TLDR IT 2026-04-17

OpenAI has significantly upgraded Codex with agentic capabilities that can control desktop apps, browse the web, and execute tasks across tools, positioning it as more than a coding assistant. New features like memory, multi-agent workflows, and 100+ integrations push Codex toward becoming a full enterprise automation layer, directly competing with Anthropic's Claude Code and accelerating the shift toward AI agents operating across the desktop and SaaS stack.

Salesforce introduced Headless 360, an API-first platform that lets AI agents execute business workflows directly without human interfaces, signaling a shift from systems of record to systems of execution. CIOs should watch pricing, SLAs, and vendor lock-in risk as Salesforce positions itself as the control layer for enterprise agents.

Most agentic AI initiatives are stalling due to poor coordination across business workflows, not tooling limitations. Success depends on tying agents to clear outcomes and integrating with existing systems instead of rebuilding from scratch.

The “95% of AI pilots fail” claim is misleading since it only reflects short-term revenue impact, while broader data shows most enterprises are successfully deploying AI and gaining productivity. AI works best when embedded into specific workflows and implemented through external vendors rather than internal builds. The real issue is not the technology but the ability to scale pilots into organization-wide impact.

Organizations succeeding with AI are heavily over-investing in data and analytics foundations and shifting toward trust-based governance models for AI agents. Without trusted data and embedded controls, AI delivers no value.

Make migration planning easier with the Azure Copilot Migration Agent. Assess readiness, risk, ROl, and landing zone automation through natural language prompts. Accelerate your migration process with Azure Copilot and have migrations that are better scoped, better justified, and less error-prone. Download the playbook to learn more and get started.

InsightFinder raised $15M to tackle a growing issue: debugging AI systems now requires monitoring data, models, and infrastructure together. As agents become embedded across stacks, failure modes are increasingly multi-layered and harder to isolate.

Salesforce launched a new Forward Deployed Engineering (FDE) Partner Network with firms like Accenture and Deloitte to help companies scale its Agentforce AI platform from pilots to real production systems. The initiative gives partners deep technical training and direct access to Salesforce teams, aiming to close the gap between AI experimentation and measurable business results.

DDoS mitigation is shifting from centralized scrubbing to the network edge, where integrated, router-level protection can block attacks earlier, reduce latency, and significantly lower infrastructure cost and complexity.

Enterprise security requires a hybrid approach to exposure validation, combining deterministic logic for repeatability with agentic AI for adaptive testing. This combination ensures consistent, measurable security outcomes while allowing systems to dynamically respond to modern attack techniques, moving beyond the limitations of purely probabilistic or static testing models.

Canva is repositioning itself as an agentic workflow platform where AI can create, edit, and manage design work across tools, turning design into an automated layer within broader business workflows.

Cisco issued an urgent security advisory for two critical vulnerabilities affecting Identity Services Engine and ISE Passive Identity Connector.