[object Object]

The Zurich release is the largest functional shift in two years and the upgrade-planning conversations reflect it. Customers debating whether to skip a release should not skip this one — Agentic AI, Build Agent, and the AI Control Tower together change how the platform is operated, not just what features are available. Reading the release notes once is not enough; the operational implications take a sprint to digest.

What Zurich Ships

Zurich reached full GA in March 2026 and is positioned as the start of the Agentic AI era for the Now Platform. The headline shift is from AI as a suggestion engine to AI as an execution engine — agents that take action within configured guardrails, not just propose actions for humans to take. Agentic Playbooks weave AI into individual tasks; the AI Control Tower governs the surface; Build Agent compresses the citizen-developer loop.

Zurich major capability surfaces:
  - Agentic Playbooks (multi-step autonomous flows)
  - AI Control Tower (governance plane)
  - Build Agent (natural-language app generation)
  - Process and Task Mining (workflow inefficiency detection)
  - Vault Console and Machine Identity Console
  - Multi-turn Virtual Agent context retention

Build Agent

The new Build Agent enables natural-language app generation on the platform — developers describe app requirements in plain language and the system generates a working application handling table design, logic, basic integrations, and baseline ATF tests. Accelerates app creation for both professional developers and citizen developers, with the same guardrails that App Engine Studio enforces today. The agent is a productivity tool inside the existing governance frame, not a license to skip it.

AI Control Tower

Governance nerve center for every AI agent across the platform. Monitor, manage, audit. Centralizes what was previously scattered across modules — an HRSD agent’s controls lived in HRSD, an ITSM agent’s lived in ITSM. Critical for compliance and operational confidence at scale; without a single pane, agentic deployment grows past auditable. The Control Tower exposes cost-per-agent, error rate, scope, ownership, and incident history in one console.

Agentic Playbooks

The platform’s pattern for embedding AI agents inside individual workflow steps. Each step in a playbook can suggest, execute, decide, or ask. Confidence thresholds gate autonomous action; below threshold, the agent degrades to ask-the-human. Workflow Studio is the authoring surface for new playbooks; legacy Flow Designer flows can call playbook steps but new playbook authoring should target Workflow Studio.

// Calling a playbook from a flow step
var playbook = new sn_agent.AgenticPlaybook('IncidentTriagePlaybook');
var result = playbook.execute({
  incident_sys_id: current.sys_id,
  confidence_floor: 0.75
});
if (result.escalated) routeToHuman(result.escalation_reason);

Process and Task Mining

The platform reviews actual workflow execution logs and surfaces inefficiencies — bottleneck steps, rework loops, rubber-stamp approvals, dead branches. Recommendations are draft proposals for the operations team, not auto-applied changes. The combination of mining (find inefficiencies) plus playbooks (automate confidently where appropriate) is the closed loop the release is selling.

Implementation Sequence

Plan a Zurich upgrade with three named workstreams: governance (AI Control Tower stand-up before broad agent enablement), security (Vault Console and Machine Identity Console adoption), and feature enablement (one playbook, one Build Agent pilot, one mining workflow). Trying to adopt every Zurich capability in the same sprint produces a wave of half-implemented features and burned operational credibility.

Common Failure Modes

Upgrading and turning on agentic features without standing up the AI Control Tower first — agents proliferate without governance and the audit team finds them. Build Agent enabled for all citizen developers on day one without scoping defaults — apps generated in Global scope accumulate technical debt fast. Treating Process Mining recommendations as auto-applicable — they are draft proposals; auto-applying them is how compliance controls get optimized away.

Cost Considerations

Agentic AI is gated to the Enterprise Plus tier and the per-call cost of agentic workflows is meaningful at scale. Establish per-agent cost caps and surface them on the AI Control Tower dashboards. The “AI is included” framing applies to many surfaces but the consumption-driven costs accumulate; budget conservatively and revisit monthly.

What to do this week: read the Zurich release notes for any module you operate and list the three capabilities most relevant to your roadmap; prioritize the upgrade discussion against that list, not the full feature surface.

[object Object]
Share