From AI chaos to AI control: key takeaways from ServiceNow Knowledge 2026

Bridging the “confidence gap”

Bill McDermott set the tone early by addressing the elephant in the room: AI chaos.

Many organisations have adopted AI through disconnected tools and chatbots that lack governance, organisational context, and meaningful control.

The next step is AI control, moving from AI that simply summarises information to AI that can act within a governed framework.

For the public sector, where accountability is non-negotiable, that shift isn’t optional. It’s essential.

1. Human-first innovation: The Idris Elba perspective

One of the highlights of the week was hearing from Idris Elba.

Beyond the star power, his message around human-first innovation really resonated with me. The technology may be powerful, but its purpose should always be to support people, not replace them.

In our world, that means AI shouldn’t just process citizens, tenants, or patients. It should empower them.

Whether it’s healthcare, housing, or local government, there’s always a human story behind the workflow. That requires empathy, accountability, and clear intent.

2. ServiceNow Otto and the rise of agentic AI

One of the most talked-about innovations at the event was ServiceNow Otto.

What became clear throughout the sessions is that Otto represents a major step toward agentic AI.

Many AI assistants today still behave like “expensive search boxes with manners”. Otto feels fundamentally different. Built on an AI-native architecture, it’s designed for autonomous orchestration across systems and workflows.

For customers, that means AI that doesn’t just answer questions about policy or process, it can execute actions across multiple systems while remaining tied to permissions and governance.

That’s the difference between a chatbot and a true digital worker.

3. Securing the “who” and the “what”: Armis and Veza

The demonstrations around Autonomous Security and Risk were another standout moment for me.

The acquisitions of Armis and Veza help bridge the gap between SecOps and IRM by connecting visibility, identity, and governance.

• Armis – the “what”: delivering visibility across connected assets, from IT and OT environments to IoT and medical devices.
• Veza – the “who”: bringing identity intelligence to understand exactly who, or which AI agent, has access to those assets.

Together, they create a much more complete security model. We saw examples where vulnerabilities could be identified, permissions validated, and remediation workflows triggered automatically, all while remaining aligned to organisational policy.

That combination of speed and governance is incredibly powerful for public sector environments.

4. Measuring AI ROI with AI Control Tower

One of the most significant announcements was the evolution of AI Control Tower.

For the first time, organisations can begin measuring the real ROI of AI activity directly within the platform.

AI spend has often felt like a black box, with value based on assumptions rather than measurable outcomes. Control Tower changes that by tracking AI activity, token usage, and workflow outcomes in one place.

It also provides governance across wider AI ecosystems, including integrations with platforms such as Microsoft Copilot, Google Gemini, Anthropic Claude, and OpenAI ChatGPT.

That means organisations can:

• Track AI spend against real outcomes and time savings.
• Reduce duplicated or overlapping AI tools.
• Identify unsanctioned AI agents operating outside governance controls.

For public sector organisations especially, that visibility is critical.

5. The platform as enforcer, humans as architects

One distinction came through consistently across the event: AI shouldn’t own policy, it should enforce it.

ServiceNow’s platform enables organisations to define the rules, governance, and intent, while AI agents execute tasks within those guardrails.

That’s the balance that matters most.

Humans still define the “what” and the “why”. The platform simply ensures the AI “how” stays within the lines.

Final thoughts

After a week of practical workshops, demonstrations, and hands-on sessions, one thing became very clear to me: AI is becoming the ultimate leveller.

It gives public sector organisations the opportunity to deliver service outcomes that previously required far larger teams and budgets.

At Phoenix Software, my focus is helping organisations navigate that transition successfully. Because ultimately, success won’t depend on AI alone, it will depend on the operating model, governance, and guardrails wrapped around it.

Knowledge 2026 made one thing certain: the era of AI chaos is ending. Governed, autonomous work is already here.