SolarWinds research finds 84% of IT professionals say AI has met or exceeded ROI expectations, yet only 7% say adoption costs matched their forecasts
AI may be delivering on its promise in IT service management, but SolarWinds’ latest research suggests that the economics of adoption are proving more complicated than the headline productivity gains suggest.
The company’s 2026 State of ITSM Report, based on a survey of more than 800 IT professionals globally, finds that 84% of respondents say AI has met or exceeded their ROI expectations. At the same time, 52% say their overall workload has increased since adopting AI, while only 7% say the actual cost of AI adoption matched what they had planned.
After an average of roughly 16 months using AI in their ITSM environments, organisations appear to be discovering that the technology is not simply eliminating work. Instead, it is shifting where that work happens and creating a new layer of operational overhead.
The productivity paradox
The productivity gains are tangible. Respondents say AI saves an average of 3.2 hours a week on detecting and flagging issues, 3.0 hours on end-user requests and 2.9 hours on ticket triage.
But much of that time appears to be returning in another form.
Nearly half of respondents, 48%, say they spend time managing and maintaining AI tools and integrations, while 47% are reviewing and validating AI-generated outputs. Another 37% are involved in training and fine-tuning AI models.
The costs associated with these activities are also proving difficult to capture in initial budgets. Staff training was identified as a surprise expense by 48% of respondents, followed by data quality and cleanup at 47%, and ongoing tuning and maintenance at 45%.
These are not necessarily one-off implementation costs. More than four in five respondents, or 83%, now spend at least three hours a week keeping their AI systems operating reliably.
The result is a distinction between productivity and workload. AI can make individual processes faster without necessarily making the overall IT operation lighter.
AI is still responding to problems
The research also points to another limitation in current ITSM deployments. AI is being used primarily to respond to incidents rather than prevent them.
The two areas where respondents reported the greatest impact were identifying issues before they affect users, at 31%, and prioritising and routing issues, at 23%. Both represent improvements in responding to problems that have already emerged.
Only 19% identified preventing issues before they occur as the area where AI has had the greatest impact.
That gap suggests that AI adoption and AI maturity are not necessarily moving at the same pace. Organisations may have deployed AI into service management workflows without yet having the data foundations, infrastructure or operational processes required to push those systems towards continuous prevention.
There are signs, however, that spending is moving in that direction. Eighty-five percent of respondents say their AI-in-ITSM budgets have increased year over year, including 36% who report a significant increase. Agentic workflows, meanwhile, are expected to see the highest investment growth among the AI capability categories covered by the survey.
The economics are becoming harder to ignore
The findings suggest that the next phase of enterprise AI adoption may be less about proving that AI can save time and more about understanding the total operating model required to sustain it.
Only 21% of respondents measure AI through outcome or experience-based metrics. Organisations that measure AI primarily through activity rather than outcomes are 2.4 times more likely to report that their workload has increased after adoption.
That distinction matters because an organisation can improve the number of tickets processed or incidents flagged without necessarily improving the broader service experience or reducing the cost of operating IT.
Data quality is another fault line. Cleanup and data preparation rank among the leading unexpected costs, suggesting that the value of AI in ITSM remains closely tied to the quality of the operational data feeding it.
For IT leaders, the implication is less about slowing AI investment than being more selective about where it is deployed. High-volume, clearly defined workflows such as ticket triage, issue detection and incident documentation offer relatively straightforward opportunities to establish measurable gains.
The research also points towards consolidation. Integrating AI into existing service workflows rather than adding multiple disconnected tools can reduce the maintenance burden created by the technology itself.
People remain another part of the equation. Eighty-two percent of organisations provide formal AI training and structured change management, while 66% of respondents say bonuses and performance reviews are tied to AI efficiency gains.
“We’re at an inflection point in IT service management. AI adoption is no longer the hard part — the hard part is building the organisational discipline to make AI actually deliver,” said Brad McGinity, GM of ITSM, SolarWinds.
The broader lesson from the research is that AI’s ITSM payoff cannot be measured simply by how much work a model can automate. The more consequential question is whether the time saved is greater than the new work required to govern, maintain, validate and improve the systems doing that automation.
For IT organisations moving deeper into AI, that may become the real ROI test.