AI pilots expose fragmented IT data and configuration gaps
Artificial intelligence initiatives in IT operations frequently fail because they surface existing data silos and inconsistent records rather than creating new technical problems.
Key points
- Discrepancies in device counts across endpoint, security, and management tools confuse AI models.
- Ivanti research indicates nearly half of security pros say IT teams lack urgency on security concerns.
- AI assistants trained on inconsistent historical ticket data replicate routing errors at scale.
Artificial intelligence pilots in IT operations often underdeliver because they expose pre-existing data fragmentation rather than failing due to model errors or poor use-case selection, according to Computerworld. The technology surfaces configuration management database issues and inconsistent service desk categorizations quickly and publicly.
Most IT environments accumulated tools independently, with endpoint, security, service management, and procurement teams adopting separate platforms that were never designed to share data. This creates a landscape where every team holds a partial view of the same environment, leading to conflicting records that confuse automated systems.
Data Silos and Tooling Gaps
The endpoint console might show 9,400 devices while a security scanner sees 8,700 and the configuration management database lists 10,200, including hundreds of retired machines. No single system is entirely wrong, but none are complete. Experienced analysts often work around these gaps using institutional knowledge, which remains invisible to AI systems and causes immediate operational trouble.
Fragmentation is not purely technical. Ivanti’s 2026 State of Cybersecurity research found that 48% of security professionals say IT teams do not respond urgently to security concerns. Additionally, 40% report that IT lacks an understanding of the organization’s risk tolerance, creating alignment gaps that shape how data is collected and labeled.
Impact on AI Outputs
When AI meets these data gaps, the outputs reflect the underlying inconsistencies. An AI assistant listing laptops missing critical updates may double-count machines or miss others if the endpoint tool and scanner disagree on device names. The technology performs as designed, but the organization lacks a shared picture of reality.
Automated ticket-routing models trained on inconsistent historical data learn that inconsistency and repeat it at scale. If three teams use the same category field in three different ways, the AI adopts those variations. Aligned organizations that resolve these data issues see measurably better outcomes from their artificial intelligence investments.
What should you do?
Organizations should audit their configuration management databases and asset records to identify discrepancies between endpoint consoles, scanners, and service desks before deploying AI tools. Teams must align on data labeling standards to ensure consistent categorization across all systems.
What to do and how to stay safe: AI
- Reconcile configuration management database records with endpoint and security scanner data to eliminate duplicate or retired device entries.
- Standardize service desk category fields across all teams to ensure consistent historical data for AI training.
- Verify that IT and security teams share a common understanding of risk tolerance and response urgency protocols.
- Monitor AI outputs for double-counting or missed assets that indicate underlying data fragmentation issues.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why do AI pilots in IT operations often fail?
They fail because they expose existing data fragmentation, such as outdated configuration management databases and inconsistent service desk categorizations, rather than due to flawed AI models.
How do data silos affect AI performance?
AI assistants may double-count devices or miss updates if endpoint tools and scanners disagree on device names, while ticket-routing models replicate inconsistent historical categorization at scale.
What does Ivanti research say about IT and security alignment?
Ivanti’s 2026 research found that 48% of security professionals say IT teams do not respond urgently to security concerns, and 40% say IT lacks understanding of organizational risk tolerance.
