Dell Technologies CEO expects AI ubiquity in the enterprise

AI for the business

Lowe's, an American hardware and home improvement retailer, uses Dell's hardware to support AI applications for its more than 1,700 stores and 300,000 retail employees.

Quick access to product or service information for store associates, regardless of their expertise or department, is the linchpin for Lowe's AI strategy, said Seemantini Godbole, executive vice president and chief digital and information officer at Lowe's, during Monday's keynote.

Each store has a small deployment of Dell server hardware to support AI agents in handheld devices from Zebra Technologies. These agents enable employees to find product information and answer questions based on Lowe's data. The AI also provides in-store alerts to notify employees if customers are seeking help.

"AI is following the data," she said. "We thought we'd do a micro data center in the back of the store."

Godbole expects to build a variety of data centers for specific use cases within the business beyond retail operations, including for engineering and executive functions, while making sure new AI adoption matches specific needs.

"We'll have lots of different [deployment] sizes because one size doesn't fit all," Godbole said. "We don't want to die a death of 1,000 pilots."

Engineering has already found a use for AI by enabling it to conduct code reviews without manager oversight, she said. The change has made the process less confrontational.

"You tend to take [code reviews] less emotionally from a machine than a manager," Godbole said.

Santosh Agrawal

Optimistic feedback from early adopters like Lowe's is inspiring other customers to follow, said Santosh Agrawal, co-founder and managing director of Esconet Technologies Ltd., an IT consulting and technology vendor in New Delhi.

"The reality is what Michael said on stage," Agrawal said. "Almost every customer is looking forward to creating their own private AI and has plans to use AI in their business."

The major challenge he's found is that many buyers are at varying levels of preparedness for AI deployments. Offerings like the Dell AI Factory and other integrations presented by Dell can help his company deploy hardware for AI purposes more quickly, Agrawal said.

"There are customers who have a clear vision as to what they exactly want and what models they want to use," he said. "There are others who don't know much about it and lean on partners like us and OEMs like Dell."

The hype is moving so much faster than business [implementation].

Many enterprises are adopting and implementing AI into the business at a more measured pace than the rate of LLM releases and innovation, said Mike Leone, an analyst at Enterprise Strategy Group, now part of Omdia.

Dell's approach of bundling its hardware with partner AI offerings can increase the speed to deployment, he said, but many organizations still need to find worthwhile use cases for AI and train end users on its effective use in their roles.

"The hype is moving so much faster than business [implementation]," Leone said. "A lot of organizations still haven't made progress on generative AI, and now we're introducing agents. That complexity is driving a greater need for partners."

AI security

The hype combined with customer inexperience is creating new security hazards and challenges, warned Maxim Balin, senior consultant and product manager for cybersecurity edge and AIOps at Dell Technologies.

Data and infrastructure for AI will present a new set of challenges for many customers, which their existing security might not be prepared for, Balin said at a Monday conference session.

Maxim Balin

New attack vectors will include hackers manipulating AI agents into giving worthless responses through manipulation campaigns, he said. Another attack might come from poisoning the AI model's data -- introducing enough garbage or incorrect information into data sets to avoid immediate human detection, but slowly ruining responses.

Preventing these attacks will require new models of thinking about existing cybersecurity, such as zero trust.

"I'm meeting with customers who said they've spent a fortune and asked why [their security investments] are not good to deal with AI," Balin said. "But they're not designed to deal with logic [attacks]."

Expanding security models such as zero trust into the AI space will be a challenge for vendors, said Krista Case, an analyst at The Futurum Group. Existing identity vendors CyberArk, Okta and 1Password will need to consider the identity and role of AI agents and the intentionality behind their actions.

"We're still seeing the formulation of the best strategies for AI security," Case said. "But how do we start to understand the intent of an AI agent and if it's malicious or not? There isn't a silver bullet yet."

Still, security and infrastructure teams can lay the foundation for securing AI by accepting it as a cross-departmental challenge, Balin said.

"Never approach this challenge by looking into a specific area," he said. "We need to control the entire stack. We're only at the beginning."

Tim McCarthy is a news writer for Informa TechTarget covering cloud and data storage.