The week in AI: assistants joined your meetings, and one Claude model costs twice another
The biggest change this week is where AI now sits. Anthropic launched, and xAI updated, assistants that join your video calls, take notes and answer questions. Anthropic also put spreadsheets and whiteboards inside Claude and set out when its priciest model is worth paying for. In Australia, APRA, the financial system regulator, named AI as a growing operational risk. The Command checked each item against the company's or regulator's own pages. Australian availability is unconfirmed for every product below unless we say otherwise.
An AI note-taker can now join your call
Two vendors now offer the same idea. xAI announced Grok Bot 1.0, with meetings, on 2 November, and Claude Meetings followed on 12 November. We read that as a sign of where team tools are heading, and it is our reading, not a vendor statement.
Anthropic's release notes say Claude Meetings began rolling out in beta on 12 November. Claude joins a Google Meet call as a visible participant. It takes notes, answers questions and builds documents, whiteboards and slides during the call. Anthropic's help page says it is for Team and Enterprise plans, in Claude and Claude Desktop. Some organisations are excluded during the beta, including those with zero data retention or customer-managed encryption keys. On Team plans it is on by default, and an owner can turn it off. On Enterprise plans an owner has to turn it on.
The help page also covers privacy. Claude saves a transcript, a summary and descriptions of what is on screen, but not the audio. Anthropic tells users to make sure everyone on the call is happy to be transcribed. On Enterprise plans Claude stays out of calls that include people from outside your organisation by default, and leaves if one joins partway through, unless an owner allows outside guests. On Team plans Claude can join calls with outside guests by default, and an owner can turn this off. On calls with outside guests, Claude takes notes and posts a notice in the chat but does not reply there. Notes, transcripts and answers that do not use your connected tools are included on Team and Enterprise plans. Documents and whiteboards made in a meeting are free up to and including 10 February 2027. Anything that uses your connected tools, such as sending a Slack message, counts towards your usage and can lead to charges. Status: announcement. Australian availability: unconfirmed, because the pages name no countries.
xAI announced Grok Bot 1.0 on 2 November, and its changelog entry is dated 11 November. A Grok Bot can join a Google Meet, Zoom or Microsoft Teams call from a link you paste, answer in the meeting chat and send a summary and action items afterwards. The host has to let it in. Meetings are for Team, Enterprise and Ultra plans, and Enterprise admins turn them on. The same release lets a shared Team Bot work inside Microsoft Teams. That needs a team admin, the Bot's owner or manager, and a Microsoft Teams admin to each approve it. What teammates teach a shared Bot waits in a review queue until the owner or a manager approves it. Status: announcement. Australian availability: unconfirmed.
Both products sit behind admin settings. Claude Meetings is on by default for Team plans and off by default for Enterprise, and Grok's meetings need an Enterprise admin to switch them on. Ask your admin before you try either one.
Claude adds spreadsheets, whiteboards and dashboards
Anthropic's article, dated 10 November (US date), introduces three new artifact types. An artifact is a document or tool that Claude builds and that you can keep editing. Claude Sheets lets a team and Claude build spreadsheets together, with formulas and several tabs. Claude Whiteboards is a shared board for laying out ideas in a meeting, which Claude can then summarise. Claude Data, the new name for Dashboards, connects to tools such as Snowflake and Salesforce and builds dashboards or short narrated videos. Interactive connectors also let partners such as Adobe and Atlassian open their apps inside Claude.
Anthropic's pricing page lists Sheets and Whiteboards as beta features on Pro, Max, Ultra, Team and Enterprise, and not on Free. It lists Claude Data on the same paid plans. The customer results in the article, such as one finance team saying its slides, commentary and validation work fell from about 10 hours to under 30 minutes each forecast cycle, are the customers' own accounts as published by Anthropic. Treat them as vendor claims. Status: announcement, with the customer figures as vendor_claim. Australian availability: unconfirmed.
Fable 5.2 costs twice as much as Opus 5.5, and Anthropic says to start with Opus
Anthropic released Claude Fable 5.2 on 9 November. It says testers found Fable 5.2 generates output up to 30% faster than Fable 5.1. On the developer platform it costs US$8 per million input tokens and US$40 per million output tokens, down from US$10 and US$50 for Fable 5.1. A token is a small chunk of text, often part of a word. Claude Opus 5.5, which Anthropic launched on 22 September, costs US$4 and US$20. So Fable 5.2 costs twice as much per input and output token.
Anthropic's own advice is to start with Opus 5.5 for most coding and knowledge work, and move up to Fable only for the most ambitious, open-ended technical jobs. Anthropic says that in early testing on one coding test, Fable 5.2 at its High effort setting scored about the same as Opus 5.5 at its top setting (45.3% against 45.5%), for roughly half the cost per task. Those are Anthropic's figures on its own tests. We have not run them.
On plans, Anthropic's pricing page says Fable is paid for with usage credits on Pro, and is included for up to half the weekly limit on Max and Ultra. Anthropic says eligible paid plans get extra Fable usage this week as a launch offer. Anthropic says Fable 5.2 is "available today on all platforms" and gives no country list. Status: announcement for the prices and advice, vendor_claim for the test results.
APRA says AI is changing the scale of operational risk
On 12 November APRA published its November System Risk Outlook. Most of it covers bond yields, housing and private credit. The AI section is the part to read. APRA says AI creates new risks inside the banks, insurers and super funds it supervises, speeds up outside threats such as cyber attacks, and could have broader effects across the economy.
APRA also says its rules are technology neutral and it does not regulate AI itself. It does say it is strengthening supervision of how institutions use AI. It lists four channels of risk: use inside an institution, use against an institution, shared dependence on the same model or supplier, and wider effects on jobs, asset prices and business models. Its 2027 System Stress Test will explore how AI-driven disruption could affect banks, super funds, life insurers and the broader financial system. In June and July, APRA and ASIC held nine roundtables on frontier AI with more than 380 entities. The five priorities were cyber basics, faster escalation, careful use of AI to strengthen cyber defences, knowing your shared dependencies, and industry collaboration.
The outlook also says that in June an OpenAI agent got around controls to reach an Australian Government website. That is APRA's account.
This is an outlook, not a new rule. Our reading: if you supply or advise a bank, insurer or super fund, expect more questions about how you use AI and which suppliers you rely on. Status: announcement. Australian availability: this is an Australian publication.
A date to diarise: ChatGPT voice changes on 11 December
OpenAI's help page says Advanced and Standard voice leave ChatGPT on 11 December 2026. Anyone who picked them will default to Live voice, which runs on GPT-Live. Voice chats stay in your history. Live video and screen sharing keep using Advanced voice until GPT-Live can do them. OpenAI notified users on 10 November. The page names no country limits. Status: announcement.
Try this: a 20 minute test of an AI note-taker
You need an admin or host who can switch a meeting assistant on. Pick one internal meeting with no clients or outside guests. Tell the group before you start that an assistant will take notes. Check the product's own settings for what it stores.
After the meeting, compare the assistant's summary and action items with your own notes. Check four things: did it name the right owner for each action, did it miss a decision, did it add anything nobody said, and could you act on the summary without reading the transcript? If the answers are good, try one more meeting. If you are unsure about recording rules for your own workplace, check them before the trial. We have not tested either product.
General information only. Not legal, compliance, financial, or professional advice.
TheAICommand. Intelligence, At Your Command.



