TLDR IT 2026-05-05

Team collaboration differentiates your business, yet tool sprawl and disconnected data often hinder productivity and AI effectiveness.

In Transform Work Management and Collaboration Technology Into a Strategic Asset, a study conducted by Forrester Consulting and commissioned by Atlassian, 280 IT leaders share how they're rethinking their digital workplace strategy.

This Forrester Opportunity Snapshot reveals:

• What IT leaders are prioritizing to reach business outcomes.
• Why disconnected data inhibits AI scalability for 58% of leaders.
• How CX, agility, and security drive top IT business goals.

Download the Forrester study

Cisco Systems is acquiring Astrix Security to help enterprises discover and secure AI agents, MCP servers, API keys, service accounts, and OAuth tokens before they become unmanaged backdoors. Cisco plans to fold Astrix into Cisco Identity Intelligence, Cisco Secure Access, and Duo Identity and Access Management, a sign that “who/what has access” is becoming a major control layer for enterprise AI adoption.

Google launched an AI Control Center for Google Workspace that gives admins visibility into which AI agents can access Gmail, Drive, Calendar, and other Workspace data, along with controls to approve, restrict, or revoke that access. Governance is becoming a real admin function as companies race to deploy AI tools across productivity stacks.

Anthropic's heavy use of AI-generated code and its willingness to still pay top engineers up to $570K highlights a broader shift: engineers are moving from writing code to orchestrating AI agents. Software development is scaling AI faster than other functions because it already has what most enterprises lack: governance, observability, testing frameworks, deployment controls, and audit trails.

Enterprises are increasingly routing routine tasks to smaller, specialized models while reserving large models for complex reasoning, cutting inference costs by as much as 90%, reducing latency, and keeping sensitive data closer to home. The bigger shift is toward multi-model AI stacks where companies optimize for cost, speed, and privacy instead of defaulting every workflow to the biggest model available.

Pipelock is an open-source firewall that sits between AI agents and the internet to block credential leaks, prompt injection attacks, SSRF attempts, and poisoned tool responses before agents take action. The latest release adds request redaction and streaming-response scanning, signaling that security teams are starting to build infrastructure specifically for agent traffic, not just model access.

IBM and Oracle Corporation expanded their long-running partnership with new integrations focused on hybrid cloud modernization, AI deployments, and simplifying Oracle workloads across enterprise infrastructure. A lot of enterprise AI spend is still flowing toward boring-but-critical modernization work, not flashy new apps.

AWS CEO Matt Garman pushed back on SaaSpocalypse narratives, arguing AI will reshape software experiences rather than replace SaaS altogether. Amazon is betting that future workplace software will become more proactive, embedded, and agent-driven

OpenInfra Foundation's StarlingX 12.0 adds partial precision timing support, unified OIDC authentication, and easier backup/restore workflows, making it easier for enterprises to run distributed edge infrastructure without needing identical hardware across every site.

AMD and Intel Corporation are working on shared extensions to make AI workloads run more consistently across both chip ecosystems.

Cisco added a temporary “Galaxy Mode” to its AI Assistant.

Companies are aggressively cutting redundant SaaS vendors as procurement teams prioritize platform consolidation and lower software spend.