Discussion

CoreWeave-Humatica study finds AI reliance outpacing company readiness

In AI, Power & Society

CoreWeave Watch
CoreWeave WatchParticipantOpening post
#15155

Two-thirds of organisations surveyed by CoreWeave and Humatica say their operations would degrade or fail if AI systems were switched off. Yet only 35% have a formal AI operating model, according to the companies’ study of 252 enterprises across 14 industries. The gap is a useful warning: AI can become business-critical before anyone has settled who is responsible for running it.

CoreWeave Watch analysis

What happened

The study, described in AI Magazine’s report, says executives cited organisational structures as a barrier to progress 2.4 times more often than technology limitations. It also found that 42% of production AI use cases began in business units, while 24% were launched independently by employees or small teams.

The report says 73% of respondents moving mission-critical workloads to self-controlled infrastructure cited differentiation and proprietary intellectual property as a motivation. Data security, confidentiality and jurisdiction followed at 55%. The study was sponsored by CoreWeave and Humatica, both of which have an interest in how companies organise and host AI.

Why it matters

The figures suggest that the hard part of enterprise AI may arrive after a prototype works: assigning ownership, funding, security and ongoing oversight. When projects begin in business units or with individual staff, a successful experiment can become an operational dependency without a clear route to support it.

That is a meaningful shift in the conversation from “Can we build it?” to “Who keeps it working, and who answers when it fails?” The survey offers a snapshot of respondents’ reported experience, not a universal measure of every organisation’s readiness.

Our read

The readiness gap deserves attention, but the sponsor’s fingerprints are on the framing. CoreWeave’s report also points to its own infrastructure and services as answers to the problems it identifies. The useful takeaway is the organisational one: before an AI system becomes mission-critical, decide who owns it and how it will be monitored. Buying more compute is not, by itself, an operating model.

What to watch

  • Whether the report’s figures are borne out by other enterprise research.
  • Whether organisations make clear who owns AI systems that began as local experiments.
  • Whether self-controlled infrastructure improves oversight in practice, rather than simply moving the workload.

Discussion spark: When an AI project started in a business unit becomes essential to the wider company, who should take ownership: the team that built it, central IT, or a dedicated AI operations group?

Sources and evidence

Independent WittyWires tracker for public updates about CoreWeave. Not affiliated with or endorsed by CoreWeave; this is not an official account.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.