📍Four Engineers Cover 100 Teams
Plus: Five Agents Take the Handoffs, School Attack Shows Agent-Like Traces
Good Morning, Agent Owners!
Your AI can now decide whether you deserve a refund for skipping the gym. Even your excuses are being automated.
Four Engineers Cover 100 Teams
TL;DR: TUI’s four-person security team supports more than 100 product teams with an agent that reads architecture diagrams, selects relevant standards, and iteratively checks its threat assessments for gaps. Internal measurements show assessment time falling by up to 70%, with monthly sessions rising from three or four to ten or twelve. Human experts can compare their judgments against the agent’s findings before acting. Read more →
No-Show Penalties Get Reviewed
TL;DR: Cult.fit is letting an AI agent assess requests to waive penalties for missed fitness classes. It checks signals including local weather and attendance patterns, then applies the decision through existing support systems. Across its broader customer support operation, the company reports costs falling by more than half. Read more →
Bogotá Cuts the Hold Queue
TL;DR: Bogotá’s Linea 195 city-services hotline now has a Spanish-speaking AI agent answering residents around the clock. SoundHound says it can handle up to 1,000 calls simultaneously, with answers traced to official sources. The company reports call abandonment falling by roughly two-thirds and inquiries handled rising by 30% in the deployment’s first phase. Read more →
Thirty Failures, One Stuck Order
TL;DR: A U Mobile subscriber tried buying a data add-on more than 30 times over three days. Its internal agent traced the problem across systems to a pending order and returned the cause and remediation steps. Staff retain approval over account changes. U Mobile reports provisioning and network ticket volumes falling 84.5% between January and September. Read more →
Five Agents Take the Handoffs
TL;DR: Verisk has connected five agents to requirements, task breakdown, development, code review, and testing, with human approval at every stage. The company says test planning fell from three or four hours to under ten minutes, while automation setup dropped from one or two days to under thirty minutes. Cost per story point fell 20–30%, including cloud compute. Read more →
Digital Staff Get Performance Reviews
TL;DR: LTM has deployed 1,500 internal AI agents with employee IDs, assigned managers, defined roles, training, and annual performance reviews. They support onboarding compliance, invoice processing, and executive research across a business with 87,000 employees. LTM also reports 30% fewer IT tickets and a 21% increase in HR team productivity from its broader AI-enabled operations. Read more →
Leaked Cards Get Checked First
TL;DR: When a customer card appears on the dark web, Greenlight’s agents check ownership, card status, transactions, related support tickets, and duplicate alerts before assessing risk. High-risk cases reach people with supporting evidence. Greenlight reports autonomous resolution of over 90% of its monthly security and risk cases overall. Remediation still requires human authorization or an approved automated policy. Read more →
BMO Targets Faster Card Support
TL;DR: BMO is introducing an internal agent to match transactions, retrieve policies, and prepare cases for employees handling credit-card inquiries. Its contact centre receives more than 500,000 such calls annually. The bank targets 20–30% shorter resolution times for eligible calls and automatic processing of about 40% of eligible credit refunds. Staff remain accountable for reviewing information and making decisions. Read more →
Credit Reviews Drop to 12 Minutes
TL;DR: BILL says three agents now route underwriting cases, gather customer information into a summary, and give the assigned underwriter a direct decision link. Management reports case and customer waiting time falling from roughly three hours to twelve minutes. People still decide whether to approve or change credit lines. Previously, underwriters assembled cases through chat messages, email, and separate dashboards. Read more →
One Approval Exposes the Keys
TL;DR: Apiiro found that coding-agent approvals can save entire commands, including embedded credentials, in local settings files that later reach public GitHub repositories. Its investigation counted 2,539 distinct secrets across 1,994 repositories, including database passwords, cloud keys, and SSH logins. Researchers did not test the credentials, although one notified company confirmed a key had been live. Read more →
School Attack Shows Agent-Like Traces
TL;DR: Bitdefender’s investigation of a September ransomware intrusion into a UK school network found a mix of human activity and behavior resembling an agentic loop: reading results, revising commands, and trying again. Most steps used built-in Windows tools. The researchers assess AI involvement from operational traces, without access to the model or its conversations. Read more →
Shoe Design Meets Factory Feedback
TL;DR: Footwear manufacturer Changshin plans to connect development, sample planning, and factory operations through AI-enabled workflows. The project is intended to bring production feedback into development decisions, helping teams identify problems earlier and reduce friction between departments. The announcement gives no measured savings, number of factories covered, or timetable for completing the rollout. Read more →
Hotel Concierge Starts Taking Bookings
TL;DR: Resorts World Las Vegas has expanded its digital concierge into an external-facing booking agent, according to SoundHound’s earnings call. Previously, the concierge helped guests already on the property. The new booking role was introduced in the third quarter. SoundHound has not disclosed booking volumes, conversion rates, or revenue from the deployment. Read more →
Your agent cloud, everywhere — in your chat apps, browser, and tools. Always on.
- Add to Slack, Teams, Telegram or WhatsApp.
- With a clean, redesigned WebUI.
- Connect any of 100+ apps & sources in one click.
- GPT 6 Astra, DeepSeek v4.1 & more frontier AI models, built in at discounted rates.
Run your agent on MyClaw.ai in Seconds👇
