AI Seer: enterprise adoption stalls
- Deloitte’s January 21, 2026 enterprise AI report found companies are still stuck in pilot mode, with only 25% moving 40%-plus of pilots into production. - McKinsey’s 2025 survey showed 88% of companies use AI in at least one function, but nearly two-thirds have not scaled it enterprise-wide. - Microsoft, IBM, and AWS say inference and token bills are becoming the bottleneck as agentic workloads push AI spending from pilots to cost control. (microsoft.com)
Enterprise AI use is broad in 2026, but most companies still have not pushed those systems into large-scale production. (deloitte.com) (mckinsey.com) Deloitte said on January 21 that only 25% of surveyed organizations had moved 40% or more of their AI pilots into production. Its survey covered more than 3,000 director- to C-suite-level leaders involved in AI programs. (deloitte.com) McKinsey’s November 5, 2025 survey found 88% of respondents said their organizations regularly use AI in at least one business function. Nearly two-thirds said they had not yet begun scaling AI across the enterprise. (mckinsey.com) The bottleneck is no longer just building a model. It is paying to run that model every time an employee, customer, or software agent sends a request. (microsoft.com) (ibm.com) Microsoft said last month that inference is what drives cloud spend when companies run millions of daily requests across copilots, analytics pipelines, and agentic workflows. The company described production inference as a capital-allocation problem, not just an engineering task. (microsoft.com) IBM said executives expect average computing costs to climb 89% between 2023 and 2025, and 70% cited generative AI as a critical driver. IBM also said every executive in its survey had canceled or postponed at least one generative AI initiative because of cost. (ibm.com) Amazon Web Services has started publishing cost-control playbooks for Bedrock customers, including token limits, budget alarms, and circuit breakers to stop runaway usage. AWS said token-based pricing can produce unexpected bills if usage is not tightly monitored. (aws.amazon.com) That cost pressure is rising as companies experiment with AI agents, which can call models repeatedly inside a single workflow. Deloitte said 85% of companies expect to customize agents for their own business, while McKinsey found 62% are already at least experimenting with them. (deloitte.com) (mckinsey.com) The spending is still growing fast. Menlo Ventures estimated enterprise generative AI spend reached $37 billion in 2025, up from $11.5 billion in 2024, with more than half going to applications rather than base infrastructure. (menlovc.com) The picture in 2026 is not that companies stopped buying AI. It is that they are moving from buying access to proving that each prompt, workflow, and agent can earn its keep. (deloitte.com) (aws.amazon.com)
