Enterprise organisations have deployed AI agents at scale whilst deliberately operating without adequate governance frameworks to manage them. VentureBeat Research's survey of 573 technical leaders at companies with 100+ employees reveals the stark reality: 86% report GPU utilisation at 50% or below, indicating massive infrastructure investment paired with underutilised capacity. This isn't accidental inefficiency—it reflects a deliberate choice to move fast and build controls later. The gap between deployment velocity and governance maturity creates a critical vulnerability. For CX teams already running Salesforce Agentforce or similar autonomous systems, this raises an uncomfortable question: are your agent interactions being monitored and validated with the same rigour as your human agent interactions, or are you operating in the same evaluation gap that's characterising the broader enterprise AI landscape?
The implications for customer experience operations are twofold. First, the infrastructure overspend signals that many organisations are treating AI buildout as a capital expenditure problem rather than an operational efficiency problem—they're buying capacity without optimising utilisation, which suggests immature deployment strategies. Second, and more critically, the conscious decision to deploy ahead of controls means that customer-facing AI agents in support, sales, and service environments are operating with minimal oversight. This creates compounding risk: as agents gain autonomy faster than companies can verify them, the likelihood of brand damage, compliance violations, or customer dissatisfaction increases proportionally. For Zendesk and Freshdesk administrators managing high-volume support queues, the question becomes whether your organisation's AI governance keeps pace with agent decision-making authority, or whether you're inadvertently running customer interactions through unvetted systems.
The broader market signal is that enterprises are willing to absorb significant waste—both in capital and operational risk—to avoid being left behind in the AI race. This creates opportunity for vendors and consultants who can help CX teams retrofit governance into existing deployments, but it also exposes a dangerous assumption: that customer experience can tolerate the same evaluation gaps that technical infrastructure can. It cannot. The 86% underutilisation figure should prompt CX leaders to ask not just whether their AI investments are delivering ROI, but whether they're delivering safety and consistency at the customer touchpoint.
Enterprise companies are running AI agents ahead of the controls needed to manage them — and they deployed that way knowingly. That is the central finding from VentureBeat Research's June survey of 573 technical leaders at companies with 100 or more employees, fielded across five parallel surve