Weekly signal

This week (covering July 13–21, 2026) pushed agentic AI from lab demos toward production-ready business automation: major vendors shipped builder and governance primitives for multi-agent workflows, and a wave of runtime security and control-plane products launched to manage agents operating across enterprise systems. Key signals: a major ERP/app vendor delivered an "agentic applications" builder for business processes; data and AI-governance vendors released dedicated OS-style tooling for enterprise agent governance; several startups published runtime control planes and zero‑trust runtime security for agents; and RPA/automation platforms continued incremental releases adding LLM-driven file and orchestration primitives that integrate into enterprise workflow automation.

What changed

1. Oracle expanded its Fusion stack with an AI-native Builder and "Fusion Agentic Applications" that let teams compose specialized agents that can act on business objects, policies, approvals and logged actions — explicitly positioning agents as outcome-driven automation inside ERP/business apps.

2. Governance and operating-system plays: Alation announced AIOS, an "intelligence operating system" focused on governed AI workflows and audit-ready evidence for agentic deployments. That frames governance as an integral runtime component, not an afterthought.

3. Runtime control & security launches: multiple startups released agent control planes and zero‑trust runtime platforms — Lineation’s runtime security control plane and Alterion’s Draco target real‑time visibility, boundary enforcement, and interventions across heterogeneous agents and clouds. These products solve gaps enterprises face when agents interact with sensitive systems/data.

4. Platform continuity: UiPath updated its Agents features (LLM-driven file analysis, agent tooling) and other vendors signaled product integrations to let agents operate over files, workflows and existing automation assets, lowering the friction to convert RPA and workflow automations into agentic flows.

5. Knowledge layer for codebases: Concho AI launched a product that converts application architecture and code understanding into a knowledge layer usable by agents — useful for modernization and change‑management automations.

What to do with it

• Reassess deployment risk: treat runtime control and identity as first‑class requirements for any agent pilots; evaluate Lineation/Alterion proof points for boundary enforcement and audit trails.
• Map agents to business objects: start experiments that pair agent behaviors with transactional business objects (ERP, CRM) so you can capture audit, approvals and rollback patterns like Oracle’s model suggests.
• Upgrade governance workflows: require evidence chains (who/what changed, why) from agent runs and test Alation‑style OS controls for lineage and compliance reporting.
• Incrementally convert RPA to agentic flows: pilot agent tasks that need LLM reasoning over files or multi‑step orchestration, leveraging UiPath’s agent primitives rather than rewiring core systems immediately.
• Invest in a knowledge layer: for modernization or developer productivity, evaluate tools that convert code and architecture into agent-consumable context (Concho) before commissioning autonomous change agents.