Office Upgrade

 Word and Excel, part of the Microsoft Office suite, have been two of its most popular products for users, span­ning students to businesses. So, when it brought agentic AI into them this week, it marked the leapfrogging of a concept to real world utility. Agent Mode could radically improve the user experience. In Excel, for example, if you have a compli­cated spreadsheet, then conversational commands in ordinary language will execute processes that would otherwise need expertise. Like asking it to pull out a slice of data, like a future projection or delayed payments, and make a table or graph out of it. While much better than earlier AI assistance, accuracy in its benchmark tests still trail behind a human being’s. That will change as the technology is refined.

In Word too, the document can be modified to the form the user wants. As Microsoft wrote in its blog, “Just tell Copilot what you need—like ‘summarize recent customer feedback and highlight key trends,’ and Copilot handles the heavy lifting: drafting content, suggesting refinements, and asking clarifying questions along the way.”

At present, in the initial rollout, this feature is available for Office 365 subscribers in the web version. Desktop applica­tions will get it at some point in the future.

 Coding Supremacy

 Anthropic’s latest AI model Claude Sonnet 4.5, claimed to be the best coding model, can reportedly operate autono­mously for 30 hours straight. The firm claims this iteration spent that many hours coding a chat app like Slack, spit­ting out about 11,000 lines of code, and stopped only once it had completed the task. This would give the firm a big leg up in its bid to corner the market on AI agents.

Apple has been known to be working on a new Siri based on large lan­guage models, but it has so far struggled. According to reports, it has now created a ChatGPT-style chatbot for internal Siri testing. Called Veritas, this chatbot is said to be testing new assistant features, including better contextual search and deeper integration with personal data.

 Facial Recognition

Scams that use AI to create fake celebrity endorsements on social media platforms are known to be a big problem. Last year, Meta had introduced a facial recog­nition technology to combat such ads in some markets. It is now extending this technology to also identify social media accounts that impersonate famous pub­lic figures. This feature will first roll out in the EU, the UK and South Korea.

 A Chatbot for SIRI

Apple has been known to be working on a new Siri based on large lan­guage models, but it has so far struggled. According to reports, it has now created a ChatGPT-style chatbot for internal Siri testing. Called Veritas, this chatbot is said to be testing new assistant features, including better contextual search and deeper integration with personal data.