TL;DR
- Vibe coding for everyone – natural language app creation, but with enterprise-grade quality, so teams get real tools to boost productivity.
- Security and context as pillars – protecting sensitive data, generating contextualized outcomes, and tracking not just human work but digital workers too.
- Turning traditional workflows into autonomous ones – freeing up teams’ time and reallocating it to higher-value work.
After publishing our article on the AI Control Tower, we realized something was missing. It’s useful to look at that product in detail, but ServiceNow didn’t just launch a single feature, it introduced a whole suite of tools that work best together. They cover the full lifecycle of AI agents: building, tuning, monitoring, and governance. They’re also key for making multi-agent systems work properly, avoiding overlap, ensuring all agents operate with the same context, and defining where human intervention is still required.
The Zurich release confirmed the direction Bill McDermott has been pointing to: taking enterprise automation to the next level with agentic AI and embedding agents into Now Platform workflows. But not at any cost. This release put special attention on protecting sensitive data with the new Vault Console, and added stronger security and compliance features to manage the risks that come with more autonomy.
So in this article we’ll review how ServiceNow frames the full lifecycle of agents—and why that matters for enterprises preparing for multi-agent systems.
What’s New with Zurich Release?
The Zurich release marked a turning point for ServiceNow, not only because it closed the alphabetical cycle of releases but because it pushed the platform deeper into the agentic AI era. All of the new capabilities announced are generally available now through the ServiceNow Store, which means customers don’t have to wait to start experimenting.
Amit Zavery, ServiceNow’s President, COO, and CPO, called it “a pivotal step in the race to scale AI beyond experiments and into tangible business value.” The message was loud and clear: a new era of productivity is coming, but above all, one of experimentation.
ServiceNow is setting the stage for teams to generate value using plain language to spin up new apps or agents in just minutes – and all with the security and context that ServiceNow provides as the platform that connects every other platform in the enterprise.
Let’s look at the new products announced in the Zurich release:
- Build Agent + vibe coding: Employees can turn natural language prompts into production-ready apps—“an onboarding app that tasks HR, IT, and Facilities” is the demo example ServiceNow used. Unlike other low-code platforms, every app includes audit trails, compliance, and security from the start.
- Developer Sandbox: A safe playground for experimentation. Teams can build and test without breaking production environments or exposing sensitive data.
- Vault Console: A guided workspace for discovering, classifying, and protecting sensitive data such as PII. It brings together encryption, anonymization, and policy templates in one place, instead of scattered across different tools.
- Machine Identity Console: Centralized visibility into all API integrations and machine credentials, reducing risks tied to service accounts and outdated authentication methods. Amanda Grady, VP of AI Platform Security, emphasized that “it’s really important that the scope of these integrations is limited only to what’s necessary.”
- Agentic Playbooks: New workflows that embed AI agents into existing processes but preserve human oversight. ServiceNow often uses the credit card fraud example: agents freeze the card and send a replacement, while humans can still intervene when judgment is required.
- Task & Process Mining: Integrated directly into the platform, these tools reveal how work gets done across the enterprise, showing where humans add value and where AI agents could accelerate execution.
Together, these features strengthen the foundation laid by the AI Control Tower earlier in 2025, giving enterprises both visibility and guardrails while moving from pilots to scaled adoption.
Three big themes that define the strategy
1. Democratizing development without losing rigor
The idea of democratizing apps is really about strengthening experimentation. The goal is to give teams and developers a safe testing ground where they can build apps or agents without the risk of affecting other parts of the system. And because the tools are available inside a governed environment, you also avoid problems like the growing shadow AI we talked about in this Reddit post.
2. Strengthening the foundation of trust
ServiceNow also doubled down on two fundamentals: security and context. With the new Vault Console, admins get a clearer way to spot and track personal data that needs to stay protected. It helps separate what can be used in a specific workflow from what must remain off-limits elsewhere. Without that layer, AI agents could easily turn data flows into a mess—acting on information that was never meant to be shared.
The Machine Identity Console takes it further by shining a light on every API connection in play, even the ones agents and bots call behind the scenes. For a CIO, this brings valuable tools to avoid compliance headaches and gain a clearer view of API usage and costs.
3. Blending autonomy with human oversight
Another key direction is the way ServiceNow balances automation with human judgment. The goal isn’t to cut people out, but to let AI agents handle the routine steps while humans step in where nuance or empathy is needed.
In the Zurich release, ServiceNow illustrated this with a fraud management scenario. An AI agent can instantly freeze a stolen credit card, trigger a replacement, and notify the customer. That’s speed no human team could match at scale. But when it comes to gray areas—like a flagged transaction that might just be a customer traveling—humans still make the call. That balance makes automation safer and more reliable in practice.
Tools or Teammates? How the ServiceNow AI Agent Stack works
In the Zurich release notes, Jon Sigler, EVP and GM of the ServiceNow Platform and AI, emphasized that AI shouldn’t be treated as just another tool, but as a teammate: “The future of the agentic enterprise belongs to organizations that seamlessly blend human expertise with AI capabilities. ”
Equally important, Dave Wright, Chief Innovation Officer at ServiceNow, pointed out in a recent article that companies need to strategically reallocate the time freed up by AI rather than simply seeing it as an opportunity for cost-cutting. He also highlighted an uncomfortable truth: “According to McKinsey, while 78% of organizations report using AI, over 80% of them saw no tangible enterprise-level impact on earnings. A study by BCG paints a similar picture: only 4% of companies create substantial value from AI initiatives, while 74% struggle to show any value at all.”
That’s exactly the cycle ServiceNow wants to break with its AI stack.
The strategy has a simple goal: helping business teams create value from their own expertise. Take an HR manager who can launch an onboarding agent that assigns tasks across IT, HR, and Facilities—without waiting for developers or writing code. The point isn’t just speed. It’s about doing it in a safe environment where AI experts can check the app before it goes into production.
This also means creating a culture where people feel free to test ideas. Some worry that adding more control slows innovation, but ServiceNow is taking the opposite approach. By giving teams, the tools to experiment in a governed space, they can try new things without compromising security or compliance.
When someone submits a new idea, it’s logged in the AI inventory and reviewed by experts, usually in an AI Center of Excellence. Their role isn’t to stop projects but to make sure they’re safe and useful. They look at things like:
- Does it use sensitive or personal data?
- Could it create bias or unfair results?
- Are the system’s decisions transparent enough?
- What rules or internal policies apply?
This way, experiments keep moving, but with the right guardrails. Innovation continues, and companies avoid problems like shadow AI or projects that collapse later because risks weren’t considered.
Inside the ServiceNow Agentic AI Stack
Here’s how the different pieces of the stack work together (and why they matter for enterprises preparing for multi-agent systems):
Build Agent + Developer Sandbox
These two go hand in hand. Build Agent allows teams to create apps or agents just by describing them in plain language. For example, an operations manager could request “set up a workflow to track equipment maintenance,” and the agent handles design, logic, and compliance automatically. The Sandbox gives developers and business teams a safe place to test those apps without breaking production. For a CIO, this translates into faster delivery cycles with governance built in from the start.
Configuration Management Database (CMDB) + Workflow Data Fabric
Agents can’t work in isolation; what they need is context. The CMDB acts as the enterprise map, showing how systems, processes, and teams are connected. Workflow Data Fabric takes this further with its zero-copy architecture, allowing AI to access information across sources without risky data migrations. Together, they ensure that agents act with the right context and can actually coordinate across departments.
Task and Process Mining
The Zurich release added more visibility into how work really happens. Task Mining captures what people do—every click, every step. Process Mining looks at how entire workflows move through the organization. Combined, they reveal where human expertise is essential and where agents can make the biggest impact. For leaders, it turns “gut feeling” into data-driven decisions about where to apply automation.
Vault Console + Machine Identity Console
Security and trust are non-negotiable. Vault Console helps admins discover, classify, and protect sensitive data, like personal information that should only be used in a specific workflow. Machine Identity Console provides visibility into every API and bot integration, making it easier to spot risks and prevent shadow integrations from spiraling out of control. For CIOs, this means stronger compliance and fewer surprises.
AI Agent Fabric
This is the communication backbone. It ensures ServiceNow’s agents can talk to each other and also to third-party agents from platforms like Adobe, Microsoft, or Zoom. It uses protocols like MCP and A2A so that agents can coordinate tasks, share context, and act in real time. Instead of isolated tools, companies get a multi-agent system that behaves like a connected workforce.
AI Orchestrator
If Agent Fabric is the communication layer, the AI Orchestrator is the coordination brain. It decides which agent should act, in what order, and with what context. For example, in a customer service scenario, one agent may pull the customer’s account data, another may verify identity, and a third may generate the response. The Orchestrator makes sure these agents don’t step on each other and that the workflow runs smoothly from start to finish. For CIOs, this avoids duplication of effort, ensures consistency, and turns a set of disconnected agents into a real AI workforce.
AI Control Tower
The Control Tower sits on top of the stack to provide visibility and governance. It lets companies monitor which agents are running, how they’re performing, and whether they meet compliance and security standards. Think of it as the guardrails that allow experimentation without chaos—enterprises can innovate confidently because they know every agent is visible and accountable.
Agentic Playbooks
This is where the stack meets day-to-day execution. Playbooks embed AI agents into traditional workflows, automating steps like verifying identity or processing service requests while keeping humans in the loop for exceptions. In Zurich, ServiceNow highlighted the fraud management case: an AI agent can freeze a stolen credit card instantly, while a human decides on special cases. It’s a practical example of autonomy with oversight.
Conclusion: Humans Still Hold the Hardest Job – Find Value Opportunities
Even as AI takes on more tasks, the hardest work still belongs to people. Technology can automate steps, monitor processes, and even build apps in minutes – but it can’t decide where the real opportunities lie. That’s still a job for human workers: spotting ways to cut costs, designing new revenue streams, or simply unblocking a workflow that slows the entire business down.
ServiceNow’s stack is designed to encourage this shift. By giving teams safe environments to experiment and embedding governance at every step, it promotes a culture of innovation without fear. Employees can try, test, and iterate – knowing their ideas won’t break the entire system or put sensitive data at risk. That cultural foundation is just as important as the technology itself.
At Inclusion Cloud, we share this vision. As ServiceNow partners, we help organizations embrace multi-agent systems in a way that’s both secure and practical. Our role is to accelerate adoption and make sure the stack delivers tangible value for your business.
Visit our ServiceNow services page to see how we can help you put this ecosystem into motion and build a workplace where humans and AI agents work side by side to create real impact.
