Agents Enter Incident Workflows, Identity Becomes Continuous Control, and SDLC Turns Governed

PagerDuty, Auvik, and ConnectWise Put Agents Inside the Incident Workflow

PagerDuty’s Nov. 3 launch of Paige, Auvik’s Nov. 5 automation framework, and ConnectWise’s Nov. 2 RMM agent show the next step in the same progression: agents are now being embedded directly into the incident workflow. PagerDuty’s SRE agent can detect, investigate, and remediate incidents, open incident channels and conference bridges, auto-resolve predefined low-risk issues, pause for approval on riskier actions, and generate post-incident reviews. Auvik’s agent analyzes alert history to cut recurring noise by up to 80%, uses topology, configuration history, and syslog data to investigate, and recommends device-specific diagnostic commands. ConnectWise’s agent investigates the device tied to a ticket, runs read-only diagnostics, tests hypotheses, and stages fixes from a vetted catalog, with technician approval required for any device-changing action.

The practical implication builds on last week’s governance controls: the market is no longer just wrapping AI remediation in guardrails, but placing agents inside the incident lifecycle itself. For working operators, the leverage now sits in approval logic, action allowlists, rollback paths, and exception handling. Teams that can convert troubleshooting knowledge into governed remediation playbooks will move faster without expanding blast radius or sacrificing auditability.

HKMA and Cloud Vendors Turn Agent Identity Into Continuous Access Control

On Nov. 2, Hong Kong’s HKMA turned agent governance from guidance into an operating requirement with a proposed risk-based “Know Your Agent” regime for banking AI agents. It demands a distinct agent identity, verified attribution to a human or institution, and auditable records of delegated authority, scope, duration, and revocation. That is a sharp escalation: banks must prove not only that an agent is legitimate, but that every action stayed inside an explicit mandate and could be suspended or revoked.

The rest of the week showed the stack aligning to that model. On Nov. 5, Google Cloud introduced centralized agent credentials through Auth Manager and previewed Agent Identity JWTs. On Nov. 3, Snowflake added authentication, fine-grained authorization, identity propagation for every MCP/tool call, and observability across models, tools, decisions, and traces. On Nov. 5, CrowdStrike made AI Agent Discovery generally available for ownership and shadow-AI visibility, and on Nov. 9, Microsoft expanded Entra with sponsor requirements, inactivity monitoring, periodic reconfirmation, and drift reporting for agent access.

For IT teams, this is the next step beyond overprivileged runtimes: agent work is shifting from provisioning to continuous identity operations. The people who can connect IAM, SIEM, and AI governance into one control loop will be the ones setting the standard for secure deployment.

Governed Agentic Workflows Become the New SDLC Control Plane

GitLab this week expanded Duo Agent Platform to cover planning, coding, review, security, and CI/CD, turning AI from a coding helper into an execution layer for the software delivery lifecycle. The release adds goal-driven workflows through the /goal command in Duo CLI, Agentic Chat, and headless mode, plus event-triggered automation for merge conflicts, draft-to-review transitions, approvals, and new work items. It also extends support to external MCP-compatible agents such as Claude Code and Cursor, which can discover GitLab resources, inspect CI/CD artifacts, and invoke GitLab agents and flows.

GitLab paired that autonomy with controls: admins can set tools to Always Allow, Always Ask, or Always Deny by role and action category at the instance, group, and project levels, while logging sessions, tool calls, approvals, and anomalous activity for centralized review or SIEM streaming. That matters because GitLab says Duo Agent Platform weekly active users grew 10x after January 2026 GA, while its April 2026 survey found 85% of DevSecOps professionals saw AI shift bottlenecks to review and validation and 80% said governance lagged adoption.

For teams, the work shifts toward exception handling, policy tuning, and audit review. Career value now comes from AI governance, change control, and SIEM-ready oversight, not just coding or pipeline administration.