Pressure to implement AI raises IT tensions

Enterprise spending on generative AI services, software and infrastructure will explode over the next few years, jumping from $16 billion in 2023 to $143 billion in 2027, according to research firm IDC. But there’s trepidation on the part of IT teams tasked with deploying AI in the enterprise. The implications of developing, implementing and utilizing AI technology can be immense for networks, infrastructure, and software development, say industry players.

A study released by Juniper Networks, for example, found that 87% of the 1,000 global executives surveyed feel rushed to implement AI technology, and 74% feel that their corporate policies are unable to maintain pace with the potential risks and rewards of AI. In addition, 82% of the executives said they feel pressure to rapidly implement AI across a wide range of applications.

“When you consider how fast solutions are evolving and what they are capable of, it’s understandable why the push for rapid onboarding of AI is creating a tension point in many enterprises. It’s also understandable why policies for such powerful technology are often a sticking point,” wrote Sharon Mandell, senior vice president and CIO with Juniper’s global information technology team, in a blog about the study, which was done in conjunction with Wakefield Research and released this week.

While the urgency is palpable, it’s important to find ways to proceed cautiously so you don’t risk being left behind, Mandell added. “Keep in mind, however, that you don’t have to completely reinvent the wheel when it comes to AI and company policies,” Mandell wrote. “For example, most companies already have clear policies on what data employees can or can’t share with third parties. In many cases, it may be possible to simply restate policies in clear terms noting that they also apply to external generative AI tools.”

Remember to also consider software purchase policies and add addendums for additional reviews for any solutions with embedded AI, Mandell stated.

Enterprise networks not ready for AI workloads

Inadequate AI networking infrastructure has resulted in data issues, higher costs, and delayed implementation, the Juniper study found.

Juniper competitor Cisco reported a similar result in its own recent AI study, which found most current enterprise networks are not equipped to meet AI workloads. Businesses understand that AI will increase infrastructure workloads, but only 17% have networks that are fully flexible to handle the complexity, Cisco reported.

“23% of companies have limited or no scalability at all when it comes to meeting new AI challenges within their current IT infrastructures,” Cisco stated. “To accommodate AI’s increased power and computing demands, more than three-quarters of companies will require further data center graphics processing units (GPUs) to support current and future AI workloads. In addition, 30% say the latency and throughput of their network is not optimal or sub-optimal, and 48% agree that they need further improvements on this front to cater to future needs.”

“Enterprises recognize the need to harness this technology to propel their businesses forward. However, amidst what seems like unlimited potential, IT leaders can be at a loss as to what concrete steps to take next,” according to Dell Oro Group research director Siân Morgan, who wrote a blog this week, “Enterprises Brace For AI.”

Enterprises are only just beginning to develop strategic plans that include the benefits of AI applications, according to Morgan. “However, investments in AIOps can be made today, and will dramatically improve an organization’s efficiency,” Morgan wrote.

“AIOps make use of advanced analytics and ML algorithms to support the complex tasks of network and data center operations, helping to increase data center storage efficiency, predict network performance issues, or even automatically suggest and apply fixes to problems,” Morgan wrote. 

“The foundation of AIOps is accurate input data. Network mapping ensures that all IT resources are identified, understood, and visualized, and that the relationships between them are captured, even as configurations change,” Morgan wrote. “AI/ML algorithms applied to the combination of network mapping data and real-time usage metrics can automate a wide range of operations tasks – and may even lead the industry to the nirvana of network management: closed-loop, or fully automated, operations.”

Another issue is that AI feels very different from other breakout technologies of recent decades, such as cloud, Internet of Things (IoT), and mobile, Mandell wrote.

“AI is not just about implementing a new tool or application for efficiency; it’s also about analyzing the impact it may have on their entire organization,” Mandell stated. “The fear of the unknown and the uncertainty of the consequences make AI adoption a much more complex and thought-provoking challenge for CIOs than most previous technology breakthroughs.”

According to the Juniper study, some of the AI challenges that IT teams face include:

  • Only 1% of respondents say they are not worried about any AI vulnerabilities, including privacy breaches, data poisoning, data leaks or other cyber attacks.
  • 87% say it may not be possible to know if their company’s AI output is accurate.
  • 89% say employees trust AI more than they should.
  • 90% of leaders say all or most of their AI outputs are influenced by bias – and just 1% say there is not impact from bias.
  • 78% of those surveyed say they are experiencing errors, almost twice as many leaders believe it’s more likely there are results of inaccuracies in the information AI systems are sourcing from compared to issues with the AI algorithm.

Data Center, Generative AI, Networking



Source link