Since the meteoric rise of ChatGPT in 2023, the debate around AI has largely focused on what the technology can and can’t achieve. Indeed, there have been plenty of opinions on this in each direction – and there is every reason to think that debate will continue in earnest.
But in the boardrooms of large organizations around the world, the conversation is now quickly moving on. Now, executives are keen to roll out AI at scale and demonstrate ROI.
Even in 2025, that’s still easier said than done…
Governance: The Next Big AI Hurdle?
“In highly regulated industries such as banking, finance, and healthcare, concerns around PII and sensitive data make governance even more critical. But AI isn’t just about governance hurdles. It’s also about how different organizations view and prioritize adoption through their own unique lens, balancing risk, cost, and operational readiness.”
Mahathi Veena, ServiceNow Technical Architect and MVP
At this point, AI technology is widely available and commercially viable. While it’s not 100% reliable or effective, organizations are increasingly getting a sense for where it can and can’t be used and what value can be gained from it. But before they operationalize these new tools at scale, there are a number of very complex governance and compliance questions that first need to be answered.
Crucially, these have very little to do with the quality or reliability of the AI models themselves:
- Bias: To avoid reputational damage and poor-quality outputs, it’s important for organizations to understand, monitor, and mitigate the biases of their AI models. This can be difficult because AI models are quite opaque by design, making it difficult to understand how and why decisions have been made – and what assumptions they’re based on.
- Data protection: AI requires data, and a lot of it. But these models are subject to the same data protection requirements as non-AI systems. This includes existing laws like GDPR/CCPA, as well as newer AI-specific legislation like the EU AI Act. Under these frameworks, AI data is subject to several restrictions around data minimization, user consent, transparency, and anonymization.
- Data quality: AI is only as good as the data and information it’s trained on. Building effective, reliable, and auditable AI systems requires high-quality data that is complete, accurate, and de-duplicated. Today, getting high volumes of data is comparatively easy. But cleaning that data and removing errors is a much more difficult and time-consuming problem to solve.
- Risk management: By now, we’re all aware of the limitations of AI. Hallucinations (particularly in Generative AI models) are a common issue – and it’s important for organizations to understand the risk of models that simply get things wrong. These risks can’t be 100% eliminated, but organizations increasingly need to understand, quantify, and manage it before AI can be rolled out at scale.
- AI scope: If you put AI agents in charge of everything, there’s a high risk of errors, reputational damage, and poor decision-making. If you trust it with nothing at all, you risk falling behind and ceding a strategic advantage to competitors. Ultimately, all organizations need to balance risk with innovation by working out where they fit on this spectrum.
So here’s the situation as many organizations see it: The pressure is on to adopt AI and demonstrate measurable results from the investment. There’s plenty of effective technology out there and a number of business processes and use cases that are ripe for automation. But first, they need robust answers to the questions above.
That’s where ServiceNow comes in.
How to Solve the AI Governance Conundrum
“Many tech companies are trying to solve AI compliance, but often they do it in bits and pieces – either by focusing on individual models or bolting on monitoring tools. What makes ServiceNow different is its integrated approach.
The AI Control Tower and its ecosystem tracks AI assets, but also actively manages risks like bias and privacy – and even automates remediation within existing workflows. It’s more than just box-ticking, it’s about building trust into everyday operations.”
Mahathi Veena, ServiceNow Technical Architect and MVP
While AI products, apps, and models have been released thick and fast over the last couple of years, very few tech vendors can offer serious answers to the challenges listed above.
Fundamentally, these governance barriers are difficult challenges to solve. They require a lot of tools to monitor, manage, and centralize both AI tools and the data they’re built on. Effectively, a tech vendor will struggle to solve these problems if they don’t already have high-level visibility over most of the systems, apps, and models that are being used across the enterprise.
By now, you’ve probably realized where this is going. ServiceNow is uniquely placed to solve these difficult operational challenges, because its products are already deeply embedded in the back-end systems of so many large organizations.
In short, ServiceNow has a clear USP here, and the company isn’t shy about putting it front and center of its strategy. First of all, the company has published a raft of content explaining what AI Governance involves and how it works, including ‘Building an Enterprise AI Governance Plan’, ‘What is AI Governance?’, and the AI governance infographic. At the same time, executives have been increasingly discussing these governance problems in interviews and press releases.
But it’s not just words. Most importantly, the most high-profile product releases of the last year have been laser-focused on helping large enterprises solve these problems.
The Roadmap: How ServiceNow is Prioritizing AI and Governance
Over the last year, AI governance and compliance have been at the heart of ServiceNow’s product strategy. Arguably, these have been some of the most important and consequential announcements of the last 12 months.
Many of these are straightforward tools to manage risk and data protection across the many Generative and Agentic AI tools that ServiceNow has rolled out since 2023. But, crucially, the company’s ambition goes further, since many also incorporate third-party agents, models, and products.
Here’s a look at the most important AI governance tools released over the last year:
Now Assist
In November 2024, ServiceNow added several new tools and features to the existing Now Assist platform. Specifically, these were designed to monitor compliance and governance for the Generative AI features that had become a fixture in ServiceNow since 2023. The features include:
- Now Assist Guardian: A set of guardrails that monitor AI data privacy at ‘runtime’ (i.e. when AI tools are being used). The key goal is to avoid sensitive data being leaked via AI prompts or chat results.
- Now Assist Data Kit: A tool to manage the quality of datasets that AI models are trained on. This lets you create a single source of truth for data – eliminating issues and reducing hallucinations later down the line.
- Now Assist Analytics: These dashboards help to track adoption, usage, deflections, and performance of AI features.
AI Control Tower
Now Assist tools are principally designed for Generative AI features. But when it comes to Agentic AI (i.e. models trained to make strategic and context-driven decisions), a slightly different approach is needed.
To manage these models, ServiceNow released the ServiceNow AI Control Tower in May 2025 at its flagship annual conference: Knowledge. The AI Control Tower offers a number of features and tools to improve AI governance:
- Top-down visibility over all AI agents, models, and workflows in the enterprise, including both ServiceNow and third-party tools.
- Tools to help train AI agents to make context-based decisions, and enforce guardrails about when decisions should and shouldn’t be taken by agents.
- Data protection safeguards around personally identifiable information (PII) usage in AI models and products.
- Reporting tools to monitor the usage of AI models and quantify their ROI.
Zurich:
ServiceNow’s latest product release, Zurich, was released in September 2025. It features many new tools to help roll out and operationalize new AI tools across the organization:
- Build Agent: An AI tool to help non-technical people create apps without building the code from scratch, a practice known as ‘vibe coding’. ServiceNow’s unique approach is to bake governance in by design, with audit trails, security, and compliance tools all available.
- ServiceNow Vault Console: This helps to identify and protect sensitive data across enterprise workflows and AI tools. It principally aims to scan and detect PII, in order to reduce the risk of it being leaked through AI Agents or elsewhere in the workflow. This is similar to Now Assist Guardian, though more specific to Agentic than Generative AI.
- Machine Identity Console: A new tool to help protect your APIs and integrations from bot-related attacks and insecure inward connections. It features a number of checks to identify suspicious connections, such as validating the authentication method.
Solving the AI Compliance Conundrum
“As AI regulations evolve globally, platforms like ServiceNow that embed governance into operational workflows will be critical for sustainable, enterprise-grade AI adoption.”
Mahathi Veena, ServiceNow Technical Architect and MVP
The governance and compliance challenges we discussed in this blog are difficult problems that large organizations around the world are working hard to solve. If large and regulated organizations want to roll out AI at scale, answering these questions is effectively non-negotiable.
That answer won’t be the same for everybody, because risk looks different for every organization. But clearly, ServiceNow is putting a huge amount of emphasis into helping businesses understand what the answers are for them.
Today, it’s still early days for Now Assist and the AI Control Tower. ServiceNow is clearly leading the market, but in time, we can expect other tech vendors to try their hand at solving these problems in their own way. Until then, it’s clear that ServiceNow has gone much further than other tech companies in solving the operational barriers to AI adoption and deployment.