AI is beginning to reshape IT operations faster than many organizations can adapt. Vendors promise autonomous service desks, predictive outage prevention, and ticket volume reductions of up to 80%. A glimpse of this future is already visible in the capabilities that IT infrastructure managed services providers are beginning to implement. Analysts project that the AIOps market will reach nearly $38 billion by 2031, putting additional pressure on CIOs to act.
However, there is a significant gap between what AI vendors promise and what enterprises are achieving. Organizations that rush to deploy AI tools without a clear architecture often encounter agent sprawl, inadequate knowledge bases, and unverified vendor claims that undermine expected returns. Gartner estimates that deployments with weak knowledge bases can deliver less than half of their projected savings.
Organizations making meaningful progress share a common approach: they treat AI as a strategic sourcing and governance decision, not simply a technology investment.
This article examines what AI can deliver in IT operations today, where organizations encounter challenges, and how CIOs can capture value while managing risk.
What AI Is Actually Doing in IT Operations Today
AI capabilities are emerging across multiple layers of the IT operations lifecycle.
Monitoring, Detection, and Automation
AI-powered observability tools correlate signals across infrastructure, applications, security, and networks to identify anomalies before they develop into outages. As organizations build confidence in these capabilities, they can introduce automation to address recurring issues. Tasks that once required teams of engineers to investigate may increasingly be handled by models trained on historical incident data and organizational knowledge bases.
Service Desk and Resolution
The service desk is a common starting point for AI adoption, and an area where vendor claims can be particularly aggressive.
Virtual agents can handle password resets, account unlocks, software provisioning, and common troubleshooting workflows. Best-in-class deployments report ticket deflection rates of 55–65% without human involvement. Resolution rates vary by task complexity, reaching 75–90% for password resets but typically ranging from 25–45% for more complex application issues.
The AI Operations Center
Forward-looking organizations are moving beyond individual AI tools toward an integrated AI Operations Center (AIOC). This model coordinates AI agents across IT operations, including detection, correlation, escalation, resolution, release management, patching, asset management, security, and documentation.
An AIOC is not simply a product an organization purchases. It is an architecture that must be designed around its operational requirements.
Agentic AI represents another important development. Rather than assisting employees with individual tasks, agentic systems can take autonomous, multistep actions. This capability creates opportunities for greater efficiency, but also increases the importance of governance, oversight, and accountability.
Services with strong potential for AI automation include Level 0 and Level 1 service desk activities, incident correlation, root cause analysis, change risk scoring, configuration drift detection, and post-incident reporting. Activities requiring human judgment, client relationships, or regulatory accountability will continue to need human involvement.
The Business Case for AI in IT Operations
The potential return on investment from mature AI deployments is compelling, although results vary significantly depending on the organization’s starting point and implementation approach.
| Metric | Typical Range | Top-Quartile |
| Ticket deflection rate | 40–60% | 55–65% |
| Cost per Tier 1 ticket (human) | ~$22.50 | — |
| Cost per AI-resolved ticket | 1.80–4.50 | — |
| Reduction in cost per ticket | 38–52% | 52%+ |
| MTTR reduction | 35% | 52% |
| Password reset time | 2.5–4 hours to 3–8 minutes | 97% faster |
| Time to positive ROI | 9–18 months | 6 months |
| Three-year ROI | 150–250% | 250%+ |
Sources: HDI 2024; Forrester composite (15,000-employee enterprise); IDC top-quartile deployments.
For an enterprise handling more than 5,000 tickets per month, Deloitte estimates that mature AI deployment can generate $1–1.5 million in annual savings through ticket deflection. Forrester’s composite case study reports a three-year net present value of $4.2 million, with payback in 6–18 months.
However, these figures assume a strong knowledge base and clean, accessible data. Gartner’s analysis indicates that organizations with poor knowledge management can underdeliver on projected AI savings by at least half.
The technology may be capable of delivering these results, but the data, processes, and governance behind it determine whether organizations can realize that value.
The Hidden Risks of AI Adoption in IT Operations
The business case for AI is real, but so are the risks. Several recurring challenges can prevent organizations from achieving their expected returns.
Vendor Hype vs. Actual Performance
Vendors routinely claim autonomous resolution rates of 70–85%, while actual enterprise deployments typically achieve 40–60%. The gap often comes down to the work required before implementation: clean knowledge bases, well-structured workflows, and effective change management.
Organizations that overlook these prerequisites risk paying for capabilities they are not prepared to use.
Agent Sprawl
As organizations adopt individual AI tools, they can quickly accumulate overlapping capabilities. A virtual agent here, an AIOps platform there, and a generative AI copilot for developers can create a fragmented environment without clear ownership or governance.
Who manages each agent? What happens when two agents provide conflicting guidance? Who reviews decisions made autonomously?
Without clear answers, agent sprawl becomes an operational and governance challenge.
The Build, Buy, Hire, or Hybrid Decision
Many organizations default to purchasing a vendor platform without first evaluating whether they have the talent to implement it, the data to support it, or the contractual protections to manage it.
The decision to build internally, buy a solution, hire specialized talent, or use a hybrid approach requires careful consideration. A managed service provider that embeds AI into its delivery model represents a different type of relationship than one that does not. Yet many existing IT contracts were written before these distinctions became relevant.
Managed Services Displacement Risk
If an existing managed service provider uses AI to reduce its delivery costs under a fixed-fee contract, the client may not receive any of those efficiency gains.
If the provider uses AI without the client’s knowledge, additional security and governance concerns may arise.
Either way, existing sourcing agreements may not reflect the changing economics and responsibilities of AI-enabled service delivery.
Overpromising and Under-delivering
When vendors and internal AI champions set unrealistic expectations, organizations risk stakeholder fatigue and abandoned initiatives when results fall short.
Sustained investment requires realistic baselines, consistent measurement, and a willingness to let demonstrated performance guide future decisions.
A Smarter Path Forward: The Windsor Group Framework
For more than 40 years, Windsor Group has helped enterprise IT organizations navigate complex sourcing decisions and transitions. AI in IT operations represents a significant shift in how technology services are delivered and managed. We bring the same disciplined approach to this opportunity: structured assessment, clear architecture, and accountable execution.
Our framework includes three phases.
1. Assess: Understand Your Environment Before You Implement
Before evaluating AI vendors or tools, organizations need an objective baseline.
Key questions include:
- Where are the highest-volume, lowest-complexity transactions?
- What is the current cost per ticket across service tiers?
- Is the existing knowledge base (and CMDB) accurate, complete, and suitable for AI consumption?
- What are incumbent managed service providers contractually required to deliver, and how does actual performance compare?
This assessment identifies both opportunities and potential obstacles. Typically completed in 4–8 weeks, it establishes financial and operational objectives to guide the strategy and decision-making process.
The Windsor Group Assessment also evaluates multiple provider solutions, giving clients an informed view of what AI can make possible in their environment. This helps clients identify which vendors they want to evaluate more closely in the next phase.
2. Architect: Design Around Outcomes, Not Features
The architecture phase determines which AI capabilities belong in each layer of the IT operations model, which are best delivered by a managed service provider, and which the organization should own directly.
It answers the Build, Buy, Hire, or Hybrid question with a focus on specific business and operational requirements.
It also establishes the governance framework before AI agents are deployed, including how decisions will be monitored, overridden, and audited.
Windsor Group develops the RFP and manages the competitive sourcing process in collaboration with the client’s Procurement team, helping the organization evaluate providers and select the right partner.
3. Activate: Implement With Accountability
Activation is where many AI programs encounter difficulties. Selecting a provider is only the beginning. Organizations also need appropriate contracts, governance, and performance measures to ensure the engagement delivers its intended value.
Windsor Group supports clients through vendor selection, contract renegotiation to capture AI-driven efficiency gains from existing managed service providers, and program governance to keep performance aligned with business objectives.
We are an independent advisor, not an implementation partner. Our role is to help clients make informed decisions and hold their implementation partners accountable for the results they promise.
What AI Means for Your IT Contracts and Sourcing Strategy
For organizations already working with managed services providers, AI is changing both service delivery and the economics of those relationships.
If you have not reviewed your IT outsourcing contracts through an AI lens in the last 12 months, now is a good time to start.
Your sourcing strategy should address several important questions:
- Are your managed service providers using AI in their delivery models? If so, are you receiving the resulting efficiency gains through improved pricing or service performance?
- Do your SLAs reflect AI-enabled performance? Resolution time expectations established before AI became viable may no longer reflect what your providers can deliver.
- Do your contracts address AI governance? What data can providers use to train or operate AI models? Who owns the outputs? What liability provisions apply if an AI-driven decision causes an incident?
- Are you paying for AI capabilities your organization cannot use? Some organizations pay for AI-enhanced managed services without validating whether those capabilities are actually available or delivering value.
- What protections do you have if an AI vendor underperforms? AI tools embedded in operational workflows can create unexpected vendor dependencies without appropriate contractual safeguards.
- What happens at the end of the engagement? Do you receive a perpetual license, ongoing support, and/or training on the AI environment?
The managed services market is undergoing a structural shift as AI changes delivery models and operating costs. Organizations that respond proactively with clear performance metrics, well-structured contracts, and informed sourcing decisions will be better positioned to capture the benefits.
About Windsor Group Sourcing Advisory
Windsor Group is an independent IT sourcing advisory firm with more than 40 years of experience helping enterprise IT organizations navigate complex technology and vendor decisions. We have worked alongside technology leaders at organizations including Aflac, Eversource Energy, Cleveland Clinic, and Intermountain Health.
We do not sell technology. Our focus is helping clients make well-informed sourcing decisions and hold their providers accountable.
If you are in an existing Managed Services engagement, or AI in IT operations is on your agenda for the next 12–24 months, we welcome the opportunity to talk. Our engagement model begins with a rapid AI Readiness Assessment, typically completed in 4–8 weeks. It provides an objective view of where AI can deliver value in your environment, where challenges may arise, and what steps to consider next.
Ready to evaluate your AI opportunities? Schedule a 30-minute introductory conversation with Windsor Group.