PDF download link (best viewable locally with your local PDF reader instead of in-browser): https://solutions.servicenow.com/ls/2c5dbe1c-0663-450f-b42c-a7e0907534ff/HMx3B1Sz3BSfapTb
Use them, mark them up, and bring them into your own sessions.
Version 4's core idea is straightforward: apply agentic capabilities when non-deterministic reasoning is suitable, and deterministic approaches when consistency is essential. It is a fundamental reimagining of v3, presented on a single reference view so you can see the entire process rather than just the trendy parts. The aim was never to implement agentic methods everywhere - solving each problem in the most effective way is the actual goal.
What has changed since version 3? The previous version presented a "ServiceNow and Microsoft better together" perspective. Now, in version 4, the approach is vendor-neutral, showcasing Microsoft as part of an expanding partner ecosystem within an open catalog rather than the focus itself. Additionally, version 4 has been updated to align with the Australia release, integrating agentic and workflow automation across all domains.
By area:
- Customer and industry workflows (CRM) are structured around an outcome-driven value chain: Lead & Opportunity, CPQ, Order Fulfillment, Customer Service, Field Service, and Customer Success & Renewals, all based on industry data models such as BIAN, ACORD, HL7, and TM Forum.
- Service Operations (ITSM + ITOM) is structured around two closed loops: Predict-to-Correct and Request-to-Fulfill. Agentic capabilities sit on top of existing workflow: alert grouping and analysis, incident triage and resolution planning, and an autonomous L1 zero-touch path with self-healing. OT is shown honestly as an integration-and-workflow layer over third-party sensing.
- Multi-modal Change now visualizes its dynamic risk evaluation and approval policy engine, featuring agents for change risk, impact analysis, and conflict assessment. Control status and third-party vendor risk signals contribute to the change risk score, ensuring approvals are based on actual exposure.
- AI Control Tower (Govern): the most expanded area. The AI Control Tower anchors the Govern band: AI asset lifecycle, model/dataset/agent inventory, guardrails, and advanced AI risk assessments. An AI Agent Fabric with Model Context Protocol handles agent interoperability, and Now Assist runs on a bring-your-own-LLM footing.
- Operational Resilience & Risk: The IRM domains are organized into a unified, consistent matrix that includes IRM core, Business Continuity, Privacy, Third-Party Risk, Regulatory Change, Operational Resilience, and Audit, and covers the same lifecycle stages. Agentic support facilitates risk identification, issue resolution, and regulatory alert analysis with alert-to-control mapping, while AI/agent risk connections integrate with the AI Control Tower.
- Security Operations has been reorganized into an Exposure → Detect → Respond → Recover framework. Two key acquisitions are now part of it, highlighted to show integration: Armis, which provides ongoing visibility into cyber-assets and OT/IoT exposure, and Veza, which handles identity and access governance, covering non-human and AI-agent identities, enforcing least privilege, and conducting continuous access reviews.
- IT & OT Asset Management has been organized into five value stream rows based on lifecycle asset entities: Software (SAM), Hardware (HAM), Cloud Cost/FinOps, OT, and Enterprise (EAM). OT Asset Management is newly introduced, utilizing Armis for cyber-physical asset intelligence. HAM, OT, and EAM share a common agent foundation. Additionally, Contract Management has been elevated to a cross-cutting layer.
- Integrations: the large set of vendor logos in the sensing layer is a catalog of what you can connect, not a list of what you have to deploy. You pick the sources and connectors that fit your estate. The breadth is there to give you coverage and choice, and you are not signing up to run it all.
I refresh this artifact at least once a year as the platform evolves. If you've used an earlier version, I'd like to know what held up in practice and what you'd change. That's what makes the next one better.
