Major firms halt AI rollouts as hidden data and confusion expose implementation challenges

AI Rollouts Expose Decades of Hidden Data Problems

AI in business is revealing uncomfortable truths about data management that many organizations never anticipated. At Fidelity Investments, the deployment of generative AI tools brought AI rollouts to a screeching halt just two days after launch when legal teams discovered the technology was surfacing decades-old PowerPoint presentations and PDF research notes from forgotten SharePoint sites 1

1

. Steve MacIntyre, senior vice president at the 400,000-employee firm, explained that AI proved to be a "tremendous search engine that runs at speed," suddenly making unstructured data that nobody thought mattered become valuable again 1

1

.

Similar data governance challenges emerged at EY, where chief architect Wim Geurden described discovering multiple petabytes of data across SharePoint sites with no lifecycle management and half with no identifiable owners 1

1

. The global consulting firm faced the added complexity of data ownership across independent member firms, forcing them to shut down access and implement licensing restrictions while they sorted through what AI had exposed 1

1

. Both organizations learned that their AI implementation challenges weren't about the technology itself, but about securing and governing information that had accumulated over years.

Confused AI Strategy Creates Employee Skepticism and Confusion

The disconnect between executive enthusiasm and practical implementation is creating what researchers call an "activation gap." According to Cognizant data presented at Fortune's COO Summit, 93% of jobs are already being disrupted by AI—six years ahead of projections—yet the promised productivity gains haven't materialized 3

3

. Malcolm, an AI engineer at a data analysis firm, experienced this firsthand when executives insisted on using generative AI for customer database categorization despite his recommendation for traditional machine learning, resulting in a process that was less accurate and more expensive but allowed the company to claim AI adoption in the workplace 2

2

.

Dan Boyles, CEO of consultancy Hello AI Collective, described sitting with an oil and gas company's C-suite where none could agree on their AI strategy—the CEO cited competitive pressure, sales wanted revenue growth, and marketing sought to eliminate contractors 2

2

. This lack of clarity cascades down to workers. Research by the civil servant union FDA found less than a third of civil servants had been consulted on AI rollouts, meaning "change is being done to workers, not with them" 2

2

. Culture Amp's research revealed that while nine out of 10 HR professionals expect to increase generative AI use, a third say no one currently owns AI strategy at their companies 2

2

.

The Automation Illusion Confronting Fortune 500 Executives

At Fortune's COO Summit, leaders from Nike, Sysco, and Box described what they termed the automation illusion—the dangerous gap between AI promises and operational delivery 5

5

. Venkatesh Alagirisamy, EVP and COO of Nike, warned about "speed without clarity," noting that hype around AI drives organizational energy to adopt without purpose, potentially pushing companies in the wrong direction 5

5

.

Laura Clayton McDonnell, president of corporates at Thomson Reuters, highlighted the reliability crisis: "We're going to move fast, we're going to get these answers really quickly, but what about making sure that output is reliable, it's accurate?" 5

5

. For professionals like lawyers and accountants, AI-generated errors create liability rather than productivity. Research from Freshworks found that 86% of mid-market IT leaders say managing AI complexity has actually increased their team's workload, with 80% reporting that AI outputs introduce what the report terms "AI slop"—noise, errors, or rework 4

4

.

Managing AI Agents Sparks Unprecedented Debate Among COOs

The question of whether to treat agentic AI as colleagues or workflow tool has divided leadership. Eric Kelleher, President and COO of Okta, named his AI agents Leo, Sloan, Hank, and Walker, including them in business reviews alongside human staff, believing this approach transforms AI from tool to colleague 3

3

. Francine Katsoudas, Chief People Officer at Cisco, pushed back forcefully: "I would not look at AI as a colleague. I think we should look at AI and agents as part of the workflow, but not a colleague" 3

3

.

Aayush Bhatnagar at Sysco revealed he added seven AI agents to his direct reports four weeks prior, with defined roles like escalation agent and delivery agent, admitting "I lost some sleep that night, thinking that our traditional laws of leadership, principles of leadership, do not apply to these agentic agents" 5

5

. Research compounds the confusion: a Harvard Business Review experiment found that humanizing AI shifts accountability away from individuals and reduces review quality, while Boston Consulting Group research showed workers scapegoat AI colleagues and become more careless 3

3

.

Compliance and Accountability Concerns Mount

A Harvard Business Review report on agentic AI warns that agents can "potentially make changes to business records and data sources," essentially admitting AI could rewrite history and decouple organizations from verifiable fact and reliable audit trails 4

4

. The compliance implications are staggering: 1,494 known examples worldwide of fake AI caselaw have been presented in completed court cases, with EY Canada retracting a fraud report due to AI-generated citations and Deloitte part-refunding the Australian government for a hallucination-filled report 4

4

.

At EY, addressing these risks required going "way beyond the labeling of confidential information" to include geo-restrictions, line-of-business labeling, and contract linkages for client data 1

1

. Geurden explained they needed historical versioning to know "what was there when the AI ran," though codifying this into technological structures remains "very, very cumbersome" 1

1

. The return on investment questions intensify as organizations struggle to demonstrate value while absorbing unexpected governance costs and managing the workload AI was supposed to eliminate.🟡 injurious_image_id=🟡None🟡, not_found_image_id=🟡None🟡, low_quality_image_id=🟡None