Enterprises Face AI Evaluation Gap Amid Shift to Smarter Systems
WHO:
CNBC - Aravind Srinivas, CEO of Perplexity
VentureBeat - 573 technical leaders at companies with 100 or more employees
VentureBeat - 157 qualified enterprise respondents at companies with 100 or more employees
WHAT:
CNBC - "The model alone is no longer the product," said Perplexity CEO Aravind Srinivas, emphasizing the importance of the orchestration system that integrates the model with various tools.
VentureBeat - 66% of enterprises permit some production deployment of AI agents without human review, and only 5% fully trust the automated evaluations that would make those release decisions.
VentureBeat - 51% of enterprises expect their primary control plane for enterprise agents to be hybrid by the end of 2026, up from 34% in the previous survey.
VentureBeat - 32% of enterprises are considering non-Nvidia accelerators as an emerging compute option in the next 12 months.
VentureBeat - 25% of enterprises already run a governed semantic layer in production, while 34% are still building one, and 41% have not started.
CNBC - The AI race is shifting from focusing on larger models to cheaper, smarter systems.
VentureBeat - 86% of enterprises that run their own GPUs report utilization of 50% or less.
VentureBeat - 54% of companies had an agent security incident or near-miss in the past 12 months.
VentureBeat - 71% of enterprises say a quarter or fewer of their deployed agents can complete multi-step tasks on their own.
VentureBeat - 66% of enterprises allow AI agents to push code or system changes to production based on automated evaluations alone.
VentureBeat - 69% of companies allow agent credential sharing somewhere in their agent fleet during runtime.
VentureBeat - 57% of enterprises traced at least one confident, wrong agent answer to missing or inconsistent business context.
WHERE:
VentureBeat - The survey was conducted across five parallel surveys of the agentic stack.
WHEN:
VentureBeat - The survey findings are based on data collected in June 2026.
VentureBeat - The full Q2 reports across all five VB Pulse trackers will be released at VB Transform, July 14–15, 2026.
WHY:
CNBC - The shift in AI competition is due to companies moving from testing AI to using it in real products and workflows.
VentureBeat - Enterprises are retrofitting to catch up with their own standards due to the rapid deployment of AI agents without adequate controls.
HOW:
CNBC - AI products are becoming systems that can decide which model to use, when to use it, and what outside tools or company data sources are necessary.
VentureBeat - Enterprises are budgeting for changes, with roughly six in 10 planning to switch or add vendors in each of five control layers within the next 12 months.
(UPDATED July 11, 2026 10:20 UTC)
In a significant shift in the tech landscape, Perplexity CEO Aravind Srinivas highlighted that "the model alone is no longer the product," stressing the need for an orchestration system that combines the model with various tools. A recent survey revealed that 66% of enterprises allow some deployment of AI agents without human review, yet only 5% fully trust the automated evaluations guiding those decisions. Looking ahead, 51% of enterprises anticipate their primary control plane for enterprise agents will be hybrid by the end of 2026, a notable increase from 34% in the previous survey.
--
(UPDATED July 11, 2026 07:38 UTC)
The AI race is evolving as companies shift their focus from larger models to more affordable and smarter systems, according to Aravind Srinivas, CEO of Perplexity. A recent survey revealed that 86% of enterprises running their own GPUs report utilization of 50% or less, while 54% experienced an agent security incident or near-miss in the past year. The full Q2 reports from five VB Pulse trackers will be unveiled at VB Transform on July 14–15, 2026, highlighting how enterprises are adapting to the rapid deployment of AI agents without adequate controls.