Gartner says AI budgets are growing faster than the rules to control them

Gartner’s newest batch of 2026 AI research puts a number on what many enterprise operators are feeling in budget meetings: the spend is arriving faster than the controls. In a Aug. 26 press release, Gartner said AI spending by customer service leaders surged 38% even as overall service and support function budgets rose just 2%.

That gap is the story. When AI grows like a carve-out inside a mostly flat cost center, the question stops being “should we pilot” and becomes “what do we stop paying for, and how do we keep the new bill from ballooning?”

The cost problem is less about AI, more about accounting for AI

Since early 2026, Gartner has cautioned that many enterprises are deploying AI without strong controls on spending. In a March 9 conference update from the Gartner Data & Analytics Summit in Orlando, the firm reported that more than half of IT leaders worry about cost overruns, while just 44% of organizations have put financial guardrails in place or adopted AI FinOps practices.

Those two facts, the 38% spend surge in service and the 44% guardrail adoption rate, fit together in an uncomfortable way. If AI is shifting from experiments to production workloads in customer operations, cost visibility and cost responsibility have to move with it. Otherwise, genAI becomes the new shadow IT, except the meter is running in tokens and GPU hours instead of SaaS seats.

Customer service is becoming a frontline AI budget center

Gartner’s Aug. 26 survey framing is notable because it ties AI growth to a specific function with a hard budget baseline: service and support rose 2%, while AI within that envelope rose 38%. For contact center and customer service operations leaders, that’s a public benchmark for internal planning. AI is no longer an “innovation” line item. It’s competing directly with staffing, knowledge management, WFM upgrades, QA programs, and outsourcing.

Gartner’s newsroom also includes an August 2026 Q&A that highlights another operating limit: how much automation customers will accept. Gartner reported Aug. 4 that 87% of customers say companies that use genAI in customer service must still offer access to a human agent. That steers teams toward hybrid setups, where escalation routes, agent assist, and knowledge workflows matter as much as the model choice.

Hospitals show where this goes next: inventory without counting

Healthcare supply chain is another place Gartner is explicitly calling out process change. In an Aug. 26 press release, Gartner said AI will soon make hospital inventory counting obsolete. Even without the implementation details in the teaser, the implication is clear for materials management teams: AI is being positioned as a replacement for periodic cycle counts and manual reconciliations, which means new data dependencies, new audit routines, and new integrations into ERP, EHR, and item master governance.

If that prediction holds, it also changes procurement conversations. Hospitals that historically bought RFID, scanning, or smart cabinet solutions to improve count accuracy may start asking vendors a different question: how quickly can the system reach “trusted inventory” without dedicated labor, and what data does it need to get there?

Security spend is being pulled along behind it

Gartner’s August news cycle also points to governance and security becoming separate line items in procurement. On Aug. 26, Gartner forecast that spending on securing AI will hit $4.8 billion in 2027. For CIOs and security leaders, the operational message is that AI security tools are moving out of “features in existing platforms” and into standalone buys, with distinct budget owners and evaluation criteria.

That matters because many teams are still trying to govern genAI through policies alone. A growing third-party market suggests enterprises are expecting to instrument AI systems with monitoring, access controls, model governance, and data protections that look more like traditional security programs: scoped requirements, controls mapping, and vendor risk reviews.

Where this lands in 2027 planning for CIOs and ops leaders

• Treat AI consumption like a utility, not a project: ask internal platform teams to show token usage, model mix, and unit cost by workflow, then decide which workflows get “premium” models versus cheaper defaults. Gartner’s March 2026 note that only 44% have guardrails is a warning sign for scale-ups.
• Rewrite SOWs for customer service AI to include human escalation and QA: Gartner reported 87% of customers still expect access to a human agent. Make “handoff time,” “agent assist adoption,” and “deflection with satisfaction” contract metrics, not slideware.
• For hospital and clinical supply chain projects, pressure-test the data layer first: Gartner’s “inventory counting obsolete” claim only works if item masters, location hierarchies, and transaction capture are dependable enough to trust AI outputs.
• Expect a separate line item for AI security: Gartner’s $4.8B 2027 forecast for securing AI suggests dedicated tooling and services are becoming standard. Decide whether the buyer is the CISO org, platform engineering, or a joint steering committee before renewal season.

One near-term signal worth watching is whether enterprises start reporting AI cost controls as explicitly as they report AI pilots. Gartner’s Business Wire notice about its Aug. 4, 2026 earnings call is aimed at investors, but it also hints at where operators will get more benchmarks: when large research and advisory firms and their customers begin treating AI governance maturity as a measurable operating metric, not a qualitative aspiration.