AI Agent Platforms Move from Experiment to Operating Layer
AI agent platforms are emerging as orchestration layers that coordinate multiple specialized AI agents, tools, and data sources to automate complex enterprise workflows end to end while preserving human oversight, security, and regulatory control. OutSystems and Filigran are now pushing this model into mission‑critical domains with platforms that treat agents as a governed workforce rather than add‑on features. At the OutSystems ONE 2026 conference, the company introduced its Agentic Systems Platform, built on the Enterprise Context Graph, to help organizations become AI‑native without tying core business logic to a single model provider. Filigran, meanwhile, launched XTM One as an AI‑native agentic layer for Continuous Threat Exposure Management, turning scattered security tasks into a continuous loop from threat intelligence to remediation. Together, these launches signal a shift from ad‑hoc AI use toward structured, autonomous workflow orchestration.
OutSystems Builds an Agentic Systems Stack for Enterprise Automation
OutSystems’ Agentic Systems Platform introduces an AI‑centric stack that spans engineering, workflow orchestration, and industry solutions. At its ONE 2026 event in Amsterdam, the company detailed how the Enterprise Context Graph underpins Agentic Enterprise Orchestration, enabling enterprises to architect and govern a flexible workforce of AI agents across business processes. The new OutSystems Agent Experience layer exposes A2A and MCP tools so developers can build, orchestrate, and govern their agentic portfolios, with early services for coding, publishing, and extensibility now live. Deeper collaboration with AWS, including native integration with Kiro and AWS Transform for legacy modernization, highlights an architecture designed for digital sovereignty, full runtime isolation, and self‑hosting options. According to OutSystems CEO Woodson Martin, the goal is to keep proprietary logic and data independent from any single AI provider, preserving both cost control and strategic flexibility.
Filigran’s XTM One Turns CTEM into an Agentic Workflow
Filigran’s XTM One focuses on security, introducing an AI‑native orchestration layer that automates Continuous Threat Exposure Management across the XTM Platform. It connects OpenCTI and OpenAEV into a single workflow so AI agents can ingest and enrich threat intelligence, generate and validate attack scenarios, and guide remediation from one interface. Instead of analysts moving manually between tools and dashboards, XTM One coordinates prepackaged agents for intelligence ingestion, summarization, scenario building, and report creation. Early benchmarks on the XTM Platform show organizations achieving up to 70% faster threat detection and response cycles and up to 80% less preparation time for offensive security testing. Filigran positions XTM One as “AI as the operating system for threat management,” arguing that the volume of CVEs, threat actors, and campaigns now exceeds what human teams can process without end‑to‑end automation.
Sovereignty, Security, and Customization as Design Principles
Both launches underline how AI agent platforms must handle enterprise automation without sacrificing control, compliance, or data protection. OutSystems emphasises digital sovereignty through a distributed architecture with runtime isolation and self‑hosting, allowing AI workloads to run where regulatory posture demands while working across multiple model providers such as Amazon Bedrock. Filigran’s XTM One mirrors these concerns in cybersecurity: Bring Your Own LLM support lets customers choose between Filigran‑provided models or their own, and on‑prem deployment keeps sensitive threat intelligence within internal infrastructure. Security and risk teams can build custom agents, workflows, skills, and integrations rather than accept fixed automation. This combination of workflow orchestration, agentic systems, and deployment flexibility shows how next‑generation AI agent platforms aim to become core control planes for enterprise operations, not sidecar tools.
