Large Language Models

Today's Stories

1. 1

   Mastercard Launches Agent Pay for Machines

   Mastercard launched Agent Pay for Machines, a service that lets AI agents and software systems make high-speed payments across cards, bank accounts, and stablecoins. The service supports machine-driven commerce, including microtransactions worth fractions of a cent, and involves more than 30 firms, including Coinbase, Ripple, Stripe, and Cloudflare, signaling a shift toward AI as economic participants.

   Broad adoption will depend on trust, controls, and settlement reliability, as noted by Mastercard. The service's success hinges on whether merchants and enterprises accept AI agents with credentials and spending limits.

2. 2

   Costa Rica Adds Booking.com, Expedia, ChatGPT to 13% VAT list

   Starting Tuesday, a 13% tax will be added to bills from Booking.com, Expedia, ChatGPT, and 17 other digital platforms newly added to Costa Rica's VAT register. The list grew from 161 to 181 providers. This closes a gap where foreign digital services often avoided the VAT that local businesses must charge, leveling the playing field. The tax is not new; it has applied since October 2020 under Law 9635, but now includes major travel and AI services.

   The expansion is expected to significantly increase tax revenue next year, as travel platforms handle larger sums than streaming subscriptions. Watch your card statements after September 8 to see how each provider applies the tax.

3. 3

   ServiceNow Becomes Core Partner in Multi-Agent AI Push

   ServiceNow (NOW) has been named a core partner in a multi-agent AI workflow initiative. The announcement positions the company as a key player in coordinating AI agents that work together to automate business processes. This move signals that ServiceNow is expanding beyond its traditional IT service management niche into broader enterprise automation. For business readers, multi-agent AI workflows could reduce manual handoffs between systems, potentially lowering operational costs and speeding up decisions.

   The success of this partnership hinges on whether ServiceNow can prove its platform reliably orchestrates agents across diverse enterprise environments. Watch for early customer deployments and any metrics on workflow efficiency improvements from the initiative.

4. 4

   Chugai Pharma Cuts Dev Work from Months to Days with Claude Code

   Chugai Pharmaceutical has made Claude Code the core of its internal system development, after first adopting Gemini Code Assist. The company says this AI-driven approach has significantly improved productivity and removed bottlenecks. The move reflects a belief that without early adoption of generative AI, developers risk falling behind in productivity. By handling everything from requirements to operations with AI, Chugai aims to stay competitive in a rapidly changing field.

   The company claims tasks that once took one to two person-months now take one person-day. The success hinges on whether such AI-driven development can scale across more internal projects.

5. 5

   ZEALS' Omakase AI Debuts at MUFG Bank

   ZEALS has started providing its multimodal AI agent platform "Omakase AI" to MUFG Bank. The first use begins on September 7, 2026, for digital guidance about an account opening campaign. The AI agent aims to let customers ask questions naturally, in their own words, via voice or text, like talking to a bank teller. It is designed to meet banks' needs for safe, accurate, and secure customer interactions, with controls on response scope and a hallucination detection feature.

   The success hinges on whether the AI delivers the promised natural dialogue quality and safety in a real banking context. Watch how customers respond to this new digital experience during the campaign from September to November 2026.

6. 6

   Mitsubishi Electric's AI agent for AC maintenance trialed by 1,000+

   Mitsubishi Electric announced on June 25, 2026, at AWS Summit Japan that over 1,000 people are trialing an AI agent that supports building air conditioning maintenance. The agent uses customer-reported issues to present necessary information to field engineers and also helps create reports for customers and internal use. This trial addresses the problem of knowledge being tied to specific individuals (a common issue in maintenance work), potentially making expertise more accessible and reducing dependency on certain employees. It shows a practical application of AI in a non-tech industry, which could set a precedent for similar service operations.

   The trial's success hinges on whether the AI agent proves reliable enough for daily use and whether it can be scaled beyond the current 1,000+ users. Watch for any announcements about general availability or expansion based on the trial's outcomes.