95% of AI Pilots Fail, Sam’s Club Quiet AI Rollout, Simulated Students Crush Exams

A peer-reviewed study prompted 11 language models to answer 489 NAEP math and reading questions as if they were 4th, 8th, or 12th graders. Across all grades the simulated students outscored real test-takers and missed questions for reasons unrelated to typical human misconceptions.

Without grade prompts, LLMs posted reading scores 33 to 40 percentile points above average students, and no prompt fully aligned AI answers with human performance. Even when instructed to mimic specific grades, the models failed to replicate common errors such as misusing the order of operations.

The study lands as 60 % of K-12 teachers report classroom use of AI, including quiz generation and tutoring support. Researchers caution that the accuracy gap and atypical error patterns could mislead educators who depend on AI simulations to tailor lessons.

A viral MIT study reports that 95% of corporate AI projects collapse, a pattern echoed by former consultant Marc Malott whose own AI rollout stalled after early gains. He says the push to capture quick returns derailed momentum and choked off further innovation.

A McKinsey survey shows 80% of executives see no earnings lift from generative AI, while an Upwork study finds top AI users are twice as likely to quit and 88% feel burned out. Malott ties both findings to organizations squeezing for immediate payback, which erodes trust and kills discretionary effort.

Malott highlights SharkNinja, Johnson Hana, and Shopify as rare firms that embed AI by redesigning incentives around collaboration and decentralized decision-making. He maintains that sustainable ROI surfaces only when leaders reward shared experimentation instead of harvesting one-off efficiency bumps.

The borough of Westwood, New Jersey is now using Vialytics’ AI road-management app to decide when streets need repair or repaving. The software analyzes images captured by the Public Works superintendent’s phone as he drives the routes.

The system records potholes, faded striping and 13 other defects, rates overall road quality, and can generate work orders immediately. Westwood is one of 40 New Jersey municipalities deploying the locally developed tool.

Officials say the app will recoup its cost by trimming manpower and giving leaders hard data for budget decisions. The data-driven scores also let the superintendent rebut resident paving requests with objective evidence.

Sam’s Club is rolling out generative AI tools directly to its frontline managers. The Walmart-owned chain says the system speeds decision-making and removes repetitive work from the sales floor.

Managers already use the software to compress financial analysis from hours to minutes. Beginning next year, staff can enroll in OpenAI’s certification program through Walmart Academy to deepen AI skills.

The rollout makes Sam’s Club one of the first U.S. retailers to arm store managers with generative AI. Walmart U.S. CEO John Furner says retail’s future hinges on employees who can wield AI, spotlighting upskilling as a strategic priority.

The Royal Ballet and Opera in London will launch RBO/Shift, a yearly technology festival probing artificial intelligence’s role in opera. Its inaugural edition runs June 4–7 and blends a symposium with live performances, tech demos, and artist discussions.

Associate director Netia Jones is curating the event, while director Oliver Mears frames AI as progress the art form cannot ignore. Both cite opera’s history of adopting stage innovations—from 18th-century flying dragons to modern VR—to position the festival as the genre’s next evolution.

By dedicating four days to AI, one of the world’s premier opera houses places the technology firmly in the cultural mainstream. The program also surfaces industry tensions, as performers like mezzo-soprano Katia Ledoux warn that AI-generated visuals undercut creative jobs and misrepresent artists’ work.