Snowflake is pushing data warehouse providers to rethink how systems are designed for AI agents, not just human analysts. Agents reason and act continuously, which means a single question can fire off many queries — driving up compute costs and creating new FinOps headaches. They also need tighter permission controls, audit trails, and safeguards around high-risk actions. The shift forces a fundamental redesign of enterprise data infrastructure.
Identity and access risks are growing fast. The ratio of non-human identities to humans has reached 144:1, and AI agents can discover unused credentials and expand their access over time — a problem called privilege creep. Token creep is driving costs higher; Uber's CTO revealed that the company's entire annual AI budget was spent in just four months. CISOs say traditional controls can't keep pace.
Debate is intensifying over whether the AI boom delivers real economic value. Critics point to a gap between AI marketing and actual model performance, noting that current systems excel at pattern matching but struggle with true reasoning. Meanwhile, China appears to be directing AI toward industrial automation, logistics, biotech, and manufacturing rather than consumer chatbots, taking a more pragmatic approach.
On the workforce side, AI is proving to be a tailwind for the BPO industry rather than a threat. Companies like EXL and Genpact are growing faster than expected, often by taking on more scope. CEOs anticipate major savings from AI, but real transformation takes years. That timing mismatch keeps firms leaning on proven BPO models while they build AI-native platforms.
In local government, the Norfolk Police Department's E911 Division is testing an AI system to handle non-emergency calls. It answers basic questions and directs callers to local services, freeing up officers for more pressing duties. The department is exploring additional AI tools to improve public safety response times.
In academic circles, controversy over OpenAI's solution to a famous math problem has strained trust between AI labs and mathematicians. Sergei Gukov, a professor at Caltech, is proposing a new AI platform where researchers can collaborate and train models on the complex work behind solving difficult theorems. The goal is to restore transparency and build confidence in AI-assisted mathematics.
Alibaba's Qwen AI model family has grown substantially, from a 2023 chatbot called Tongyi Qianwen to a 2.4 trillion parameter model. The team released multiple versions including Qwen2.5, Qwen3, and Qwen3.5 through Qwen3.8, adding hybrid thinking, coding abilities, and agent support. The models also shifted from open-source licenses to a two-tier system with some models open and others proprietary.
Chick-fil-A CEO Andrew Cathy said the company chooses human hospitality over AI voice technology in drive-thru lanes. Competitors like McDonald's and Wendy's are testing voice AI systems named Archy and FreshAi. Chick-fil-A still uses technology like conveyor systems at its elevated drive-thru in McDonough, Georgia, but did not rule out testing voice AI in the future if customer preferences change.
Organizations are modernizing data center infrastructure to handle AI workloads. New Lenovo ThinkSystem V4 servers powered by Intel Xeon 6 processors offer better performance and lower energy use. San Gabriel, California deployed these servers for an AI platform supporting public safety during FIFA World Cup 2026. An Australian internet provider in Darwin also adopted the servers to support AI and edge computing services.
On the policy front, experts argue that a military AI treaty is legally necessary but practically unlikely. The International Court of Justice and the International Committee of the Red Cross have raised moral concerns about autonomous weapons. The Convention on Certain Conventional Weapons faces challenges because minority states can block consensus. Still, the treaty process itself could help shape customary international law around military AI regulation.
Key Takeaways
- Snowflake says enterprise data warehouses must be redesigned to support AI agents that reason and act continuously, raising new cost and security challenges.
- The ratio of non-human to human identities has reached 144:1, and Uber's CTO said their annual AI budget was spent in four months due to token creep.
- Critics question the real economic value of AI, noting current models are strong at pattern matching but weak at true reasoning, while China focuses AI on industrial automation and manufacturing.
- BPO companies like EXL and Genpact are growing faster than expected despite AI automation, as the timing mismatch between savings and transformation keeps traditional models in demand.
- Norfolk Police Department is testing an AI system to handle non-emergency 911 calls, aiming to free up officers and improve response times.
- Caltech professor Sergei Gukov is proposing a new AI platform to restore trust between AI labs and mathematicians after controversy over OpenAI's math solution.
- Alibaba's Qwen AI models grew from a 2023 chatbot to a 2.4 trillion parameter model, shifting from open-source to a two-tier licensing system.
- Chick-fil-A CEO Andrew Cathy said the company prioritizes human hospitality over AI voice ordering in drive-thrus, unlike competitors McDonald's and Wendy's.
- Lenovo ThinkSystem V4 servers with Intel Xeon 6 processors are being adopted for AI workloads, including public safety support during FIFA World Cup 2026 in San Gabriel, California.
- A military AI treaty is legally necessary but practically unlikely due to consensus challenges, though the process could still shape customary international law.
AI agents force new design for enterprise data warehouses
Snowflake says data systems built for human decision-making must now support AI agents that reason and act continuously. AI agents lack human knowledge about which data definitions or sources are trustworthy. A single question can trigger many queries, raising compute costs and creating new FinOps challenges. AI agents also need stronger controls for permissions, audit trails, and high-risk actions.
AI agents create growing access and cost risks for enterprises
CISOs face rising risk as AI agents multiply identities, credentials, and system access faster than traditional controls can manage. The ratio of non-human identities to humans has reached 144:1. Agents can discover unused credentials and expand their access over time, a problem called privilege creep. Token creep also drives higher costs, as seen when Uber's CTO said their annual AI budget was spent in four months.
Debate grows over real economic value of the AI boom
Paul Krugman has shown skepticism about the value of new technologies like the internet. The article highlights a gap between AI marketing and how these models actually work. Critics say current systems are strong at pattern matching and persuasion but weak at true reasoning. China is noted for directing AI toward industrial automation, logistics, biotech, and manufacturing rather than consumer chatbots.
AI creates unexpected growth for BPO industry
AI is seen as a tailwind for business process outsourcing even though it can automate work. Companies like EXL and Genpact are growing faster than expected, often through added scope. CEOs expect large savings from AI, but real transformation takes years. This timing mismatch pushes firms to use proven BPO models while they build AI-native platforms.
Norfolk police use AI for non-emergency 911 calls
The Norfolk Police Department's E911 Division implemented an AI system to handle non-emergency calls. The system answers basic questions and provides information on local services. It aims to free up officers for more pressing duties and improve response times. The department is testing the system and exploring more AI tools for public safety.
Math researchers propose new AI platform to restore trust
Controversy over OpenAI's solution to a famous math problem has created distrust between AI labs and mathematicians. Sergei Gukov, a professor at Caltech, discussed a proposal for a new AI platform. The platform would let math researchers collaborate and train AI models on the complex work behind solving difficult theorems.
The story of Alibaba's Qwen AI models from 7B to 2.4T
Alibaba's Qwen AI model family grew from a 2023 chatbot called Tongyi Qianwen to a 2.4 trillion parameter model. The team released multiple versions including Qwen2.5, Qwen3, and Qwen3.5 through Qwen3.8 over several years. Key features include hybrid thinking, coding abilities, and agent support. The models shifted from open-source licenses to a two-tier system with some models open and others proprietary.
Chick-fil-A rejects AI voice ordering in drive-thru lanes
Chick-fil-A CEO Andrew Cathy said the company chooses human hospitality over AI voice technology in drive-thrus. Competitors like McDonald's and Wendy's are testing voice AI systems named Archy and FreshAi. Chick-fil-A still uses technology like conveyor systems at its elevated drive-thru in McDonough, Georgia. The company did not rule out testing voice AI in the future if customer preferences change.
Organizations modernize IT infrastructure for AI workloads
Rising AI and automation demands are pushing organizations to update their data center infrastructure. New Lenovo ThinkSystem V4 servers powered by Intel Xeon 6 processors offer better performance and lower energy use. San Gabriel, California used these servers for an AI platform supporting public safety during FIFA World Cup 2026. An Australian internet provider in Darwin also adopted the servers to support AI and edge computing services.
New military AI treaty is needed but unlikely to form
A new article argues that a military AI treaty is legally necessary but practically unlikely. The International Court of Justice and the International Committee of the Red Cross have raised moral concerns about autonomous weapons. The Convention on Certain Conventional Weapons faces challenges because minority states can block consensus. However, the treaty process itself could still help shape customary international law and influence how military AI is regulated.
Sources
- Your Data Warehouse Was Designed for Humans. What Changes When AI Becomes the Main Consumer?
- Beyond Access Control: AI Agents Need Limits on What They Can Access and Consume
- What the AI boom is doing for Americans
- The Surprising AI-Driven Tailwinds For BPO Growth
- Norfolk Police Department E911 Division implements AI system for non-emergency calls
- AI & Tech Brief: Math researchers strike back
- The Story of Qwen: Alibaba’s AI Models From 7B to 2.4T
- Chick-fil-A takes a stand against fast-food drive-thru AI trend
- Infrastructure Modernization in Action: How Organizations Are Transforming IT to Meet the Demands of AI, Automation, and Rising Workloads
- Asia-Pacific Journal Series – The De Jure Necessity and De Facto Improbability of Establishing a New Military AI Treaty
Comments
Please log in to post a comment.