Insights

What satellite imagery, weather data and AI can forecast about crops, from a peer-reviewed rice study in Nepal's Terai to NASA Harvest and FAO pilots.

Traffic-signal AI from Google and UPS's ORION routing show what real-time data can save in transport, and where cities and fleets say the limits are.

How sensors and AI predict machine failures, from an 85-study 2026 review to a heavy-industry case that cut false alarms by 90%. What works and where it fails.

Open models like Gemma 4 can run on phones and laptops, but how well do they handle Nepali? What benchmarks and connectivity data show, and what is unproven.

The OECD's Digital Education Outlook 2026 says AI can lift student output without lifting learning. What it found, who gains, and where evidence is thin.

AI now scores payments for fraud. The US Treasury and a BIS pilot report gains, but false declines and deepfakes limit it. What the evidence shows.

How weather, meter and grid data feed AI forecasts of demand and wind output, what the IEA and DeepMind report, and the limits, including data-centre demand.

Google's Gemma 4 is an open-weight model family under the Apache 2.0 license. Here is what that means for privacy, cost and control, and what to check first.

The FTC has reportedly opened a probe into OpenAI, Anthropic and others over AI agents acting beyond instructions. What is known and what businesses should do.

GovAI researchers told Fortune that labs often run top models internally with safeguards off. What is confirmed, what is not, and what to ask AI vendors.

BI is moving from charts you read to agents that act on company data. What Databricks announced, what Deloitte and Gartner found, and what is unresolved.

EU AI Act transparency duties apply from 2 Aug 2026; high-risk rules slip to Dec 2027 and Aug 2028. A dated timeline and what non-EU firms should check.

Two reports give very different pictures of enterprise AI market share. What Menlo Ventures and a16z measured, why they disagree, and how to read them.

A 2024 Nature paper found AI models degrade when trained on AI-made data. What the evidence shows, who it affects and what teams can do about it.

A Gallup survey of 2,232 US teachers found weekly AI users report saving 5.9 hours a week. What it measured, what it did not, and the caveats.

The FDA cleared an AI sepsis early-warning system in May 2026. The evidence behind it, how it works, where it falls short, and what a rival alert showed.

Three recent AI tutoring trials found real learning gains when the tutor was designed well, and harm when it was not. The evidence and what is unknown.

Can AI answer questions from messy company databases? An ICLR 2025 benchmark of 632 real tasks shows the gap, who benefits and where it still fails.

Automation follows fixed steps, agents choose their own, orchestration coordinates them. What Anthropic, Microsoft and Gartner say, and when to use each.

A UCLA randomized trial of AI scribes found a 9.5% cut in note time for one tool, modest burnout gains and occasional note errors. What it means.

A randomized trial of 100,000+ women found AI-supported mammography reading linked to 12% fewer interval cancers. What it shows and where it stops.

About half of US adults use AI chatbots, yet globally under one in five working-age people do. What Pew, Microsoft and Stanford data show, and what it misses.

A dated guide to the AI conferences, launch milestones and policy deadlines coming up in October and November 2026, and how to follow them without chasing hype.

A May 2026 Nature paper shows Google's AI writing expert-level scientific software. The evidence, who benefits, and the limits scientists point out.