The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Only 28% of AI use cases at infrastructure and operations (I&O) organizations fully succeeded and met ROI expectations, while 20% failed outright, according to Gartner’s survey of 782 I&O leaders conducted in November and December 2025. The figures are not a universal failure rate: the remaining use cases were not classified in either category in Gartner’s summary. The pattern points to a practical problem more than a simple model problem: AI has to fit real operations, workflows, data, and budgets to produce measurable value.
Why doesn’t AI deliver ROI for IT departments?
Many AI initiatives begin with ambitious expectations—immediate cost reductions, automation of complex work, or a solution to long-standing operational issues—before teams have established whether the task, data, and surrounding processes are ready. Gartner found that 57% of surveyed I&O leaders reported at least one AI failure. Among leaders facing setbacks, 38% cited persistent skills gaps; 38% said poor data quality or limited data availability directly contributed to failure.
Scope and operational fit are also decisive. An AI feature that sits outside the systems employees use, or that does not match a stable and well-defined process, can add friction instead of removing it. Gartner’s Melanie Freeze put it plainly: “AI that doesn’t fit into the organization’s operations simply can’t deliver ROI.”
Recommended Free Tools
Integration and sponsorship turn capability into value
Gartner associates successful use cases with integration into existing systems and workflows, business-executive support, governance, and alignment with operational needs. Freeze said ROI depends less on model sophistication than on how well AI is “integrated, governed, and aligned with real operational needs.”
That distinction matters because a technically capable tool does not automatically change an operational result. If staff must switch systems, correct unreliable outputs, or create a new process around the tool, the organization may incur extra work without realizing the anticipated savings.
Which IT AI use cases appear more established?
Gartner points to generative AI in IT service management (ITSM) and cloud operations as areas where use cases are more mature and business value is more established. It reports that 53% of I&O leaders’ AI wins occur in ITSM. That is a survey finding about where respondents reported wins, not a guarantee that any particular ITSM deployment will pay off.
By contrast, respondents most often observed failures in auto-remediation, self-healing infrastructure, and AI agents managing complex workflows within or across systems. These tasks can be unpredictable and dependent on context, making them harder to automate safely and reliably.
Compare the work before choosing the use case
| Use case area | What the survey evidence indicates | Practical implication |
|---|---|---|
| Generative AI in ITSM | Gartner identifies ITSM as a current area of success; 53% of I&O leaders’ AI wins were reported there. | Consider bounded service workflows where the outcome and human review are clear. |
| Generative AI in cloud operations | Gartner describes cloud operations as an area with more established business value. | Assess whether the task fits existing cloud operations workflows and has a measurable operational outcome. |
| Auto-remediation and self-healing infrastructure | Gartner identifies these among common failure areas. | Be cautious with autonomous changes where errors can have broad or difficult-to-reverse effects. |
| Agents handling complex, cross-system workflows | Gartner identifies these among common failure areas, citing complexity and unpredictability as challenges for current AI tools. | Start with constrained tasks and defined oversight rather than assuming end-to-end autonomy. |
What broader CIO findings say about measurement and capacity
A separate signal comes from CIO.com/Foundry’s 2026 State of the CIO survey, which included 662 IT leaders and 249 line-of-business users. Only 19% of respondents said AI initiatives met or exceeded business goals. This is a broader CIO survey—not an I&O-only measure—and its result should not be combined with Gartner’s figures as if the populations and questions were identical.
Respondents also identified organizational barriers to scaling AI: 32% cited ill-defined ROI metrics and 40% cited a lack of in-house expertise. Fewer than half had formal AI success metrics. Among respondents who were measuring success, the leading measures were operational efficiency or process improvement (40%), employee productivity (34%), and cost reduction (30%). These are survey responses, not independently verified performance outcomes.
How should IT leaders measure AI ROI?
Define the business outcome and its baseline before expanding a pilot. The metric should match the job being changed: for example, incident resolution time for a support workflow, service quality for an end-user process, or cost for a clearly scoped operation. A deployment can improve one measure while worsening another, so track workload and operating expense alongside the headline benefit.
Rank #4
- Choose a bounded operational problem. Tie the proposed use case to a business or service goal, and specify what work the AI will and will not perform.
- Record the baseline. Measure the current process before deployment, including time, cost, service quality, and any relevant error or escalation rate.
- Count the full cost. Include integration, infrastructure, model use, training, output review, maintenance, and governance—not just the initial build or license.
- Name an accountable owner. Involve the operational team and relevant business, security, legal, and finance stakeholders so that success, risk, and ongoing support are not left ambiguous.
- Set decision thresholds before the pilot. Decide what result would justify expanding, changing, or stopping the work, then reassess as real usage and costs emerge.
Gartner recommends treating AI use cases as a managed product or portfolio, tying them to business goals, and evaluating feasibility, risk, cost, and expected impact with relevant stakeholders. Gartner’s Freeze says high-performing I&O leaders start with “realistic AI business cases and upfront preparation.”
Free tools Windows power users keep installed
One-click scans. No signup required.
Why positive ROI can coexist with more work
A SolarWinds ITSM survey, reported by ITPro in August 2026, offers a useful counterpoint: 84% of respondents said AI met or exceeded ROI expectations, yet 52% said their overall workload had increased and only 7% said adoption costs matched what they had planned. These are self-reported survey responses, not audited financial results, and the survey differs from Gartner’s in population and measurement.
The results can coexist because perceived return, staff workload, and total cost are different measures. Respondents reported time savings in issue detection, end-user requests, and ticket triage, while also describing work to maintain integrations, review outputs, train models, and manage reliability. As SolarWinds’ Brad McGinity told ITPro, “AI adoption is no longer the hard part — the hard part is building the organizational discipline to make AI actually deliver.”
A decision checklist before scaling an IT AI project
- Operational fit: Is the task predictable and bounded, or complex and likely to change with context?
- Workflow integration: Can the capability work in the systems and processes people already use?
- Business case: Is there a named owner, a baseline, and a measurable outcome tied to an operational need?
- Data and skills: Are the data available and reliable, and do staff have the capability to deploy and support the workflow?
- Governance: Are oversight, escalation, permissions, and responsibility for errors clear?
- Total cost and workload: Do expected gains outweigh implementation and ongoing costs, including the human effort needed to review and maintain the system?
A project that cannot answer these questions is not ready to scale simply because a pilot appears impressive. The relevant test is whether the organization can sustain a measurable operational improvement after integration, oversight, and ongoing costs are counted.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
