TL;DR — Key Takeaways
- Nearly half of enterprises surveyed by Futurum Group say they are running over budget on AI spending, underscoring the difficulty of predicting deployment costs.
- Organizations are often absorbing AI cost overruns by cutting elsewhere in IT, particularly contractors, support operations and application development.
- Few organizations are responding to overruns by pausing AI initiatives, suggesting enterprises still view AI investment as a strategic priority.
A survey of 1,636 global enterprise technology decision makers conducted by the Futurum Group finds nearly half (47%) lead organizations that are over budget on their artificial intelligence (AI) spending, compared to 32% that report spending approximately in line with the plan and 6% that are spending below plan. A total of 10% said their organization has no formal AI budget.
Among those respondents that work for organizations that are over budget, 48% ask for more budget, while 43% absorb the overrun and settle it later. Only 17.2% pause or reduce the AI initiative itself.
When the gap is funded by reallocating within IT, external labor is cut first: 61% said they help close that gap by reducing external contractors and consultants. Additionally, 60% make cuts to IT support and helpdesk, while 58% cut software and application development. Only 23% of over-plan are reallocating funds from a non-IT business unit budget.
The survey suggests that CIOs are the fastest growing segments of their budgets, says Mitch Ashley, vice president and practice lead for CIO and technology buyers at Futurum Group. “Only a quarter are able to cover the gap from a business unit budget,” says Ashley.
It’s not yet clear how budget dollars for AI might be allocated in 2027 now that organizations have a better understanding of actual costs. There are, however, multiple approaches that range from building and deploying custom open weight AI models in an on-premises IT environment that might enable organization to deploy more AI applications without breaking the bank.
Alternatively, organizations will need to prioritize a narrow range of projects that provide the most value to the business. Regardless of approach, tradeoffs will need to be made as AI services become more expensive with each advance. The simple fact of the matter is that the providers of these AI services are not going to be able to indefinitely subsidize the cost of building and deploying advanced AI models.
Organizations are also now starting to grapple with the security implications of AI, especially as AI agents are deployed. Most organizations in addition to applying guardrails to those AI agents will need to apply more granular controls across the data that AI agents access. IT leaders should also assume that as AI becomes more regulated there will be an increase in compliance costs.
CIOs will also need to work with business leaders to determine how much intellectual property to share with those AI models. While the providers of AI services routinely note they are not using customer data to train their models, the models they build are exposed to metadata and the underlying code that drive specific workflows. Those insights could potentially be used to enable the providers of an AI service to add an application to their portfolio that, in effect, turns them into a competitor of an existing customer.
At this juncture, there is no putting the proverbial AI genie back in the bottle. The challenge and the opportunity now is reimagining the role IT plays in an organization as AI continues to rapidly evolve in ways that today are difficult to not just predict but also anticipate.
