With Agentic AI, we have a unique opportunity to create a self-optimizing ecosystem of content and outcomes. Through ServiceNow's AI Platform, we’re reimagining how enterprise content interacts with AI to drive platform productivity and customer outcomes.
At Knowledge 2025, we tackled one of the most pressing issues facing enterprises today: why does AI often underperform in real-world scenarios? The session, "Fueling AI: Unlocking Content and Data to Supercharge Your Knowledge Base," co-presented by myself, Katie Ott, and a team of ServiceNow AI experts, provided insights into how organizations can dramatically boost AI effectiveness by focusing on structured, high-quality content.
Why AI Underperforms in Enterprises
Most AI implementations start with high expectations, yet many struggle to deliver value at scale. As we discussed in our session, the issue isn't primarily the technology itself—it’s the quality and structure of the data. Alexandr Wang, CEO of Scale AI, highlights this succinctly:
"Companies that invest in structured, high-quality data consistently outperform those relying on raw, unstructured data."
And the challenge is massive. According to Gartner, an estimated 80% to 90% of enterprise data is unstructured—living in PDFs, emails, chat logs, videos, and more. This vast ocean of unstructured information is largely inaccessible to AI, limiting its ability to extract insights, answer questions, or drive automation at scale.
This means most AI does not have access to the majority of enterprise data. We’re leaving massive value on the table simply because our content isn’t optimized for machine consumption.
The Three Pillars of AI Performance
AI performance hinges on three key pillars: Compute, Algorithms, and Data.
1. Compute: This is the backbone your infrastructure. Think GPUs, TPUs, and scalable cloud environments. At ServiceNow, this layer is robustly managed. Our platform is engineered for scale, speed, and security. We’re expanding our capacity through Raptor DB and seamlessly connecting more data across the enterprise via the Workflow Data Fabric.
2. Algorithms: These are the brains of the operation. Think LLMs, embeddings, clustering, and RAG pipelines. You're already leveraging them through tools like Now Assist, Virtual Agent, and Predictive Intelligence. ServiceNow not only supports our own models but integrates yours too, setting you up on solid ground from the start.
Optimized compute and advanced algorithms accelerate speed and quality of AI outputs. But here’s the reality: most organizations don’t control these two layers directly—they’re managed by hyperscalers or specialized teams.
3. Data: This is where the real bottleneck happens. Even the best models running on the fastest infrastructure can’t perform well if they’re fed outdated, inconsistent, or unstructured data. Among the three pillars, data is where enterprise AI success often stalls.
And here’s the opportunity: Data is the one pillar you can directly control. How well you curate, structure, and maintain your data pipeline directly influences AI performance. Invest here, and the other two pillars become exponentially more powerful.
Making Content AI-Ready
To transform content into a strategic asset, enterprises must rethink how they structure, manage, and govern content. During our session, we introduced the five critical tactics for AI readiness:
-
Structure your content for humans and AI: Clear metadata, predictable formatting, and structured tagging are crucial.
-
Provide clear, context-rich information: Enable AI to understand context through well-defined concepts and procedures.
-
Speak plainly: Plain, visible text is more valuable to AI than embedded or unstructured visuals.
-
Maintain high-quality operational records: AI depends heavily on the accuracy and cleanliness of inputs.
-
Continuously audit content: Regular governance ensures information remains current and relevant.
The ServiceNow Advantage
In ServiceNow, AI doesn't just pull from a knowledge base—it stitches together insights from case records, incident reports, community forums, and more. Yet, this intricate ecosystem only thrives when signals are clean, content is structured, and everything remains up-to-date.
We demonstrated this vividly through the Agentic AI Assessment app, developed by our partners at Work4Flow to diagnose and enhance your AI readiness.
Practical Framework for AI Readiness
Our structured approach to AI readiness includes:
-
SCAN your current content and configuration signals.
-
DIAGNOSE issues and gaps hindering performance.
-
REMEDIATE content using targeted playbooks.
-
ENRICH your knowledge base by auto-generating content from resolved tickets.
-
TRACK improvements continuously through actionable dashboards.
In practice, enterprises following this framework have seen substantial improvements—like a 40% increase in AI search accuracy and significant boosts in incident deflection.
Taking Action
AI success is achievable, but it starts with the foundation of structured content. By focusing on data hygiene, structured tagging, and continuous governance, enterprises can shift from merely experimenting with AI to truly scaling its benefits.
Explore the tools and tactics we discussed, to jump-start your AI transformation journey. Let's fuel AI with the right data to drive real, measurable outcomes.
Additional Resources:
- Session slides attached to this article.
-
Best practices to use your knowledge articles with Now Assist (generative AI) by Ashley Snyder
Join the conversation and share your AI readiness journey. How structured is your content today, and what's your next step?
With Agentic AI, we’re not just preparing content; we’re creating a self-optimizing loop where content drives outcomes, and outcomes drive content evolution—a true paradigm shift in enterprise AI readiness.
