.avif)
🛋️ Ikea: reskilled call center staff now drive $1.4B in sales
Ikea's AI chatbot Billie now resolves 74% of customer queries, up from 47% two years after launch. Instead of cutting staff, Ikea retrained 8,500 call center workers as design consultants. Those remote-sales centers are now its fastest-growing sales channel, generating $1.37 billion last year, up 15-20%. Customer satisfaction rose to 89%, from 60% before Billie.
🏦 Lloyds Bank: £2 billion in AI-driven savings already banked
Lloyds Bank has already banked more than £2 billion ($2.5 billion) in cost savings through AI and digital transformation, and is targeting another £2 billion by 2030. Its InvestAI tool is live inside the Scottish Widows app, giving customers automated investment guidance. Lloyds plans to invest £13 billion ($16.5 billion) over four years to scale it.
🎖️ Veterans Affairs: $1.6B deal puts AI agents in 150+ facilities
The U.S. Department of Veterans Affairs awarded Salesforce a $1.6 billion, three-year contract to expand AI agents across more than 150 medical centers. The agreement aims to cut community-care scheduling from 28 days to minutes, building on Slack infrastructure already in place. Salesforce's agents will handle patient triage, intake, and benefits verification.
📈 Mackenzie Investments: AI infrastructure now live across $269B
Mackenzie Investments, the Canadian asset manager, completed a middle office and data foundation build with BNY Mellon in July, giving its C$269 billion ($196 billion) in assets under management a live foundation for AI tools. IGM Financial, its parent, cut investment teams from 15 to 9, redirecting C$70 million ($51 million) in savings into AI by 2028.
🧭 Jeremy Yang: AI agents broaden work, not just speed it
Jeremy Yang, a Harvard Business School professor, and Perplexity researchers compared Comet's conversational assistant against its autonomous agent in a live field study. The agent didn't just work faster and cheaper. It changed what people asked for, crossing occupational boundaries into multi-step projects that rarely surfaced with a standard assistant.
🎓 Ashish Agarwal: AI skill doesn't predict who thrives with agents
A field study of 523 early-career professionals sorted participants into three groups: AI apprentices, delegators, and amplifiers, based on how they worked with an AI agent on real tasks. Critical thinking, AI literacy, and domain knowledge did not predict which group someone landed in. Only amplifiers added value beyond what the AI produced alone.
🔀 Kris Johnson Ferreira: AI should connect silos, not just tasks
Kris Johnson Ferreira and Jordan Tong, professors at Harvard Business School and the Wisconsin School of Business, studied AI use at Walmart, Ericsson, Ramp, and Medtronic. Most AI tools improve individual tasks but fail at cross-functional work that drives performance. Their fix: AI routes information and surfaces trade-offs, while humans make the call.
🤖 Claude Opus 5: near-Fable-5 performance at half the price
Anthropic's Claude Opus 5 delivers performance close to Fable 5 at roughly half the cost per task, with pricing unchanged from its predecessor. A new effort dial lets teams choose how much computing power a task uses, so routine work runs cheap while complex analysis gets full power.
🛡️ Microsoft's MAI-Cyber-1-Flash: top cyber score, half the cost
Microsoft's new MAI-Cyber-1-Flash model scored 96% on the CyberGym vulnerability benchmark inside its MDASH security system, 12 points ahead of Anthropic's cyber model, at roughly half the cost of its prior setup. The Hacker News found the model scored zero on a separate benchmark testing exploit construction, a different skill from vulnerability detection.
🎙️ OpenAI Presence: enterprise voice agents resolve 75% of calls solo
OpenAI's new Presence platform packages voice and chat agents with guardrails, simulation testing, and escalation rules for enterprise use. OpenAI says its support line resolves 75% of calls without a human, though VentureBeat notes the figure is self-reported and unverified. Early users Spain's BBVA, SoftBank, and Australia's IAG are still in pilots, not production.
🪶 Thinking Machines: Inkling-Small matches big model at 1/4 the size
Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, released Inkling-Small, an open-weight model at about a quarter the size of its flagship but within one point of it on independent benchmarks. The smaller footprint means enterprises need far less GPU capacity to run it in-house. Fine-tuning is available through the company's Tinker platform.
🩺 TikTok's AI doctors draw millions of fans, zero medical degrees
AI-generated "doctors" now appear in 40% of TikTok's top health videos, rising to 84% for searches like "health tips," per a study by UK marketing agency Hallam. The avatars have pushed disproven cancer claims about microwaving plastic and sleeping near a phone, plus a supplement that doesn't seem to exist. Top videos average 2.5 million views.
⛪ A pastor cloned himself into a 24/7 AI chatbot
Justin Lester, pastor of a Baptist church in Vallejo, California, trained a chatbot on nearly 2 million words of his own sermons, then let it counsel his congregation around the clock. It has talked someone through the night before heart surgery and out of despair over a lost job.
🐒 Wild monkeys in Costa Rica are taking AI cognition tests
Researchers at Emory and Georgia Tech built CapuchinAI, a touchscreen station using facial recognition to identify capuchin monkeys in Costa Rica's Taboga Forest Reserve, then run each through a cognitive test for a dried banana reward. The system identifies monkeys with 97% accuracy. An alpha male named Papi was first to figure it out.
🐔 AI made them rich. Now they're saving chickens.
Donors tied to the AI industry have given roughly $40 million this year to help farmed animals, according to Lewis Bollard of Coefficient Giving, a farm-animal-welfare funder. The momentum traces to a tech podcast appearance that led host Dwarkesh Patel to launch a fundraising campaign, which has raised $2.5 million and counting.
🧓 Grandparents, not Gen Z, are AI's biggest robot fans
Anthony Niemiec, a widower in Beacon, New York, has spent four years with ElliQ, an AI-powered tabletop robot that greets him each morning and keeps him company. Made by Israeli robotics company Intuition Robotics, ElliQ has found its earliest adopters among Americans in their seventies and eighties, not the younger generations AI companion devices usually target.
