Daily AI brief · October 9, 2026
Google's Gemini Agent Will Work Inside Microsoft 365
Google unveiled one Gemini agent that works across Gmail, Microsoft 365 and Slack, Cisco is bringing Claude-powered agents into Webex meetings, and OpenAI caught Russian and Iranian operators misusing ChatGPT.
Google launches one Gemini agent that works across Workspace, Microsoft 365 and Slack (9to5Google, CNBC, Android Authority) At its Gemini at Work event yesterday, Google Cloud chief Thomas Kurian introduced the Gemini agent, which Google calls a single agent for work. It can carry out multi-step tasks on its own, you call it up by typing @Gemini in Gmail, Docs, Sheets, Slides and Chat, and it also reaches into Microsoft 365, Slack and the command line. Google says it picks the best model for each job, including its own Gemini models and Anthropic's Claude models today. It is in private preview now, with wider release promised soon for some Workspace Business and Enterprise plans. For a small business, the interesting part is that the AI helper is no longer tied to the office suite you bought, so a mixed shop running Outlook and Google Drive side by side may not have to choose a camp to get one capable assistant.
Cisco brings AI agents, including Claude, into Webex meetings and calls (SiliconANGLE, UC Today) At WebexOne in Austin this week, Cisco redesigned the Webex app around people and AI agents working side by side. Agents can join spaces, meetings and calls as members of the team, and Cisco gave the example of tagging an analytics agent to pull usage numbers, then asking a presentation agent to update the slides. A Claude integration built on Anthropic's Claude Managed Agents is on the way, and most of the new features are planned for early 2027. If your team lives in Webex, you have a few months to think about a new kind of question: when an agent sits in a meeting, what is it allowed to hear, remember and act on afterward?
OpenAI catches Russian and Iranian operators using ChatGPT for influence campaigns (NPR, NBC News) OpenAI banned accounts tied to Russian and Iranian influence operations. The Russian accounts mostly used ChatGPT to write progress reports to their bosses, and those reports described a fake think tank in Latin America and posing as a regional education office in Peru to get schools to run events. The Iranian accounts invented seven fake authors and placed almost 100 articles with small media outlets. The practical part: none of this needed hacking, just convincing writing under a borrowed name, which is exactly the trick used against small businesses in fake vendor and fake official emails, so a call-back rule for unusual requests matters more every month.
Ask about the Gemini agent if you use Google Workspace. Have your AI champion check which plan you're on and ask your Google reseller or IT partner whether it qualifies for the wider release. Pick one repeat task, such as pulling a weekly summary from email and a spreadsheet, as the first test when it arrives.
Add a meeting line to your staff AI rules. "No AI assistant or note-taker joins a call with a customer unless the customer has been told, and recordings are kept only as long as we need them." Agents in meetings are coming from Cisco, Microsoft and Google, and the rule is easier to set before the first one shows up.
Be wary of polished requests from official-sounding names. A letter from a "regional authority" or a "research institute" is now cheap to fake well. If a request asks for money, data or a change in how you pay someone, confirm it with a phone number you already had.
Where it's going: AI agents are spreading into every app your team touches, so the businesses that do well will be the ones that decide early what each agent may see and do.