Major companies are learning hard lessons as they roll out agentic AI. Kyndryl and Salesforce both pivoted after trying to deploy too many agents at once, while Freeport-McMoRan took a more targeted approach and saved money processing $500 million in invoices. Adobe focused on measuring success beyond speed, checking quality and employee satisfaction alongside efficiency gains.
Netflix is leaning into generative AI to improve personalization and creative tools for filmmakers. The streaming giant is testing a conversational interface that lets users describe what they want to watch, and it has developed an AI-enabled ad format that connects advertisements to the content viewers are currently watching. Netflix ranks 17th on the Fortune 2026 AIQ 75 list.
On the policy front, Nobel laureate Geoffrey Hinton and Yoshua Bengio co-authored a report urging governments to prepare for AI systems capable of accelerating their own development. The report examines recursive self-improvement, where automated AI research could compress years of progress into months. More than 20 researchers contributed, and the authors recommend monitoring mechanisms, independent scrutiny, and emergency response plans.
Law enforcement use of facial recognition AI is drawing scrutiny after a policing tech expert told Fox News Digital that mistakes have led to at least 16 wrongful arrests. The expert called for stronger safeguards as AI becomes more common in criminal investigations.
A new study testing eight black-box large language models found that changing one user attribute in a persona prompt also shifts responses on unrelated attributes. Researchers call this cross-attribute influence and identified a neutrality gap where models fail to keep non-target attributes stable. The findings suggest that following a persona does not guarantee controlling it selectively.
Mental health providers and payers are clashing over AI auditing tools. More than half of clinicians now use AI in their work, and providers worry that payers will use AI to audit claims more deeply than before. Some providers are using AI for their own internal auditing to get ahead of potential payer audits. Both sides call for more transparency and agreed-upon standards.
McDonald's is using AI and machine learning to set prices at nearly 14,000 restaurants. The pricing engine analyzes daily transactions and customer willingness to pay, creating price differences for the same item between locations. Some franchisees say the company pressured them to use the tools, though McDonald's says the system is optional.
Wharton professor Jeremy Siegel says AI agents could help fight inflation by comparing prices, negotiating bills, and switching providers automatically. He believes this would break customer inertia and increase price competition, but notes consumers will need better financial and digital literacy to use these tools safely.
Bradley University in Peoria, Illinois will host a TEDx event on September 30 focused on AI and its impact on leadership and daily life. The University of Utah is launching a new AI ethics class called PHIL 2150 / CS 2395, co-taught by philosophy and computer science professors. The course addresses fairness, bias, authorship, environmental costs, and job displacement, with about half the students from computer science and half from other majors.
China is expanding AI cooperation with ASEAN through training and joint projects rather than just selling technology. Wang Xiaolong from China's foreign ministry met ASEAN journalists in Beijing on September 28 and named AI as a key priority for the next five years. Seven ASEAN countries joined a new AI cooperation group in Shanghai last July, and China hopes Thailand will join soon.
Key Takeaways
- Kyndryl and Salesforce pivoted after deploying too many AI agents at once, while Freeport-McMoRan saved money processing $500 million in invoices through targeted AI use.
- Netflix is testing a conversational AI interface and an AI-enabled ad format tied to content viewers are watching.
- Geoffrey Hinton and Yoshua Bengio co-authored a report urging governments to prepare for AI systems that could accelerate their own development.
- Facial recognition mistakes have led to at least 16 wrongful arrests, according to a policing tech expert.
- A study of eight large language models found that changing one persona attribute can shift responses on unrelated attributes, called cross-attribute influence.
- More than half of mental health clinicians now use AI, and providers and payers are clashing over AI auditing standards.
- McDonald's uses AI to set prices at nearly 14,000 restaurants, analyzing daily transactions and willingness to pay.
- Wharton professor Jeremy Siegel says AI agents could lower consumer costs by comparing prices and negotiating bills automatically.
- The University of Utah launches an AI ethics class addressing fairness, bias, authorship, and job displacement.
- China is expanding AI cooperation with ASEAN through worker training and joint projects, with seven countries already joining a cooperation group.
Companies learn hard lessons from agentic AI rollouts
Kyndryl, Salesforce, Freeport-McMoRan, and Adobe all tried using agentic AI and learned important lessons. Kyndryl and Salesforce both pivoted after trying too many agents at once. Freeport-McMoRan focused on clear outcomes and saved money processing $500 million in invoices. Adobe measured success beyond speed by checking quality and employee satisfaction.
Netflix uses AI to personalize streaming and ads
Netflix is using generative AI to improve personalization and creative tools for filmmakers. The company ranks 17th on the Fortune 2026 AIQ 75 list. Netflix is testing a conversational interface that lets users talk about what they want to watch. The company also created an AI-enabled ad format that connects ads to the content viewers are watching.
Police facial recognition AI linked to 16 wrongful arrests
A policing tech expert told Fox News Digital that facial recognition mistakes have led to at least 16 wrongful arrests. The expert warned that law enforcement agencies need stronger safeguards as AI becomes more common in investigations.
Study finds LLM persona simulation lacks selective control
A new study tested eight black-box large language models and found that changing one user attribute in a persona prompt also shifts responses on unrelated attributes. The researchers call this cross-attribute influence and identified a neutrality gap where models fail to keep non-target attributes stable. The study suggests that following a persona does not guarantee controlling it selectively.
AI pioneers urge governments to prepare for intelligence explosion
Nobel laureate Geoffrey Hinton and Yoshua Bengio co-authored a report urging governments to prepare for AI systems that could accelerate their own development. The report examines recursive self-improvement and says automated AI research could compress years of progress into months. More than 20 researchers contributed, and the authors recommend monitoring mechanisms, independent scrutiny, and emergency plans.
Mental health providers and payers clash over AI auditing tools
More than half of mental health clinicians now use AI in their work. Providers worry that payers will use AI to audit claims more deeply and broadly than before. Payers say documentation standards are changing because of AI. Some providers are using AI for their own internal auditing to get ahead of potential payer audits. Both sides call for more transparency and agreed-upon standards.
McDonald's uses AI to set menu prices like the Big Mac
McDonald's is using AI and machine learning to set prices at nearly 14,000 restaurants. The pricing engine analyzes daily transactions and customer willingness to pay. It can create price differences for the same item between locations. Some franchisees say McDonald's pressured them to use the AI pricing tools. The company says the tool is optional and meant to help franchisees make informed decisions.
AI agents could lower prices and fight inflation, economist says
Wharton professor Jeremy Siegel says AI agents could help lower costs for consumers. These agents could compare prices, negotiate bills, and switch providers automatically. He believes this would break customer inertia and increase price competition. He compares this to how the internet made comparison shopping easier but with less friction. He notes that consumers will need better financial and digital literacy to use these tools safely.
Bradley University hosts TEDx event on artificial intelligence
Bradley University in Peoria, Illinois will host a TEDx event on September 30 focused on AI. The event will explore how AI is changing leadership, collaboration, and daily life. Six speakers will present including Ryan Terao, Becky Buchen, Cory Boulton, Ranen McLanahan, Injy Wasfy, and Charlie Wood. The event runs from 5 to 7 p.m. and is open to the public. Tickets cost $5 for students and $30 for non-students.
University of Utah launches AI ethics class for students
The University of Utah is offering a new AI ethics class called PHIL 2150 / CS 2395. The course addresses fairness, bias, authorship, environmental costs, and job displacement linked to AI. It is co-taught by professors from philosophy and computer science. Students from many majors are enrolled, with about half in computer science and half in other fields. The class aims to help future engineers build AI responsibly.
China expands AI cooperation with ASEAN through training and joint projects
China wants deeper AI cooperation with ASEAN beyond selling technology. Wang Xiaolong from China's foreign ministry met ASEAN journalists in Beijing on September 28. He named AI and the digital economy as key priorities for the next five years. Wang said the focus should be on training local workers and sharing knowledge rather than just exporting finished products. Seven ASEAN countries, including Cambodia, joined a new AI cooperation group in Shanghai last July, and China hopes Thailand will join soon.
Sources
- Agentic AI early adopters have failed, pivoted, and learned these 3 lessons
- How Netflix is using AI to personalize streaming
- Police AI mistakes have led to at least 16 wrongful arrests, expert says
- Persona Following Is Not Selective Control: The Neutrality Gap in LLM User Simulation
- AI pioneers warn governments to prepare for a possible ‘intelligence explosion’
- Fears of Payer AI Use in Auditing Loom as Providers Establish Their Own AI Tools
- Inside McDonald's push to have artificial intelligence price your Big Mac
- AI may prove disinflationary because agents can negotiate contracts and deals to make your life cheaper, says top economist
- Bradley University to host TEDx event focused on artificial intelligence
- Exploring the ethics of artificial intelligence
- China pushes AI training, joint projects as central to ASEAN cooperation
Comments
Please log in to post a comment.