Ai 900
The latest Ai 900 coverage — news, analysis, and updates from the WindowsNews.AI desk.
Business Central’s New AI Agents Can Now Handle Invoices and Sales Orders Autonomously—See It Live July 22
Cambay Solutions is hosting a free July 22, 2026 webinar demonstrating new AI agents in Dynamics 365 Business Central that can autonomously process invoices, sales orders, and bank reconciliations. The session marks a shift toward agentic ERP, with practical implications for finance teams needing to balance automation gains against data quality and governance risks. Attendees will see live workflows and can request a personalized readiness assessment.
OpenAI Codex Windows App Sees 17x Adoption Surge in APAC; Singapore Cracks Top 5 Markets
OpenAI reports a 17x jump in Codex adoption among APAC startups since late 2025, with Singapore now among the top five markets. The surge coincides with the March 2026 Windows app release, offering a desktop interface for coordinating AI coding tasks. Windows developers and admins should adopt the tool with proper safeguards and review processes.
Copilot's New Metering: Every AI-Assisted Line of Code Costs You Starting June 2026
GitHub Copilot moves to usage-based billing on June 1, 2026, replacing premium requests with AI Credits based on token consumption. This turns Copilot from a fixed-cost tool into a metered service, demanding new budget controls, model access policies, and developer training from Windows enterprise teams. GitHub positions the shift as part of a broader platform strategy that routes tasks across multiple AI models while maintaining governance and lifecycle controls.
Microsoft Copilot Now Lets You Buy Etsy Products Without Leaving the Chat. Here's How It Works.
Microsoft Copilot now supports direct purchases of eligible Etsy items for U.S. shoppers, joining ChatGPT, Gemini, and Google AI Mode in the agentic commerce landscape. The move turns the AI assistant into a checkout surface, with orders processed as Etsy guest orders initially. While convenience is the main draw, questions remain around discovery, security, and seller visibility.
NVIDIA Agent Toolkit Brings Local AI to DGX Station—But Windows Users Are on the Outside Looking In
NVIDIA's new Agent Toolkit bundles a 550B-parameter model, agent blueprints, and 3D tools into a package exclusive to the Linux-based DGX Station GB300. Windows PCs cannot run the stack locally, but they may connect as clients; IT admins must prepare for a Linux appliance with ARM64 architecture and early-stage software. The rollout emphasizes on-premises AI for privacy and cost control, with future integrations targeting workflows across both Linux hosts and Windows creative tools.
Microsoft Azure Adds AMD Helios AI Inference and HPC VMs: A Roadmap for Windows Admins
Microsoft announced three new Azure VM families—ND MI455X v7, HDv2, and HXv2—powered by AMD's next-gen EPYC and Instinct hardware, with the Helios rack-scale platform targeting AI inference in the second half of 2026. The roadmap gives Windows admins and developers concrete AMD-based options for inference, data engineering, HPC, and EDA, though no instances are available to provision yet. Enterprises should monitor previews and plan capacity accordingly.
Microsoft Bets on AMD for Inference: Azure ND MI455X v7 VMs Arrive in Late 2026
Microsoft is adding AMD's Helios AI platform to Azure with the ND MI455X v7 VM family, built for inference and expected in late 2026. The move gives Azure customers a full-stack alternative to Nvidia, backed by EPYC CPUs and Pensando networking, but practical availability and software maturity remain open questions. Two new CPU-only VM families, HDv2 and HXv2, will also target AI data pipelines and HPC workloads.
Microsoft to Deploy AMD Helios AI Racks in Azure, Adds Three New VM Families
Microsoft and AMD announced a major partnership to bring AMD's Helios rack-scale AI system to Azure, along with three new VM families (HDv2, HXv2, ND MI455X v7) powered by Venice CPUs and MI455X GPUs. Aimed at inference, data preparation, and chip design, the new offerings give Azure customers an alternative to Nvidia starting in the second half of 2026. While specifics are pending, enterprises should begin evaluating AMD-based options for cost-sensitive and memory-bound AI workloads.
Grok Add-in Brings AI Editing to Excel, Word, and PowerPoint — Here’s What IT Needs to Do Before Installing
xAI has released a Grok add-in for Microsoft Excel, Word, and PowerPoint via the Microsoft Marketplace. While it offers plain-English data analysis and content generation, the add-in’s ability to read, modify, and transmit documents raises immediate data governance concerns. IT administrators should treat it as a third-party service integration and conduct controlled pilots before deployment.
Business Users Can Now Build AI Agents Inside Oracle Fusion—Without Breaking Governance
Oracle has introduced an AI-native builder inside Fusion Cloud Applications that lets business users and developers create agentic apps with native governance controls. These agents can execute real business processes—not just offer suggestions—inheriting security, approval, and audit frameworks. The move could finally close the gap between AI proof-of-concepts and production-ready enterprise automation.
Bun’s Million-Line Rust Overhaul: What the AI-Driven Rewrite Means for Your Windows Workflow
Bun, the JavaScript runtime, completed a Zig-to-Rust migration in 11 days using Anthropic’s Claude Code. The AI-generated port passed all existing tests and fixed 128 bugs, but 19 regressions later emerged. Windows developers should test thoroughly, pin versions, and monitor the project’s stability before upgrading production environments.
Why Microsoft 365 Copilot Needs a Permission Lockdown Before You Flip the Switch
Microsoft 365 Copilot integrates deeply with Word, Excel, and other Office apps by using each user’s existing permissions to access files via Microsoft Graph. This convenience becomes a serious risk if organizations haven’t audited who can access sensitive documents, because Copilot will silently use any file a user can open. IT admins must conduct a permission lockdown before enabling the tool.
Microsoft Picks AMD's Helios Rack-Scale Solution for Azure AI, Adds New EPYC Venice VMs
Microsoft is committing to AMD's Helios Rackscale Solution for Azure AI inference, integrating MI455X GPUs, EPYC Venice CPUs, and Pensando networking in a pre-built rack-scale system. Two new Azure VM series—HDv2 for agentic AI and HXv2 for engineering—will arrive on sixth-gen EPYC Venice silicon. Shipments start in H2 2026, but Azure availability details are pending.