Weekly briefing

AI briefing, week 43/2026: ChatGPT connectors, n8n, Copilot licences

Reading time
7 minutes
Published
Topic guide
AI models
Written by
modernagents editorial teammodernagents is an AI automation agency for SMEs.
Illustration: a folded newspaper with a magnifying glass next to a desk calendar

Three developments this week affect small and mid-sized businesses directly: OpenAI retires administrator-managed sync connectors in ChatGPT Enterprise and Edu and moves to real-time access to Google Drive, SharePoint and Microsoft Teams. n8n 2.45 moves the Schedule Trigger to a new scheduler that times interval schedules differently. Microsoft announces that Copilot can be bought without a Microsoft 365 base subscription from 1 December 2026. In addition, Google delays skills in Workspace until further notice. For each item you get the primary source and our view on whether to act now, plan or watch. This issue covers announcements from 12 to 20 October 2026.

Key takeaways

  • Act: If your company uses ChatGPT Enterprise, check the new content controls and the sharing settings in Google Drive, SharePoint and Teams. With the real-time apps, each user's existing Google and Microsoft permissions apply.
  • Plan: If you self-host n8n, test version 2.45 on a test instance first. Workflows with interval triggers can run at different times afterwards.
  • Plan: If you need new Copilot licences before December, agree which offer you order. Through partners (CSP), sale of the current add-on offers ends on 1 December.
  • Watch: Skills in Google Workspace and Gmail's AI Inbox. Google gives no new date for skills in Workspace, and AI Inbox is currently only available in the US.

01 / 05

ChatGPT Enterprise: sync connectors retired, real-time access instead

What happened. OpenAI has retired administrator-managed sync connectors in ChatGPT Enterprise and Edu and points customers to the real-time Google Drive, SharePoint and Microsoft Teams apps instead. According to the release notes of 16 October 2026, there are new content controls for real-time Google Drive, Microsoft Teams and Outlook Email access in Enterprise workspaces where these controls are enabled. Admins find them in each app's workspace settings under Safety & Security Controls. Each user's existing Google or Microsoft permissions still apply. (OpenAI: ChatGPT Enterprise and Edu release notes)

What it means for SMEs. User permissions and enabled workspace controls limit which supported content ChatGPT can access through the real-time apps. A folder that is shared too widely in your company therefore remains a risk, because the existing permissions still apply.

Our recommendation: act. Check the Safety & Security Controls of each app in the workspace settings. Then check the folders in Drive, SharePoint and Teams that are shared with all employees or with external people. What matters for permissions is described in Preserve SharePoint permissions in RAG answers.

02 / 05

n8n 2.45: Schedule Trigger on the new scheduler

What happened. In n8n 2.45, released on 20 October 2026, Schedule Trigger nodes run on the new durable scheduler by default. It is enabled unless it is explicitly disabled, and it requires the workflow publication service. Scheduled runs are stored in the database before they are due. After a restart, they run within a grace period; beyond it, the trigger's misfire policy decides. Existing triggers switch over on upgrade without republishing. Schedules of the type "every N seconds" or "every N minutes" now run at a steady interval from publication time, not on clock boundaries. The variables N8N_SCHEDULER_ENABLED=false and N8N_SCHEDULER_TRIGGER_NODE_MODE=legacy are deprecated and cause a warning at startup, but remain available as a fallback. (n8n changelog, durable scheduler)

n8n 2.44, released on 13 October, renamed the Azure and Foundry nodes to "Microsoft Foundry". This changes only the display: saved workflows and credentials keep working. There is also a new credential type for Google Vertex AI. (n8n release notes)

What it means for SMEs. A workflow that ran on the quarter hour can run at a different minute after the update. That matters when another workflow or a person relies on that time, for example when a mailbox is read before an export. After a restart, recorded runs execute within their grace period; beyond it, the misfire policy determines the outcome.

Our recommendation: plan. If you self-host n8n, install 2.45 on a test instance first. List the workflows with interval triggers and move the ones that must run at a fixed time to a cron schedule. Remove the two deprecated variables only after a successful test; keep a fallback setting for a while if you still need it. What else self-hosting involves is covered in n8n Cloud or self-hosting.

Got a workflow like this?

Tell me about it. Honest take and a rough price, usually within a day.

AI models →

03 / 05

Microsoft 365 Copilot: without a Microsoft 365 base subscription from 1 December

What happened. Microsoft announced in Partner Center that Copilot offers can be bought without a Microsoft 365 base subscription from 1 December 2026. Current offers carry "(add-on)" in their names from 16 October and reach end of sale in the CSP partner channel on 1 December. Existing subscriptions need no action; product IDs, pricing and renewal terms stay the same. (Microsoft Partner Center: October 2026 announcements)

What it means for SMEs. The new offers are available across all channels from 1 December, including partners (CSP). The end of sale of the current add-on offers concerns purchases through CSP. From December, the offer you or your partner order for new Copilot licences therefore changes. Your existing licences continue unchanged. For usage-based billing for Copilot Business from December, see the AI briefing for week 41.

Our recommendation: plan. If you need more Copilot licences before December, agree with your partner or your Microsoft contact whether you order the add-on offer now or the new offer from 1 December, and get the terms of both offers in writing.

04 / 05

Google Workspace: skills delayed, AI Inbox US only

What happened. On 13 October 2026, Google delayed skills in Workspace for Rapid Release and Scheduled Release domains "until further notice". Skills in the Gemini app have been rolling out since 13 October, with completion expected by mid-November. (Google: An update on the launch of skills in the Gemini app and Workspace)

On 15 October 2026, Google announced that Gmail's AI Inbox is coming to paid Workspace accounts. It highlights to-dos and deadlines from emails and summarises important topics. The gradual rollout of up to 15 days started on 12 October for Rapid Release domains and starts on 26 October for Scheduled Release domains. According to Google, Starter and Education editions follow in the coming months. The feature is currently only available to users in the US with English (US) as their language, and there is no admin control for it. (Google: Gmail’s AI Inbox is rolling out to paid Workspace users)

What it means for SMEs. Skills in Workspace remain delayed, while skills in the Gemini app are already rolling out. AI Inbox does not reach users outside the US for now. If you started recreating Gems as skills for Workspace after the AI briefing for week 41, expect a later start there.

Our recommendation: watch. We will report when Google gives a new date.

05 / 05

In brief: the other providers

Provider What is confirmed Relevance for SMEs
Anthropic Code Review is generally available for Claude Team and Enterprise since 15 October (Code Review is now generally available) Watch: for teams with their own software development
Anthropic Fast mode for Claude Opus 5.5 left research preview on 15 October and no longer needs a beta header; also on Microsoft Foundry. Since 14 October, an OpenTelemetry export is in beta for the Messages API and Claude Managed Agents (release notes) Watch: for developers who monitor response time and cost
OpenAI ChatGPT Business admins can manage up to 25 plugins at a time in the Admin console since 15 October (ChatGPT Business release notes) Watch
AWS Amazon Bedrock AgentCore bills short-term memory by data volume since 12 October: 1.00 US dollars per GB ingested, 0.20 US dollars per GB retrieved and 0.10 US dollars per GB-month stored. For ingestion and retrieval, billable size is at least 12 KB and at most 64 KB per event; storage uses actual size. Long-term memory prices remain unchanged (Announcing consumption-based pricing for short-term memory in Amazon Bedrock AgentCore) Plan: if you use AgentCore, recalculate costs with your own data volumes
DeepSeek DeepSeek-V4.1-Flash was updated to version V4.1-Flash-1017 on 17 October; the API model name and price stay the same (DeepSeek change log) Watch: results can shift without a code change
xAI Grok Voice 3.0 is available in the API since 14 October at the same price as its predecessor (Introducing Grok Voice 3.0) Watch

We also checked Meta, Mistral, Aleph Alpha, Qwen, Make and Zapier as well as the European Commission and the AI Office, the RTR KI-Servicestelle, the Bundesnetzagentur and the data protection authorities in Austria and Germany, and selected no item from them for this issue. Left out on purpose: benchmark claims and funding news from the providers themselves.

If you want to know what one of these changes means for your existing workflows, we are happy to talk it through in an AI consulting call.

ShareEmail

Related reading

AI models

Got a workflow like this?

Tell me about it. Honest take and a rough price, usually within a day.

Robert Van Ysendyck · Founder & Managing Director