Why ITSM Data Quality Is a Strategic Issue Today: Data Quality as a Prerequisite for AI

ITSM data quality has long since ceased to be a matter of operational “hygiene” and has become a strategic choice. Among the conditions that make it possible to integrate AI into IT processes, data quality is the most underestimated and, at the same time, the most decisive. Three elements are indispensable for any AI-based automation to produce stable results: an accurate Configuration Management Database (CMDB), a structured workflow history, and consistent operational records – the digital traces of daily IT activities, from malfunction reports to changes across different systems, through to user service requests.

Data Quality as a Factor of Success (or Failure) for AI

The figures produced by international research converge on one point: today, data quality is the primary variable that separates successful AI projects from those destined to fail. According to a Gartner survey from February 2025 conducted among 1,203 data management executives, 63% of organizations did not have – or were not sure they had – adequate data management practices for AI. Gartner also predicted that by 2026, 60% of AI projects not supported by AI-ready data would be abandoned.

This figure is not isolated. Of the 782 Infrastructure & Operations executives also surveyed by Gartner, 38% stated that the AI projects they had implemented had proven to be a failure, and another 38% explicitly cited poor data quality or availability as the direct cause of the problem. Only 28% of AI projects in the IT domain generated an ROI in line with expectations.

On an even more specific front, more than half of IT professionals do not trust the data contained in their own CMDB, even though they use it as a source for ITSM workflows and automations that are supposed to be powered by AI.

What Do We Really Mean by ITSM Data Quality?

When we talk about ITSM data quality, we are not referring to “clean data” in the generic sense. It means ensuring that the information managed within the ITSM platform – that is, everything that is tracked during daily IT activities – meets four precise and verifiable requirements:

1. Completeness: every record must contain all the fields necessary to be processed reliably, whether by a human operator or an AI agent. An incident (a report of an IT service malfunction) without an assigned category, an asset without a reference to its owner, a request without a due date are all examples of incomplete data that block automation.

2. Accuracy: the recorded information corresponds to what is actually in production: the real state of an asset, the technical and functional relationships between the different components of the IT infrastructure – the Configuration Items (CIs) – and the effect produced by a change management intervention (the planned modification of a system) after its implementation.

3. Timeliness: the data reflects the reality of the moment, not a snapshot of the past. Outdated information within the system of record – the platform recognized as the official source of data for the entire organization – effectively becomes a source of misinformation.

4. Relational integrity: the relationships between the different elements are tracked, explicit, and verifiable. For example: the link between a business service and its technical components, an incident and the asset involved, and a change and the areas it affects.

Without these four conditions, even the most sophisticated AI model finds itself operating on information that does not reliably describe operational reality: it would make decisions based on a distorted picture, and its suggestions – however formally consistent – would end up being misaligned with what is actually happening in the company’s systems.

Why Without Good ITSM Data Quality AI Cannot Scale

Keith Andes, Product Marketing @ EasyVista, observes that the two most recurring phrases that IT organizations say internally are: “Our processes are not fully implemented” and, above all, “Our data is not reliable.” This is an important observation, because it precisely defines a breaking point: if the company’s system of record is incomplete or disorganized, AI can only amplify that disorder, never correct it.

Three concrete examples further clarify the problem:

1. Disorganized CMDB and automatic incident routing: If CIs are duplicated, miscategorized, or orphaned, an AI agent that assigns tickets based on application dependencies will produce wrong assignments just as quickly as it would produce correct ones. Speed does not compensate for inaccuracy.
   
2. Misaligned knowledge base: An LLM-based knowledge synthesis system that draws from outdated articles will suggest solutions that are no longer applicable, or procedures not approved for the organization’s specific environment. The apparent authority of the AI output makes the damage harder to intercept in real time.
   
3. Incomplete operational history: If tickets are closed without a proper classification of the original cause of the problem (the so-called root cause, meaning what generated the malfunction, as distinct from the visible symptoms), the AI does not have sufficient raw material to identify recurring patterns or to improve operator suggestions over time.

In all three cases, the AI model works perfectly in the laboratory. It is when it encounters real data – disorganized, inconsistent, partial – that the promised value falls short. ITSM data quality is therefore the variable that separates a successful demo, executed by the vendor in a controlled environment, from a genuinely reliable production system.

Four Priorities for Building Solid ITSM Data Quality

Once it is understood how much data quality affects the success of AI, what remains is to understand how to intervene. Recognizing the problem is the first step; addressing it requires acting on four specific fronts, in sequence and with measurable results.

1. Measure Before Intervening

You cannot improve what you do not measure. Defining explicit KPIs on data quality makes it possible to transform a perceived problem into a manageable one. Data quality dashboards are daily working tools for operational teams and must track a series of fundamental parameters such as: how many CMDB records are complete in all their mandatory fields, how long on average they have not been updated, how many duplicate records exist, and what percentage of tickets are categorized accurately.

2. Automate Data Detection and Alignment

Manually entered data ages quickly. Maintaining a reliable CMDB requires two things: automatic discovery tools, which continuously scan the infrastructure – networks, end-user devices (laptops, desktops, smartphones, servers), cloud environments – and record the elements present; and reconciliation processes, which periodically compare what appears in the CMDB with what is actually in production, flagging differences, missing assets, or configurations that are no longer valid.

Modern ITSM platforms, including solutions such as EV Discovery by EasyVista, are designed to carry out this activity in a constant and automated manner, rather than leaving it to periodic manual checks that become outdated between one session and the next.

3. Define Explicit Data Ownership

Behind every asset, service and area of the knowledge base there must be an identified owner who oversees its quality over time. Without ownership, ITSM data quality degrades through inertia. With clear ownership in place, each anomaly report is assigned to a specific recipient, and data remediation becomes a structural practice rather than an exception handled on a case-by-case basis.

4. Integrate Data Quality into the Workflow, Not Downstream

The most effective moment to ensure the quality of a piece of data is at the time of its creation. Ticket-opening forms with validation, intelligent mandatory fields, automatic categorization suggestions, consistency checks at the time of closure: all of these best practices prevent the problem at its source, rather than forcing costly corrective activities after the fact. At this level, AI can begin to provide concrete value today, supporting – rather than replacing – the operational discipline of IT teams.

Data as an Investment, Not a Cost

Improving ITSM data quality is today one of the most strategic investments an organization can make. Keith Andes’s message, from this point of view, is direct and definitive: “Stop allocating 100% of your AI budget to AI features. AI needs a solid foundation to operate on.” This solid foundation consists of documented workflows and reliable data. Organizations that choose to invest in their system of record before investing in the most eye-catching features will achieve a double result: a more robust ITSM in the short term, and a credible foundation for any agentic development in the coming years. This is how data quality ceases to be an operational topic and becomes a strategic choice with direct effects on competitiveness.

FAQs

1. What is ITSM data quality?

It is the condition of completeness, accuracy, timeliness, and relational integrity of the operational data managed within the ITSM platform: incident records, changes, assets, configuration items, knowledge base. It constitutes the raw material on which automations, operational decisions, and AI initiatives rest.

2. Why is data quality critical for AI in the ITSM domain?

Because AI does not correct the data it operates on: it amplifies it. If the CMDB is disorganized or the operational history is incomplete, AI will reproduce – and accelerate – those errors. Gartner estimates that by 2026, 60% of AI projects lacking AI-ready data will be abandoned.

3. Where is it best to start in order to improve ITSM data quality?

From measuring the current state. Before intervening, it is necessary to know where you stand: how many CMDB records are complete in all their mandatory fields, how old the configuration data is on average, how many duplicate records exist, and what percentage of tickets are categorized accurately. Without these baseline indicators, every subsequent action proceeds blindly.

4. What is the role of the CMDB in ITSM data quality?

The CMDB is the source of truth for IT configurations and dependencies. When it is accurate, every automation and every operational decision benefits from it.