Facio Blog

Practical notes on human-reviewed AI agents.

Payload-powered product notes, security writing, HITL patterns, and operational guidance from the Facio runtime: long sessions, Placet approvals, audit trails, memory, providers, channels, tools, and Docker-first operations.

Showing 61-65 of 235 articles.

Human-in-the-loop

HITL for Multi-Tenant AI Agents: Why the Same Reviewer Pool Can't Govern Different Customers

HITL for Multi-Tenant AI Agents: Why the Same Reviewer Pool Can't Govern Different Customers

Most HITL designs assume one deployer, one reviewer pool, one policy. But enterprise AI is multi-tenant — the same agent infrastructure serves dozens of customers, each with their own data, their own compliance regime, their own reviewers. Treating them as one tenant is a privacy violation, a compliance violation, and a design defect. Here's how HITL has to evolve for the multi-tenant reality.

Engineering

MCP Spotlight: Playwright MCP Server by Microsoft — The Accessibility-Snapshot Browser Automation Default for Agents

Jul 7, 2026Engineering

MCP Spotlight: Playwright MCP Server by Microsoft — The Accessibility-Snapshot Browser Automation Default for Agents

The official Playwright MCP Server by Microsoft — ~40 focused tools covering navigation, interaction, forms, network mocking, tracing, video recording. Multi-engine support (Chromium/Firefox/WebKit), accessibility-snapshot navigation (not pixel screenshots), Apache 2.0. The browser-automation default for AI agents in 2026.

Product

Facio's Deploy Discipline: How AI Agents Ship Code Without Breaking Production

Jul 7, 2026Product

Facio's Deploy Discipline: How AI Agents Ship Code Without Breaking Production

An AI agent that can write code but not deploy it is a development tool. An AI agent that can both write and deploy is a team member. The difference is huge — and so are the risks. Production deployments are where good code meets reality: networking, dependencies, secrets, traffic patterns, monitoring. Without discipline, an agent deploying code can take down production in seconds. Facio's deploy discipline gives agents the structural patterns to ship code safely: pre-deploy verification, canary rollouts, automated rollback, post-deploy validation.

Security

Human-in-the-Loop Is Now a Legal Requirement: Engineering the HITL Architecture That Satisfies EU AI Act Article 14 Before August 2, 2026

Jul 7, 2026Security

Human-in-the-Loop Is Now a Legal Requirement: Engineering the HITL Architecture That Satisfies EU AI Act Article 14 Before August 2, 2026

EU AI Act Article 14 enters full enforcement for Annex III high-risk AI systems on August 2, 2026. HITL is no longer a best practice; it is a legal obligation with fines up to €15M or 3% of global turnover. The architecture must support four capabilities, dual verification, and tamper-evident audit trails.

Human-in-the-loop

HITL and the Cost of Saying Yes: Why Reviewer Approval Velocity Is the Wrong Optimization Target

HITL and the Cost of Saying Yes: Why Reviewer Approval Velocity Is the Wrong Optimization Target

Every HITL dashboard celebrates approval velocity — actions approved per hour, queue cleared per shift, throughput maximized per reviewer. But "yes" is the cheapest possible review decision. The system optimizes for the wrong thing. Real review quality means saying no when the action is wrong, asking questions when the context is unclear, and requesting changes when the action is incomplete. Here's why the cost of yes is the hidden tax on your HITL system.