ServiceNow's Q4 beat - how solving AI's governance bottleneck is winning enterprise budgets

Today ServiceNow announced its Q4 and FY 2025 results, highlighting a year of consistent performance in an industry full of AI noise. The company beat expectations across all topline growth and profitability metrics in Q4, with subscription revenues of $3.47 billion representing 21% year-over-year growth (19.5% in constant currency). For the full year, subscription revenues reached $12.88 billion, also up 21% year-over-year, with total revenues hitting $13.28 billion.

Current remaining performance obligations reached $12.85 billion, up 25% year-over-year, while total remaining performance obligations hit $28.2 billion, up 26.5%. The company closed 244 transactions over $1 million in net new annual contract value during Q4, nearly 40% growth compared to last year, and ended the year with 603 customers spending more than $5 million annually, representing approximately 20% year-over-year growth.

And sitting down with ServiceNow Chief Product and Operating Officer Amit Zavery ahead of the numbers being released, it’s clear that there’s a lot of important enterprise context that financial analysts might not address. It became clear during our conversation that ServiceNow is running a very similar playbook to the one that made it successful over the past decade, just adapted for the AI era. The "platform of platforms" approach that allowed ServiceNow to orchestrate workflows across fragmented enterprise systems is now being positioned as the AI Control Tower that orchestrates AI agents across those same fragmented systems.

It's a clever approach and one that plays directly to ServiceNow's strengths. While competitors are building departmental AI capabilities or chasing differentiation through model performance or claiming they have the whole stack, ServiceNow is trying to solve the enterprise-wide governance challenge that's actually preventing AI adoption at scale. What used to be "we'll help you get work done across all your systems" is now "we'll help you safely deploy AI across all your systems."

I've been writing about ServiceNow's platform advantages for a decade, and this feels like the moment where that investment could help buyers surface that enterprise-wide AI ROI that they are desperately seeking.

The platform play adapted for AI

Prior to ‘AI’ as we know it, ServiceNow had positioned itself as the layer that sits above departmental applications and enabled work to flow across them. It wasn't trying to replace your ERP or CRM or HCM - it was the workflow engine that connected them and let employees get things done without navigating multiple systems. That "platform of platforms" approach required deep integrations, a unified data model via the CMDB, and cross-functional workflow capabilities that most vendors couldn't replicate because they'd built departmental solutions.

Now ServiceNow is leveraging that integration capability and knowledge for AI. AI Control Tower, launched early last year, positions ServiceNow as the layer that sits above all your AI systems - whether they're from ServiceNow or third parties - and gives you visibility, control, and governance over how those AI agents operate. Zavery explained the value proposition when I asked about how AI Control Tower fits into the broader strategy:

We can discover all your AI systems, third party, ServiceNow, it doesn't matter. We're heterogeneous and we'll do that end-to-end. Then we give buyers the ability to manage AI proliferation - not just discover it, but who's doing what? What are AI systems allowed to do? How much are you paying for it? What results did you get? What controls do you have around it?

This is important because enterprises aren't going to standardize on a single AI vendor any more than they standardized on a single application vendor. They're going to have AI agents from multiple sources operating across their environment. Someone needs to orchestrate that complexity, and ServiceNow is betting that the same capabilities that let it orchestrate workflows will let it orchestrate AI agents.

The technical foundation makes sense. ServiceNow already has the CMDB that maps enterprise assets and their relationships. It already has the integration framework for connecting disparate systems. It already has the workflow engine for coordinating actions across multiple applications. Extending all of that to cover AI agents and their permissions is a natural evolution, not a fundamental rebuild.

However, the sell is different. Workflow orchestration was about efficiency and user experience. AI orchestration is about governance and trust. According to Zavery, that shift is important for customers:

When we started having that conversation and people started implementing AI Control Tower, they said, 'Now I can open up new agentic use cases. I can start investing in a lot of new things I couldn't before.'

The governance infrastructure removes the barrier that, in theory, has been keeping organizations cautious about AI deployment. This aligns with what we’ve been hearing from CIOs in the diginomica network. They're not struggling to find AI use cases. They're struggling to get past the security challenges and are finding it difficult to establish an answer to an enterprise-wide solution, without backing away from sunk investments. It’s worth remembering that AI in isolation will only ever generate limited returns, which is why we are seeing tension in the market around who will ultimately accrue the value. 

Security acquisitions 

2025 was ServiceNow’s most aggressive year for M&A to date. It made a number of acquisitions, many of them focused on filling gaps for AI - but perhaps not in areas you might expect. The company announced its intent to acquire Veza for identity security and Armis for $7.75 billion to address cyber exposure management. These aren't just ‘AI companies’. They're investments in the specific capabilities ServiceNow needs to credibly claim it can govern AI deployments across the entire enterprise.

I asked Zavery about this, particularly whether governance and security were proving to be bottlenecks for AI adoption. His response confirmed what I suspected. Customers were pushing ServiceNow to manage identity and access for AI agents:

Customers were asking, 'You're doing all these things…but can you also make sure user identity is managed [at a granular level]? Especially now with non-human identity with agents, because their identities change very fast depending on the use case and request."

ServiceNow's security and risk business has also been growing faster than expected, crossing $1 billion in annual revenue a couple of quarters before they'd planned. Zavery said:

Security and risk is probably the biggest problem people are facing in AI, and none of these large language models or hyperscalers solve those problems.

This is where ServiceNow's platform approach creates some advantages, in theory. If you're orchestrating AI agents across the enterprise, you need comprehensive visibility into what those agents can access, what actions they're authorized to take, and how to audit their decisions. Veza provides the identity governance piece. Armis provides visibility into the full attack surface including operational technology and IoT devices. Combined with ServiceNow's existing CMDB and workflow capabilities, you get something that looks like decent AI governance.

As I wrote following the Armis acquisition announcement in December, ServiceNow is systematically answering the question "Why wouldn't you choose ServiceNow?" for enterprise AI management. Each acquisition plugs a capability gap that might otherwise give buyers reason to look elsewhere or force them to cobble together multiple point solutions.

The platform play only works if you have comprehensive capabilities. ServiceNow is building that comprehensiveness through acquisition because developing everything internally would take too long - and as Zavery put it, the company has other things to focus on. 

Enterprise productivity

Zavery also pointed to how ServiceNow's platform advantages matter for AI in ways they didn't for traditional applications. I asked him to explain the difference between ServiceNow's "autonomous employees" (multi AI agents working in coordination to complete tasks that replicate what an employee does) and the AI agents other vendors are building. His answer:

As a human, you have multiple skills and you can do things. Even though you might be specialized in one area, you understand the context around many other things. We bring in the context of enterprise use cases and 20 years of workflow data.

Then he added:

A lot of the tools out there are all about employee productivity - I can help you find content, I can summarize this. But when you free that up, it doesn't guarantee that the enterprise will be productive. Can I do this at an enterprise level?

This is likely a nod to the departmental application vendors that are surfacing AI in their systems and the likes of Microsoft and Google - Individual productivity tools that might help employees work faster, but they don't necessarily translate to enterprise-level efficiency gains. ServiceNow is arguing you need end-to-end workflow automation with AI embedded throughout, not AI tools bolted onto existing processes.

The argument is that when you have cross-functional workflows, a unified data model, and two decades of workflow patterns, you can build AI capabilities that understand enterprise context rather than just departmental tasks. An autonomous employee in ServiceNow can handle a laptop provisioning request that touches IT, procurement, finance, and security because the platform already connects those functions.

This is harder for competitors to replicate than you might think. Building a CRM copilot that helps sales reps is straightforward. Building an AI capability that can coordinate actions across CRM, ERP, ITSM, and HCM requires the kind of cross-functional integration that ServiceNow has been building for years.

Zavery gave an example of autonomous IT, where ServiceNow eliminates level one support by handling service requests automatically end to end: 

Maybe ordering a laptop, password reset, login requests, completely end-to-end, done through an autonomous employee. Once we show that, budget shows up.

Where the budget actually comes from

The platform positioning also helps with something I've been trying to understand about how buyers will consider budget for enterprise AI. Who pays for enterprise AI when there are competing agendas, institutional knowledge, investments and relationships? When I asked Zavery about budget allocation, particularly given that competitors like Salesforce target customer-facing budgets while ServiceNow has IT heritage, his answer revealed how AI spending differs from traditional software purchases:

Budget is coming from many angles. CIOs have budget. CISOs have budget. Line of business owners want to make their departments more efficient and AI-first. You're seeing a lot of alignment at the C-suite level that this is an area they want to invest in, as long as you can show them value and great outcomes.

This C-suite alignment around AI investment creates opportunities for vendors who can demonstrate enterprise-wide value rather than departmental improvements. ServiceNow's platform approach naturally addresses enterprise-wide challenges because that's what the platform was built to do. You're not pitching point solutions to individual departments. You're pitching governance infrastructure that lets the entire organization deploy AI safely.

Partnerships that reinforce 

This week ServiceNow also announced partnerships with Anthropic, OpenAI, Microsoft, and others alongside the earnings release. On the surface, these look like standard enterprise software announcements - dime a dozen. But Zavery's comments show how the partnership strategy reinforces ServiceNow's platform play rather than threatening it.

The core principle is the same as it's always been for ServiceNow: 

We support any system out there. Any cloud, any data sources, any LLMs, any persona, any industry, any data models.

I probably sound like a broken record at this point, but ServiceNow is placing itself at the center of this ecosystem (well, it is trying to) and if it can do that - and convince buyers that governance is key - it sees an opportunity. 

As CEO Bill McDermott told me last year:

We'll have the control tower that works with all those agents. Our agents will control those agents in the form of a business process or an automated workflow.

The Anthropic partnership provides a good example of how these partnerships work in the AI era. Claude powers ServiceNow's Build Agent for application development, but customers who prefer different models can choose them. ServiceNow maintains the orchestration layer regardless. Zavery explained they do prompt engineering across multiple LLMs to deliver the same outcomes, but there might be specific use cases where one model works better:

Anthropic is pretty good at code generation. They've really been focusing on it. Our Build Agent can benefit from that. 

But if customers don't want to use Claude and prefer Gemini instead, ServiceNow is fine with that. The value isn't in the model - it's in the orchestration and governance layer.

This mirrors exactly how ServiceNow has always approached partnerships. They'll integrate with anyone and support any system, but ServiceNow remains the workflow layer that sits above everything. Now that workflow layer is becoming the AI governance layer, but the strategic logic is identical.

Whether this holds up as partners pursue their own enterprise orchestration ambitions is an open question. Microsoft in particular isn't going to hand over AI orchestration to ServiceNow without a fight. But ServiceNow's platform foundation and cross-functional capabilities give it advantages that will be hard for departmentally-focused vendors to replicate quickly.

My take

What was workflow orchestration across systems is becoming AI orchestration with governance, and that shift appears to be unlocking enterprise budgets. The growth in Now Assist ACV and deals over $1 million and $5 million appear to validate this. 

If ServiceNow can help solve the enterprise-wide governance challenge that actually determines whether AI gets deployed at scale, it could have a strong hand to play. CIOs might champion AI initiatives, but CISOs control whether those initiatives go into production. ServiceNow is positioning itself to answer the CISO's questions.

However, this partnership strategy also has inherent tensions built in. ServiceNow says it is agnostic, supporting any model, any cloud, any system. But the underlying ambition is clearly to own the orchestration layer that sits above everything else. Competitors aren't going to hand over that position, and we're already seeing aggressive responses from Microsoft, Salesforce, and others who recognize what's at stake.

For buyers, ServiceNow's platform approach does address real challenges. I'm hearing from CIOs consistently that they need enterprise-wide visibility and governance for AI deployments, not just departmental capabilities. They need answers to questions about control, security, and audit trails before they can move beyond pilots. ServiceNow has built infrastructure to provide those answers, and the platform foundation means those capabilities extend across the enterprise rather than remaining siloed in individual departments. And for buyers, ServiceNow has history with them in this way, way before AI. 

But the proof remains in production deployments that deliver measurable returns, not impressive demos or strong quarterly results. As Zavery noted, the test is straightforward - does it work, and can you prove it? The next twelve months will be telling.