Salesforce is rolling out 30 new AI features for Slack, which is a pretty loud signal that collaboration suites are becoming agent surfaces, not just chat apps. Slack admin, governance, and workflow ownership are about to matter even more because the app is turning into an execution layer, not just a communications layer.
News & Trends
NetSuite integrates external AI models using Model Context Protocol for ERP data (6 minute read)
Oracle NetSuite adopted the Model Context Protocol to integrate external AI models like Claude and ChatGPT with ERP data. Serving 43,000 customers, the update includes 100+ financial prompt templates and MCP Apps. This enables flexible, model-agnostic AI workflows while maintaining internal predictive capabilities via OCI.
Slack gets an AI overhaul (3 minute read)
Salesforce is rolling out 30 new AI features for Slack, which is a pretty loud signal that collaboration suites are becoming agent surfaces, not just chat apps. Slack admin, governance, and workflow ownership are about to matter even more because the app is turning into an execution layer, not just a communications layer.
Analysis & Opinions
AI agents hit the enterprise wall (4 minute read)
Agent failures are often less about model quality and more about disconnected enterprise systems, brittle workflows, and bad operational plumbing. It shifts the question from “which model should we buy?” to “what systems, permissions, and process boundaries are actually blocking deployment?”
RSAC 2026 shipped five agent identity frameworks and left three critical gaps open (5 minute read)
RSAC analysis highlighted three major gaps in current identity frameworks when applied to AI agents, including a lack of strong identity binding, poor lifecycle management, and limited visibility into agent actions. As agents begin acting autonomously across systems, traditional IAM models fall short, forcing enterprises to rethink identity, access, and accountability from the ground up.
Launches & Partnerships
Granite 4.0 3B Vision: Compact Multimodal Intelligence for Enterprise Documents (6 minute read)
IBM’s Granite 4.0 3B Vision is a compact multimodal model for enterprise document processing. Utilizing DeepStack architecture and the ChartNet dataset, it achieves 86.4% on Chart2Summary benchmarks. The LoRA-based model integrates with Docling to automate high-accuracy table extraction and key-value pair parsing from complex layouts.
AWS AI Risk Intelligence (AIRI) for non-deterministic agentic systems (7 minute read)
AWS AI Risk Intelligence (AIRI) automates governance for non-deterministic agentic AI systems. By operationalizing NIST and OWASP frameworks, AIRI evaluates security, transparency, and controllability through reasoning-based assessments. This will allow for continuous risk monitoring and actionable recommendations, allowing enterprises to scale autonomous agents while maintaining compliance.
Runway launches a $10M builders fund (2 minute read)
Runway is expanding from model maker to ecosystem shaper with a $10 million fund and builder program for early-stage AI startups. The generative AI market is maturing from “who has the best model” to “who can attract the best applications and developer ecosystem.”
Miscellaneous
AI supply-chain attacks keep spreading (3 minute read)
Mercor said it was hit in a cyberattack tied to the compromise of the open-source LiteLLM project, showing how attacks on AI tooling can quickly spill into downstream customers. The AI stack is now full of shared infrastructure, so one compromised package can radiate risk far beyond the original target.
WhatsApp Becomes Attack Vector (3 minute read)
Microsoft is warning of a campaign using WhatsApp messages to deliver malware that installs backdoors on Windows systems via multi-stage scripts and UAC bypass techniques. Attackers are blending social engineering with legitimate tools and trusted cloud infrastructure to evade detection, reinforcing that messaging apps are now a real enterprise attack surface, not just a user risk.
Quick Links
Claude Code Leak Exposes Playbook (3 minute read)
Anthropic accidentally exposed over 500K lines of Claude Code via an npm packaging error, revealing how its AI coding agent works.
Microsoft’s cloud licensing faces UK scrutiny (3 minute read)
The UK CMA is investigating whether Microsoft’s licensing terms unfairly favor Azure, which could matter for multicloud strategy and software procurement.