Data in Action

While AI is enabling a new era of convenience, making it easier for customers to research products, it's also reducing the role a brand’s website plays in the customer experience. As users begin to use AI platforms to make transactions on their behalf, brands have the potential to disappear into the background.

In fact, agentic AI is reshaping customer behavior just as e-commerce did decades ago.  

“I am concerned for our retailers — that they're going to lose everything between trigger and purchase,” Nikki Baird, VP of strategy and product at Aptos, said. “Yes, the consumer knows that they need something, but you may not have any visibility into any of the journey that ultimately led to you.”

Some brands see a clear role for themselves in AI-powered experiences. Other experts voiced concern about how the rise of AI agents could put up barriers between customers and brands throughout the experience, from data collection to relationship building. 

“There is no customer journey if you are relying on agents,”  said Jon Copestake, global senior consumer analyst at EY. Companies would need to design journeys based on AI’s needs, and in the process they could lose control over the customer experience.

“If they can't generate that experience for the customer, then there isn't really going to be a sense of loyalty to the retailer,” Copestake said. “That means retailers face potential disintermediation because AI and the consumers themselves won't necessarily see strength in your brands.”

AI agents could also deprive brands of the data they use to craft and personalize customer experiences, according to Baird. Third-party AI already needs inventory and location data, and as they grow more powerful, search history and intent data could become necessary as well.

The third-party platforms may then charge companies for what they learned from customer journeys made using their AI, according to Baird. The relationship would be similar to how Google AdSense works right now.

“For ChatGPT in particular, OpenAI needs revenue sources,” Baird said. “I would be extremely wary as a retailer that they're going to turn that around and charge me for it. That's coming.” 

A matter of trust

Another major challenge that may arise from AI-led journeys is how it could affect consumer trust, especially when third-party agents make mistakes.

For retailers in particular, trust is usually just a matter of getting the customer experience right, according to Copestake. However, when AI guides the experience, the people in the company have less control over its accuracy.

“If a retailer gets it right 99 times and then gets it wrong once, then the consumer will remember the once,” Copestake said. “One of the big challenges is that generative AI may say the right thing or do the right thing for consumers 99% of the time, but it's the 1% that people will pick up on the most.”

How to find the perfect response to a bad experience is the “million dollar question,” Copestake said. The best route a company can take to regain trust is to react immediately and transparently and let the consumer see that they’re trying to repair the problem. 

This is less of an issue with associates using AI in-store, according to Copestake. In that case AI is shaping how workers serve the customer, but a human is right there to validate the information. 

The larger challenge may appear when customers let AI handle the shopping process for them without any external validation, such as when a shopper transacts through a third-party platform. 

It’s in vendors’ best interest to help companies serve as the merchant of record for their services, according to Anil Jain, global managing director for global strategic industries at Google.

“If the experience is poor, you've not only upset the consumer, but you've created conflict and distrust with the retailer, even though it's not the retailer's fault,” Jain said. “A big part of this is consumer consent and trust, but also the control that both the consumer and the retailer can have.”

Home Depot is one brand working to position itself as the merchant of record, maintaining its brand voice through its product selection and support. 

“There is a desire to have a company behind that,” Home Depot CIO Angie Brown said. “That's what we mean by the merchant of record. We feel like the brand aspect of that is still very, very strong through product selection and especially from there forward.”

Clarity can suffer when AIs meet

Trust and ownership become truly complicated once a customer is using a third-party AI to make a purchase that gets handled by a company’s in-house AI.

Mistakes happen, and it may be unclear whether the fault lies with the business selling the goods or the third-party AI. One approach to solving this problem is for companies to divide their communication around AI in two, according to Copestake.

First, retailers should establish what their capabilities are along with any promises they can make about their own infrastructure, according to Copestake. From there they can craft a second message that sets expectations for customers who arrive through third-party AI.

“It's probably relatively simple for retailers to do in terms of saying the buck stops here with our AI usage,” Copestake said. “We may have built a platform where if you directly interact with our AI, then that responsibility lies with us. If you are contacting our AI through a third party AI, then clearly that's something that needs to be addressed to this issue between the two AIs.”

Data is another potential point of concern. Companies will likely only want to share necessary details with third-party AI agents, not offer unfettered access to their entire system. They will also want to retain access to information about customers arriving through third-party sources. 

One way to regulate the flow of information from one company's AI to another is via specialized concierge agents, according to David Jennings, global leader of CX and agentic AI solutions at Cisco. Concierge agents are built specifically for communicating with third-party agents, including handling their requests and recording relevant customer data.

“These specialized agents can be built by the teams or they can be built by third parties, but you still know where the customer is living all the way through the flow,” Jennings said. “They're always living in that initial concierge, as opposed to if you had multiple agents where every single time you're interacting in a disparate way.”

Opportunities to stand out remain

While companies can’t control how customers shop through AI, that doesn’t mean they should expect to relinquish control over the experience.

Customers are less likely to hand the shopping journey over to AI for tactile goods like fashion, according to Copestake. In contrast, the weekly grocery trip is more likely to become automated.

Many categories will have some journeys well-suited for AI and others where customers will seek a personal touch, Copestake said. A purchase like socks is easily automated despite being part of the fashion industry, while consumers will likely want to choose expensive foods with a shorter shelf life like meat or cheese.

Companies can make their experience stand out from generic AI advice by positioning themselves as the arbiters of brand knowledge, according to Baird. A customer who has a strong relationship with a retailer is more likely to shop with them directly.

“You need to get in front of that trigger so that they're already predisposed ask, ‘Well, what about New Balance?’” Baird said. “I don't need to go to ChatGPT and get advice because I know and trust New Balance and I just want to go there.”

Companies’ AI tools remain highly vulnerable to cyberattacks, even as enterprises race to use them in more ways, the security firm Zscaler said in a threat report.

Enterprises are also feeding AI tools vastly more data, the report found, “which paints an expanding target on AI platforms for cybercriminals across the globe.”

Zscaler recommended organizations focus on visibility, real-time defense and consistent governance controls.

One of the most striking findings in Zscaler’s report concerns how brittle many AI systems are. “They break almost immediately,” researchers wrote. “When full adversarial scans are run, critical vulnerabilities surface within minutes — and sometimes faster.”

During Zscaler’s red-teaming exercises in 25 corporate environments, it took a median of 16 minutes for an AI system to experience its first major failure, and by 90 minutes, 90% of systems had failed. In one case, it took only a single second for a system to fail.

Researchers observed failures in categories including biased and off-topic responses, failed URL verifications and privacy violations. “Models can still be coerced into exposing sensitive data or participating in harmful workflows,” the report warned.

In 72% of corporate environments, Zscaler’s first test of an AI system uncovered a critical vulnerability.

The lesson for CISOs, the report said, is that “critical risk is present from day one, even in mature environments,” and security personnel need to constantly test their systems and apply strict governance protocols.

At the same time, Zscaler’s analysis of nearly one trillion AI data transactions in its cloud environment in 2025 revealed some promising signs. Companies’ security policies blocked roughly 40% of all attempted AI transactions, a phenomenon that Zscaler said reflected “governance in action … as leaders balance the tradeoff between innovation speed and risk tolerance.”

The 989.3 billion AI transactions that Zscaler observed in 2025 represented a 91% increase from 2024, with the company tracking activity from more than 3,400 different AI tools.

The U.S. accounted for roughly 38% of the transactions, followed by India (14%) and Canada (5%). And the finance and manufacturing sectors led the way in using AI for the third year in a row, representing 23% and 20% of AI transactions, respectively, in 2025.

Sponsored content

By Lumen Technologies

Every day, businesses move staggering amounts of data between data centers, clouds and offices. Most of this traffic travels on shared networks, much like rush hour on a crowded highway. But what if your mission critical workloads need higher capacity, more speed and more security? That is where wavelengths come in.

Imagine your own data expressway

Picture a Formula 1 car on a private expressway with no traffic and no speed limits, just a clear shot from start to finish. That is the power of a wavelength, a dedicated high-capacity connection that moves your data at the speed of light directly from point A to point B. No congestion. No slowdowns. Maximum scalability and security.

What is a wavelength?

A wavelength is a dedicated high-capacity point-to-point connectivity service that uses optical fiber to carry massive amounts of data with very low latency. Through a technique called Dense Wave Division Multiplexing (DWDM), modern optical networks can transmit multiple colors of light (wavelengths) at the same time through a single fiber, creating a private multi-lane superhighway. Each lane is isolated, so your data can travel securely and without interference from other traffic.

How do wavelengths work?

A wavelength works by converting your electronic data into pulses of light, each traveling on its own lane within a fiber optic cable. You can think of it as giving your data its own color of light so it moves independently and does not get mixed with other traffic. Specialized equipment at each end converts your data into light, sends it down the fiber and then turns it back into electronic signals at the destination.

Because this light travels on a direct optical path, your data moves incredibly fast with minimal loss or delay. Today’s wavelength services offer speeds such as 10 Gbps, 100 Gbps and 400 Gbps, with even higher rates like 800G and 1.6T on the horizon. Multiple wavelengths can run on the same fiber, allowing you to move enormous amounts of data in seconds. For a visual breakdown of how massive new data demands are driving network innovation, check out this infographic.

If you’re interested in learning how these advances in high-speed connectivity will power the evolution of cloud architectures and prepare businesses for what’s next, explore our Cloud 2.0 report.

When do companies use wavelengths?

Businesses rely on wavelength services when they need scalable, high-capacity connectivity between specific sites. Common use cases include:

• AI and analytics that require transferring massive datasets between compute clusters
• Data center interconnects for synchronizing databases or large-scale backups
• Cloud access with direct private connections to major cloud providers
• Media and entertainment workflows like streaming, broadcasting and moving large video files
• Disaster recovery or network diversity with a ready-to-activate backup connection

How wavelengths compare with other network options

Wavelengths are among multiple network options your business can explore to find what best fits its individual needs.

Option	How it works	Best for
Internet	Public network that carries everyone’s traffic	Unscaled everyday web use and SaaS applications
Ethernet/Private WAN	Private connections that link multiple sites through packet switching	Branch or campus networks where flexibility matters
Wavelengths	Dedicated optical channel between two fixed points	High-volume or latency-sensitive workloads
Dark Fiber	You lease or own the fiber and operate your own optical gear	Organizations that want complete control and have in-house expertise

A wavelength sits between Ethernet and dark fiber. It delivers the control and performance of owning fiber, without the cost or complexity of managing it yourself.

Ready to put your data in the fast lane?

Wavelengths deliver the speed, security and scalability modern businesses require. If your workloads cannot afford delays, it may be time to consider a dedicated wavelength connection.

Explore Wavelengths today and learn why Lumen ranked #1 in Wavelengths Services in North America.¹

Learn more.

1Vertical Systems Group, Year-end U.S. Wavelength LEADERBOARD, 2024.

FRISCO, Texas — While the Dallas Cowboys prepare for another football season, with the NFL Draft just weeks away, the franchise’s IT team is gearing up behind the scenes to play a season of its own. 

Walking through the halls of The Star, the team’s practice facility, CIO Matt Messick described wrapping up a lengthy refresh cycle to deploy Lenovo devices including laptops and desktop computers, migrating to a new ERP system and readying the data foundation for the long game — AI adoption. 

Messick plans to create an AI dashboard for stadium operations staff ahead of the 2026 FIFA World Cup, which starts in June. Dallas will host nine World Cup games, which will be played at AT&T Stadium in Arlington, Texas — the Cowboys’ home stadium. But getting the Cowboys’ IT infrastructure ready for AI with a lean IT team has taken time, Messick said. 

Beyond fixing systems issues and providing computers, Messick said he wants to demonstrate how technology can boost intelligent insights and improve decision-making for the franchise. 

“It's showing the business value of technology,” Messick said. “That's our No. 1 goal.” 

Starting with the foundation

Part of the team’s AI preparation has included breaking down data silos.

“Our primary focus is making sure we have a true data strategy that can support anything that we want,” Messick said. 

While working on the data foundation, the IT team completed a device refresh that was delayed due to the COVID-19 pandemic. The Cowboys partnered with Lenovo in 2023 to begin the refresh. 

Lenovo served as a strategic partner for the Cowboys and helped address varying device needs across the staff, said Ben Barnes, senior manager of IT enterprise support for the Dallas Cowboys. 

Working with Lenovo, Barnes honed which type of processor would better support talent scouts — laptop-focused users who work on the road. Scouts frequently analyze video, and the Cowboys’ video platform represents an “extremely large database” that’s heavy on a PC, added Jason McKay, director of football administration and technology for the Dallas Cowboys. 

“At the beginning, I picked a certain AMD processor, one that heated up pretty rapidly, which is OK for the standard user who's sitting at their desk doing something, but not for the user who's moving around, closing it, not turning it off and putting it in a bag,” Barnes said. Through several iterations with Lenovo, Barnes said they ultimately picked an Intel processor less prone to overheating to better support scouting staff. 

The Cowboys also use Lenovo devices to provide analytics to coaches on the field and inventory management for team stores, according to Lenovo. 

In addition to cleaning up data and refreshing devices, the IT team began a two-year ERP system migration from Microsoft Dynamics AX to Dynamics 365, which is set to conclude in April. 

“It’s all getting to the point where it’s time to stop the build and start focusing on the automations and leveraging data,” Messick said. “Being with the business so we can truly leverage everything we’re building.” 

Setting up the AI touchdown

The IT team is in the early phases of its AI vision, still focused on clean, structured data and building its AI governance. 

The NFL itself has been experimenting with its use of AI, partnering with Microsoft in 2025 to enable AI use for football and business staff. It also teamed up with AWS in 2025 to deliver better insights from the annual NFL Scouting Combine and earlier in 2024 to build a digital athlete as a way to help predict and prevent player injuries. 

Messick said he has targeted specific AI use cases, from providing assistance to employees or enabling insights for scouting teams. The IT team has adopted Microsoft Copilot as the primary AI tool for staff to start experimenting with the technology, he added. 

It’s critical that CIOs get some wins from low-hanging fruit along the way to highlight the technology’s value to the business. Amid AI adoption efforts, “you need everybody to work with you,” Messick said. 

“You’re also changing the way people fundamentally work every single day,” he said. 

The FIFA World Cup provided the Cowboys’ IT team with an opportunity to bring some of the AI plans to fruition. Messick and his team spent the last two months identifying every data feed and data point the stadium generates, such as traffic flow, Wi-Fi and ticketing data. The team plans to pull all the disparate data sources together into one AI dashboard that translates command center alerts to real world actions, he said. 

“You’re really having one command center dashboard that’s translating and bringing all of this relevant data in real time to the people that need it the most,” he said. 

With a number of heavy infrastructure projects behind them, Messick said his goal is to identify where AI can drive efficiencies and make employees’ lives easier as the technology matures. 

“It’s the perfect time,” Messick said. 

Data infrastructure issues are derailing enterprise AI investments, leading to $108 billion in wasted annual AI spend, according to a Hitachi Vantara analysis. The tech provider surveyed 1,200 IT decision makers for its report.

More than 4 in 5 businesses with mature data estates reported ROI on their AI investments, compared with less than half of what the firm described as “data laggards.” Only 43% of U.S. leaders believe that their companies have predictive or automated infrastructure operations.

Despite data limitations, surveyed IT leaders expect AI spending to rise by 76% in the next two years as businesses spin up in-house platforms and plan broader deployment efforts.

Data and infrastructure have been stubborn components of AI implementations, hindering the training of models for targeted use cases and creating bottlenecks for in-demand services.

Hyperscalers are sprinting to deploy capital, ramping up efforts in the past two years with plans to match available compute capacity with soaring demand for AI services. This year, the spending blitz continues — large cloud providers plan to increase their capital investments by nearly 40% in order to adequately service demand, according to a recent report from S&P Global.

In the enterprise, data sprawl could be hampering broader AI adoption, as data quality is the most commonly cited factor in successful AI implementations, according to Hitachi Vantara.  

“The fastest way to waste AI budget is to chase models without fixing the data underneath them," said Simon Ninan, SVP of business strategy at Hitachi Vantara, in an email to CIO Dive. The company works with customers to align AI spending to business outcomes, focusing on data performance, resiliency and availability, Ninan said.

"CIOs get the most out of AI spending by modernizing their data infrastructure, setting clear ROI benchmarks and treating AI as an operational system, not an experiment," Ninan said.

CIOs and CISOs are grappling with information complexity as they look for a strong security posture amid adoption efforts, according to a Ponemon Institute study. Nearly 3 in 4 leaders said reducing information complexity can help scale AI tools.

Agentic AI adoption is accelerating across industries, but governance frameworks are struggling to keep pace, according to Deloitte’s State of AI in the Enterprise report. The company surveyed more than 3,000 business and IT leaders across 24 countries.

Nearly 3 in 4 companies plan to deploy agentic AI within two years. Despite ambitious plans for deployment, only 1 in 5 report having a mature model for governing autonomous agents, a gap that can increase the risk of vulnerabilities, according to Deloitte.

To improve accountability as technology plans unfold, businesses should “anchor agentic AI programs in measurable impact from day one,” Jim Rowan, U.S. head of AI at Deloitte told CIO Dive in an email.

With AI adoption on the rise, agentic applications are emerging as a key growth catalyst. The tech is shifting AI from a decision-support tool to one capable of autonomous action.

While the capabilities race ahead, the necessary guardrails have been slow to follow, raising the alarm on accountability, risk management and regulatory compliance, according to Deloitte.

To scale AI safely and effectively, organizations must restructure from the bottom up and embed governance early in transformation efforts. 

“Speed and governance should not be viewed as opposing forces, as there can be alignment when you establish the right foundational capabilities upfront,” Rowan said. “We’ve seen clear patterns indicating that organizations without strong governance have difficulty scaling beyond the pilot stage. Focusing on this early allows for innovation and growth at pace with the technology itself, instead of facing stalled deployments and operational risks.”

CIOs facing pressure to deploy agentic AI quickly should prioritize governance elements such as a cross-functional team that brings IT, legal, compliance and business leaders together, according to Rowan.

Executives should also establish clear autonomy boundaries, create real-time monitoring for anomaly detection and add comprehensive audit trails to “capture the full chain of agent actions for compliance and improvement,” Rowan said. 

“The fastest-moving CIOs recognize that governance upfront eliminates the far costlier cycle of deployment, remediation and rollback,” Rowan added. “These elements of a governance strategy enable speed by building trust and reducing crisis management overhead.”

Demand for strong governance is only expected to continue rising as agentic, physical and sovereign AI see ramping deployment across industries. 

A previous report from IBM found governance gaps can exacerbate the financial and reputational repercussions of breaches, while an EY survey last year found more than 3 in 5 businesses experience AI risk-related losses of more than $1 million.

Snowflake and OpenAI signed a $200 million deal to bring OpenAI models directly into Snowflake’s cloud data platform, the two companies said. 

The agreement will support agentic AI adoption across enterprises, allowing customers to create and deploy context-aware AI agents without requiring coding experience, according to the announcement.

“By combining OpenAI’s advanced reasoning with Snowflake’s enterprise-grade data platform, organizations can scale AI responsibly and give more employees access to intelligence while ensuring data stays secure and usage is governed,” Baris Gultekin, VP of AI at Snowflake, said in an email to CIO Dive. 

Enterprise leaders expect autonomous AI agents to handle complex workflows while managing complex data estates. But deploying these systems at scale has proven difficult, particularly as organizations grapple with governance and accountability concerns.

Snowflake’s multiyear partnership with OpenAI aims to bridge this gap.

By bringing OpenAI models directly into Snowflake’s platform, the companies said, customers can use AI in the same governed environment where their data already lives, building and deploying agents without moving data into standalone tools and increasing risk.

“For CIOs, the biggest challenge with standalone AI tools is fragmentation,” Gultekin said. “Data has to be moved, access controls re-implemented, and governance often becomes an afterthought. Embedding OpenAI models directly into Snowflake eliminates those tradeoffs.”

OpenAI models such as GPT-5.2 will also be accessible through Snowflake Intelligence, the company’s enterprise intelligence agent. 

Ease of use is a central focus, with employees able to interact with these agents using natural language and receive responses from the company’s internal database, without needing to write code.

Several customers including design platform Canva and fitness wearables maker Whoop, are already using the platform to accelerate research and internal decision-making.

For Snowflake, the OpenAI deal is part of a broader push to expand its enterprise AI footprint. 

In recent months the company deepened partnerships with Anthropic and Accenture to scale agentic AI adoption, and acquired AI-powered observability platform Observe, underscoring its focus on reliability and governance as AI becomes embedded in core business systems.

Although agentic AI has taken the technology ecosystem by storm, one CIO believes that small to midsize businesses should prioritize actionable AI use cases to succeed.  

“I feel like we haven’t really wrapped our heads around exactly what all we can get an agent to do,” said Matt Price, CIO of promotional products company Gold Bond Inc. “It’s still a gray area for us as to where we can really use it.” The real impact of AI for the Tennessee-based company has taken the shape of targeted use cases that do away with busy work.

He’s not alone in his evaluation of the technology. Many midsized organizations are working with a handful of AI use cases, deploying everyday tools like Gemini or Microsoft Copilot in an attempt to boost user adoption as they grapple with continuously unfolding developments, said Alys Woodward, senior director analyst at Gartner. 

“Even in the midsized, people are definitely aware [AI] is happening and they are out there and implementing,” Woodward said of AI developments, including AI agents. 

To scale its adoption of AI, Gold Bond tapped Promevo, a Google-focused partner, to lead workshops teaching employees how to use Gemini in 2024. Employee Gemini usage rose from 20% to 71% following the workshops, with nearly three-quarters of employees indicating the technology saved them up to an hour of time per day, according to Promevo. 

Since then, the company has embarked on a number of AI projects, Price said. 

Gold Bond collects around 120,000 invoices annually from its distributors. To help improve distributors’ marketing to end users, Gold Bond created a part-time role to categorize customer artwork orders and record the user data in its Oracle NetSuite ERP system. 

“We said: We think we can use AI to automate this,” Price said. 

Price built a pilot project using the Vertex AI platform in Google Cloud, powered by Gemini, and deployed a SuiteScript within Oracle NetSuite, automating the entire process. The part-time worker became a company representative as AI took over more administrative tasks, Price said. 

“Every night, we throw 600 to 700 invoices at Gemini,” Price said. “It analyzes all the artwork and categorizes it properly for us.” 

Another project where Gold Bond is seeing promising results is via image generating model Nano Banana Pro, which has allowed the company to automate some of the product design process, Price said. 

“We’re all about image editing in our business, putting brands on drinkware, things like that,” Price said.

Finding the small wins

While invoice workflow automation and image generation mark significant AI wins for the company, Gold Bond has more projects underway for 2026, Price said. 

Working with a telecommunications partner, Gold Bond plans to adopt an AI voice agent this year that will assist customers with order information, status and bill payment. The agent will launch for after-hours, weekend and holiday support to start and will eventually be incorporated during working hours, where it can escalate issues to employees and learn from interactions, Price said. 

“We have about 500 orders a day coming through the building,” he said. “The AI voice agent, we feel like it’s going to be a great win for us.” 

For SMBs to see success in their AI ventures, Price recommended executives pursue the targeted wins. 

“If you can, take small use cases,” Price said. “Find where the company is paying X amount, or using this amount of resources, and see if you can get AI to automate that small workflow and get some small wins.” 

As CIOs face high expectations to produce return on investment for their AI projects, Woodward said it’s critical for CIOs of midsized organizations to find sources of value including employee satisfaction, advancement along the AI maturity journey, faster customer service and improved efficiency. 

AI agents beyond IT

Although Price has begun experimenting with agents and is identifying ways to make the technology work within Gold Bond’s operations, he said it’s not just the IT team looking at what the technology can do. 

Price said he has what he calls “technology ninjas” in accounting, sales, customer support, manufacturing and marketing departments interested in what both generative and agentic AI tools can bring to the table. 

Social media is one area the company’s marketing team is evaluating as a potential AI use case, Price said. 

“The marketing team is working through that now, where an AI agent would post at certain times, different personas, things like that,” he said. 

When it works well, using AI agents can take pressure off the CIO as different parts of the business automate some of their own workflows, Woodward said.  

Data plays a key role in AI agent success, which is why they can be particularly effective when deployed within departments such as finance or marketing and used on that specific data, she added. 

CIOs will need to prioritize cleaning up technical debt and improving data quality to enable a better foundation for supporting future AI use cases, including agents, said Mark Moccia, VP and research director at Forrester. 

While one department might operate on a modern IT system, another might be functioning on a legacy system that won’t be able to scale AI the same way, Moccia said. 

“It could create a mismatch of expectations with the C-suite, who just saw you do something quickly, and now this one’s taking longer,” Moccia said of AI projects in different departments. “You’ve always got to reframe back to the foundational investments to be successful.” 

Governance comes into play

AI governance will be a big piece of the puzzle for successful, scaled adoption, Woodward said. 

Business users are essentially constructing IT systems as they build AI agents, and they might be unaware of considerations IT departments inherently know — such as data access, privacy controls and security measures. CIOs must ensure only business users who have gone through AI literacy training can access such systems, Woodward said. 

Less than half of IT decision-makers have formal AI governance policies in place, while only 47% make governance and compliance training available to employees, according to a Collibra survey. 

“If you’re letting people build AI agents just for themselves, you don’t need to govern that,” she said. “But if you’re going to have agents shared between departments, that’s where you need to start thinking: 'Hold on, we need to make sure people are educated around this.'”