AI will transform entire sectors. But before organisations can capture its full value, they need to confront a more immediate challenge: control.
Findings from the Flexera 2026 State of ITAM report show that, after years of steady progress, IT estate visibility and cost control are stalling for many businesses. Fewer respondents than last year report that they have complete IT asset visibility (36%) and cost pressure is making software spend optimisation a top priority by a record margin (19%).
These trends reflect the unique challenges in controlling and funding the rapid emergence of AI across business. Only 31% of organisations report having visibility into AI tools, while 59% say wasted AI spend has increased year over year.
This should be concerning for IT and business leaders. AI is not just another category of software to track. Its unique characteristics, massive upside potential and significant risks mean it warrants greater governance than previous generations of technology. If organisations treat it like a conventional SaaS rollout, they risk repeating the mistakes of cloud and SaaS sprawl, only faster, with less visibility and higher stakes.
There are several reasons why the current surge in AI adoption is difficult to manage.
The first is speed. The pace of AI innovation is unlike previous technology cycles. New models, tools, assistants, and agents are appearing every few weeks, each promising better performance or broader functionality. Business teams are understandably eager to experiment. Individuals want productivity gains. Departments want faster workflows. Leaders want to show progress. That enthusiasm is valuable, but it also creates the conditions for shadow AI.
Unlike earlier software adoption, AI often enters the organisation through many doors at once. It may appear as a standalone application, an embedded feature in an existing SaaS platform, an API call in a cloud environment, a chatbot subscription, or a custom agent built by a team outside central IT. Each instance may seem small on its own. Together, they create a fast-growing layer of cost, risk, and operational dependency that many organisations cannot fully see.
The second challenge is that AI behaves differently from traditional cloud and SaaS applications. Not only is spending spread across tools, APIs, models, and cloud environments; it is also irregular and behaviour driven. Usage can rise suddenly when a workflow is automated or when users discover a new use case. Without the right controls, costs can increase before finance, procurement, or ITAM teams know what is happening.
AI also makes value harder to predict. Traditional automation was often judged by speed, consistency, and cost. Generative AI is more variable. Output quality depends on the model selected, the quality of the data, the design of the agent, the prompts used, and the skills of the people interacting with it. If those elements are weak, organisations may pay for activity without getting reliable outcomes. In practical terms, that means wasted tokens, duplicated tools, unused licences, poor adoption, and a growing gap between AI investment and business value.
Organisations do not need to stop experimenting with AI. They need to adapt their approach to manage it as part of the broader IT estate and surface measurable outcomes.
That starts with bringing ITAM and FinOps closer together. ITAM brings the discipline of inventory, ownership, entitlement, contract management, compliance, and lifecycle control. FinOps brings consumption visibility, cost allocation, forecasting, optimisation, and accountability for cloud-based spend. AI needs both. A licenced AI tool may look like SaaS. A model running in the cloud may look like consumption spend. An agent may combine software, cloud, data, and workflow automation. Managing each part separately will not be enough.
A more effective approach is to create a shared view of AI assets and usage. Organisations need to know which AI tools are approved, who owns them, who uses them, what data they touch, how they are funded, and what value they are expected to deliver. They need accurate, connected, and current information about the IT estate. That includes application ownership, contract terms, user access, cloud resources, SaaS usage, business criticality, and risk classification. In combination with commercial data, this will enable a comprehensive view of whether adoption, cost, and outcomes are moving in the right direction.
With the right visibility in place, organisations must prioritise understanding where AI generates measurable business value. This is achieved with the following steps:
It’s all well and good having the infrastructure and processes in place to govern AI, but without strong implementation, value for money will remain elusive.
Users across the business need to have the skills to select the right model, design effective prompts, validate outputs, monitor usage, and manage agent lifecycles. Without these fundamental competencies, users will burn through tokens while generating inferior outcomes and, potentially, risks. Organisations with an effective data strategy will also yield superior results, as without a strong data foundation, the hit rate for even perfectly crafted prompts will be lower.
Agentic AI raises the stakes further. Agents have the potential to automate complex workflows, but they also introduce new questions about ownership, access, testing, cost, failure modes, and accountability. Before scaling agentic use cases, organisations should define clear guardrails: what the agent is allowed to do, which systems it can access, what human approvals are required, how performance is measured, and how costs are monitored. The goal is not to block high-value use cases. It is to make sure they are safe, measurable, and commercially viable from the start.
For many organisations, expert support will accelerate this journey. The challenge is not simply choosing a tool or writing a policy. It is connecting discovery, governance, cost optimisation, licencing, cloud consumption, security, and business value into one operating model. That requires experience across software, cloud, data, and AI.
AI will reward organisations that move quickly, but not if they move blindly. The businesses that get the most value from AI will be those that can see what they use, understand what it costs, manage the risks, and reinvest savings into the use cases that matter most. That starts with closing the visibility gaps, connecting spend to value and ensuring you have the data and skills to create value with minimal waste.
SoftwareOne helps organisations bring clarity, control and business impact to complex software, cloud, and AI environments. From ITAM and FinOps, to AI deployment and governance, our experts can help you identify waste, reduce risk, and build the foundation for smarter AI investment. Talk to us about building a practical roadmap for AI visibility, governance, and optimisation.
Discover the trends that are shaping IT cost and risk management today by downloading the full report from Flexera.
Discover the trends that are shaping IT cost and risk management today by downloading the full report from Flexera.
