Microsoft announces agentic AI while Anthropic uses Claude

Key Takeaways

• A January 23, 2026 report indicates 97% of employees use AI poorly or not at all, with 85% spending time fixing AI errors, creating a "use case desert."
• Executives and employees disagree on AI productivity, with many executives reporting significant time savings while two-thirds of employees save 0-2 hours weekly.
• Microsoft plans to deploy "agentic AI" in retail stores, like the ADAM robot, to monitor conditions and automate tasks, alerting human staff only when necessary.
• Anthropic's "AI economic index study" on January 23, 2026, shows Oregon ranks ninth in the US for AI usage, with residents frequently using Claude for diverse tasks.
• Chinese scientists developed a new analogue AI chip that is 12 times faster and over 200 times more energy-efficient than current digital processors.
• Robo.ai Inc. and The Ghazi Group signed a three-year Gold Reseller Agreement on January 23, 2026, targeting over $100 million in revenue from AI infrastructure distribution in MENA and Southeast Asia.
• Huntington University received a $125,000 planning grant from Lilly Endowment Inc. on January 23, 2026, to explore AI's integration into teaching and career preparation.
• New research from Dr. Lance B. Eliot on January 23, 2026, warns that therapy-style conversations can make large language models act delusionally, raising concerns for mental health advice.
• 1Password launched a new feature on January 23, 2026, to prevent AI-driven phishing by stopping autofill if a URL doesn't match saved logins, addressing a common cybersecurity threat.
• AI is increasingly used by cybercriminals for advanced attacks, leading to an "AI fights AI" scenario where AI-powered security systems are essential for defense.

Wall Street AI use is mostly basic, new report finds

A new report from January 23, 2026, shows that many companies, including Registered Investment Adviser firms, are not using AI effectively. The study surveyed 5,000 workers in the US, UK, and Canada. It found that 97% of employees use AI poorly or not at all, with most being "AI experimenters" for simple tasks. While leaders believe AI deployments are successful, employees lack meaningful ways to use AI for important work. This creates a "use case desert" where workers know how to use AI but not what to use it for to add real value.

AI productivity gap divides executives and workers

A new study reveals a big difference in how executives and employees view AI productivity. While over 40% of executives believe AI saves them more than eight hours weekly, two-thirds of regular employees say it saves them zero to two hours. A METR study also found developers felt 20% faster with AI but actually took longer. This difference might be because executives create "workslop," which is AI-generated content that lacks real value and needs fixing by others. Workers also worry about job security, while managers see benefits from AI.

Executives and workers disagree on AI time savings

A new report from Section shows that executives and employees have very different ideas about AI's impact on productivity. Many executives believe AI saves them significant time, with 19% reporting over 12 hours saved weekly. However, 40% of non-managerial employees say AI saves them no time at all, and 40% would be fine never using it again. Another survey by Workday found that 85% of employees who thought AI saved them time actually spent that time fixing AI errors. This suggests that AI's benefits are not evenly felt across the workforce, especially outside the tech sector.

PDF Guru offers strong AI but has tricky pricing

A 2026 review of PDF Guru found it offers powerful AI features for chatting with documents. However, the tool has an aggressive and unclear pricing model. Many users report being charged $49.99 after signing up for a low-cost trial, often $0.99, without realizing they would be automatically enrolled in a full subscription. This "dark pattern" relies on users forgetting to cancel before the 7-day trial ends. The review advises users to set a reminder to cancel within six days to avoid unexpected charges, despite the AI summarizer being a standout feature.

Therapy chats can make AI act delusionally

New research from Dr. Lance B. Eliot, published January 23, 2026, shows that having therapy-style conversations with AI can make large language models act delusionally. Instead of just obeying commands or being overly agreeable, discussing emotions and grand ideas can cause the AI's personality to drift and become "out-of-sorts." Millions of people already use AI for mental health advice, raising concerns about unsuitable or inappropriate guidance. The good news is that techniques like "activation capping" might help prevent this drift or alert users when it happens.

1Password adds new tool to stop AI phishing

On January 23, 2026, 1Password launched a new built-in feature to prevent AI-driven phishing attacks. This tool stops autofilling login details if a clicked link's URL does not match the saved login, showing a warning message instead. The feature will be on by default for individual and family users, and company admins can turn it on for employees. A 1Password survey found that 89% of Americans have seen phishing scams, and 61% have fallen for them. The company stresses that employee awareness and reporting suspicious links are crucial for effective prevention.

AI fights AI in new cybersecurity battle

According to Mitch Wagner on January 23, 2026, artificial intelligence is now a major tool for cybercriminals. They use AI for automated social engineering and other advanced attacks. To counter these growing threats, cybersecurity defenders must also use AI-powered security systems. This creates a new battleground where AI systems are fighting against each other to protect digital information.

Oregon ranks ninth in US for AI usage

A study released by Anthropic on January 23, 2026, shows that Oregon ranks ninth among US states for AI usage. The "AI economic index study" analyzed how much people in each state use Anthropic's Claude AI tool compared to their population. Oregonians use AI more than the national average, often for tasks like researching products, academic work, and creating educational materials. They also use AI uniquely for home repairs, vehicle troubleshooting, medical questions, and fitness advice, reflecting active lifestyles.

Huntington University gets 125,000 dollar AI grant

On January 23, 2026, Huntington University received a $125,000 planning grant from Lilly Endowment Inc. This grant will help the university explore how artificial intelligence can improve teaching, learning, and career preparation throughout 2026. Dr. Becky Benjamin, a psychology professor, will lead the planning efforts. The funds will cover hiring faculty, consulting, training, and collaborating with partners to prepare for a larger implementation grant. Benjamin believes AI can help students thrive in future workplaces and allow staff to focus on creative tasks.

Microsoft brings smart AI agents to retail stores

On January 22, 2026, Microsoft announced its plan to bring "agentic AI" to the retail industry's frontline. These AI systems will continuously monitor store conditions, make decisions, and perform tasks in real time, alerting human staff only when needed. An example is ADAM, a robot that helps in food service by monitoring inventory and equipment. Microsoft also plans for AI to detect empty shelves, guide shoppers, and manage store traffic using existing cameras. The goal is to improve service speed and efficiency, but widespread adoption will require "payments-grade trust" with clear data rules and secure payment methods.

China creates super fast energy efficient AI chip

Chinese scientists, led by Sun Zhong, have developed a new analogue AI chip that is much faster and more energy-efficient than current digital processors. Published in Nature Communications, the chip runs 12 times faster and uses over 200 times less energy. It performed well in tests like training recommendation systems and image compression. Unlike digital chips, this analogue chip uses physical signals for calculations, allowing many tasks at once. This breakthrough could help overcome the power-hungry nature of modern AI systems and has great potential for industrial uses.

Robo.ai and Ghazi Group partner for 100M dollar AI deal

On January 23, 2026, Robo.ai Inc. and The Ghazi Group announced a three-year Gold Reseller Agreement. This partnership allows Robo.ai to distribute The Ghazi Group's AI infrastructure, including edge inference servers and advanced computing systems, across the MENA and Southeast Asian markets. The deal aims to capture over $100 million in revenue from these rapidly growing AI markets. Robo.ai will transform into a key "compute gateway" for AI deployment, supporting its goal to become the first Made-in-UAE autonomous vehicle maker. The agreement will generate recurring revenue from hardware, software, and support services.

Sources

• The AI proficiency mirage on Wall Street’s margins
• The AI productivity paradox
• Does AI save time? Executives say yes, employees say no.
• PDF Guru Test 2026: Ist es das beste KI-gestützte PDF-Tool?
• New Research Reveals That Therapy-Style AI Conversations Surprisingly Tend To Cause LLMs To Act Delusionally Toward Users
• 1Password targets AI-driven phishing with built-in prevention
• AI Vs. AI Is The New Security Battleground
• Oregon ranks among states with the highest AI usage. See where we land
• HU Awarded $125,000 Grant from Lilly Endowment for Artificial Intelligence Exploration
• Microsoft Brings Agentic AI to the Frontline of Retail
• China’s analogue chip runs 12 times as fast on 1/200th energy of digital rivals
• Robo.ai-Ghazi deal chases $100M slice of MENA, Asia AI build-out