What's New in Fuel iX Copilots — Week of March 23, 2026

This week's edition is packed with meaningful improvements from the newsroom. We're reporting on a significant resilience upgrade that was impacting live conversations for multiple customers, a major overhaul of Google Drive knowledge base syncing, and powerful new analytics dashboards giving administrators and developers unprecedented visibility into platform usage. Alongside those, we have improvements to MCP authentication, model selection accuracy, and more. It's been a productive week in the Fuel iX newsroom!

🚀 Generally Available (GA)

Improved Conversation Resilience at Scale: Conversations now handle context memory limits more gracefully, with accurate warnings surfaced at the right moment so users can take action and continue their work. This improvement addresses a pattern where extended conversations could stop responding unexpectedly — particularly impactful for teams relying on Copilots during live, high-volume interactions.

More Reliable Google Drive Knowledge Base Syncing: Large file syncs from Google Drive to knowledge bases are now significantly more dependable. The pipeline has been redesigned to stream files directly with automatic retry logic, eliminating timeout errors that could occur with larger documents during bulk sync operations.

Faster PowerPoint Processing: PowerPoint files uploaded to knowledge bases are now processed more efficiently, with reduced complexity and improved ingestion speed for PPTX documents.

Stronger Knowledge Base Embedding Performance: Increased concurrent request capacity for the embedding service means files process more consistently, even during peak usage periods — reducing stalls during high-volume knowledge base syncs.

More Reliable Summary Generation for Long Conversations: Summary generation now handles long, document-heavy conversations without issue. A payload size constraint that could cause failures on large conversations has been increased, ensuring summaries work reliably regardless of conversation length or the number of documents involved.

More Accurate Model Availability in Copilot Configuration: The model selector now precisely reflects your organization's configured model access, ensuring what you see in the selector matches your organization's settings exactly.

Updated AI Model Availability: The SAIF (Sovereign AI Factory) model family has been temporarily removed from the Copilot model family selector while the team works to bring those models up to platform quality and reliability standards. Copilots previously using the SAIF family have been automatically migrated to the Claude model family — or Gemini CA for organizations with Canadian data sovereignty requirements. SAIF models remain available for intentional testing via pinned model and tool calling configurations, and we'll communicate when they're ready to return to the full selector.

Cleaner Shared Conversation Experience: The interface for shared, read-only conversations is now more polished — controls that aren't applicable in view-only mode no longer appear, creating a cleaner and more intuitive experience for conversation recipients.

Smoother MCP App Management Experience: The MCP interface now manages state more efficiently, eliminating screen flickering during interactions and improving visual alignment between element names and their associated controls when managing MCP app connections.

Improved MCP OAuth Authentication Reliability: The MCP authentication flow has been overhauled with stronger security controls, resolved reconnection loops, and more accurate authentication status tracking. Users can expect a smoother, more reliable experience when connecting, verifying, and revoking MCP server access.

Improved Reliability for Complex AI Workflows: Long-running AI agent workflows now have the time they need to complete. An extended request timeout has been restored, ensuring complex multi-step tasks aren't interrupted before they finish.

Model Usage Analytics for Administrators: A new endpoint enables administrators to retrieve token consumption and cost data grouped by AI model, making it easier to track spend and optimize usage across users and projects within custom date ranges.

Admin Portal Streamlined: The Admin Portal and API have been simplified by removing legacy subscription and metering infrastructure, resulting in a cleaner, more focused administrative experience.

API Usage and Cost Analytics Dashboard in Developer Portal: Developers can now monitor API consumption and spending directly in the Developer Portal. The new analytics dashboard provides visibility into total requests, token usage (input, output, and total), and estimated costs — with built-in time range filters (24h, 7d, 30d, 90d), a custom date picker, and interactive daily charts organized across Usage and Cost tabs.

Another edition from the newsroom — your trusted source for Fuel iX updates! ✍️