, that hardware stays here.”

AI is only exacerbating the problem, Cure said. Amid the growing trend of rapid code development with the help of AI tools, security teams are forced to identify bugs in larger and larger code bases.

“Because AI is generating twice as much for them, they’re getting twice as much done. So now, you’ve got twice as much to do,” said Cure.

This leads to security teams spending all of their time “fighting the same fight,” said Cure. While they attempt to keep up with security vulnerabilities and weed through thousands of false positives picked up by code security scanners, they’re often fighting against leadership’s demand for output, said Cure.

But in the face of competition, slowing down development isn’t always possible, said Cure. The only solution is investing the time, resources and cash to bolster application-layer security, whether that means hiring more talent internally or outsourcing.

“What they need to do is to be able to iterate security as quickly as they can iterate features,” he said.

What happens when agents go awry?

The results can range from frustrating to disastrous, and IBM is developing a potential remedy: a system for “coordinating a conversational agent” with a large language model for “conversation repair.” When an agent fails to understand a user query, IBM’s tech would essentially refer back to a more sophisticated AI model to set it on the right track.

When a failure is detected, IBM’s system pulls a descriptive prompt that explains the chatbot’s purpose and capabilities. That information, plus the context of the failed conversation, are sent to a more capable large language model, which determines the user’s intent as well as where the original agent went wrong.

If possible, the language model hands the task back to the agent, this time with a clarified user intent. If the task is outside of the agent’s capabilities, the language model handles it.

“Traditional (conversational agents) are limited to a specific set of topics and are not able to respond to input outside those topics,” IBM said in the filing. “They can only produce responses that are pre-written or pre-approved by humans, i.e., they are non-generative.”

While IBM’s tech uses AI to fix AI, generative models don’t always get things right. The domain knowledge of large language models tends to be far more expansive than that of a simple question-answer chatbot. But when conversation expands outside of the realm of its training data, these models can hallucinate, potentially leading to broader issues than a closed-domain chatbot running out of things to say.

Plus, while AI agents are the current hot topic of the tech world, as far as customer service goes, many consumers prefer the real thing: A Gartner study found that 64% of customers reported they’d prefer that companies not use AI for customer service. About 53% reported that they’d consider switching to a competitor because of it.

The bottom line: Before using AI to make chatbots better, remember the tech is far from a panacea. Enterprises should consider where it’s best to keep a human in the loop.