Metrics, Semantics, and Knowledge in the Age of AI

Data teams and business users face a paradox. On one hand, modern, self-serve analytics tools and AI promise speed and independence---the ability for anyone to ask a question and get an answer without waiting for a data analyst to build a custom report. On the other hand, organisations need to know their numbers are accurate, their definitions are consistent, and their decisions are based on the same shared understanding of what the data actually means.

At the heart of this tension is something called metadata and semantics---a fancy way of saying: the agreed-upon definitions, formulas, and relationships that turn raw numbers into meaningful business insights. For example, does "revenue" include refunds? Does "active user" mean someone who logged in this week, or this month? Without clear, shared answers to these questions, different teams end up with different numbers---and nobody can agree on what's actually happening.

The debate has intensified as AI-powered tools promise to skip traditional data layers entirely. But can AI truly deliver reliable insights without a foundation of governed business logic and rich contextual knowledge? The short answer is no---and this guide explains why, while charting a path forward.

The future calls for reinventing our data foundations as AI-native, flexible, and dynamic---combining the precision of a metrics layer (which ensures consistent definitions and calculations) with the contextual intelligence of a knowledge graph (which captures the relationships, strategies, and institutional knowledge that give those numbers meaning).

Think of it as building your AI a business analyst who not only knows the exact definitions and calculations of every metric (the metrics layer), but who has also been with the company since day one and understands the strategy, relationships, and larger context of how the business operates and wins (the knowledge graph).

In this guide, we will:

• Explore the difference between rigid and flexible data foundations

• Examine the tension between governance and agility

• Clarify what a true AI-ready knowledge ecosystem actually looks like

• Investigate AI's role as both a tool and a reasoning agent

• Chart a path toward a metrics layer enhanced by a knowledge graph

1. The "Right" Foundation vs. The "Wrong" Foundation

Not all data foundations are created equal. The problem isn't the concept---it's the incomplete execution.

The wrong approach relies solely on rigid metric definitions without context. Imagine a rulebook that tells you exactly what every number means, but nothing about why it matters or how it connects to other parts of the business. Teams get accurate numbers but lack the context to understand what those numbers are telling them---or what to do about it.

The right approach combines two complementary layers that work together:

The Metrics Layer --- the "what" and "how":

• Consistent definitions: Standardised calculations for all key performance indicators (KPIs) and metrics---so "revenue" always means the same thing, no matter who's asking

• Data integration: Simplified connections across different data systems (your CRM, your finance platform, your product analytics) so users don't have to stitch things together manually

• Reporting enablement: Reliable foundations for dashboards and analysis that people can trust

The Knowledge Graph --- the "why" and "so what":

A knowledge graph is essentially a structured map of how business concepts relate to one another---linking metrics to strategies, customers to behaviours, products to outcomes. Think of it as your organisation's institutional memory, made accessible to both people and AI.

• Context-aware reasoning: Understanding of how business concepts and relationships connect---for example, knowing that a spike in customer support tickets often precedes a drop in renewals

• Connected insights: Links between disparate data points (e.g., marketing spend → qualified leads → revenue)

• Natural language capability: Rich context that allows AI to interpret questions the way a knowledgeable colleague would

The result is a modular design with the appropriate governance guardrails. Teams can use either layer independently or both together, without rebuilding everything from scratch. Version control, approval workflows, and certification processes ensure that both metric accuracy and contextual integrity are maintained over time.

As data leader Kimmo Santamaa notes, "The endgame for the semantic layer is to NOT lock you in a pattern." The same principle applies to knowledge graphs---they should enhance understanding, not constrain exploration.

2. Governance and Trust vs. Flexibility and Innovation

A robust data foundation serves two critical---and sometimes competing---missions. Understanding both is key to building something that actually works for your organisation.

Ensuring Consistency:

• Unified terminology: One definition of "Net Revenue," used everywhere---so the sales team, the finance team, and the board are all looking at the same number

• Single source of truth: A centralised metric catalog that replaces the chaos of competing spreadsheets and ad hoc calculations

• Auditable lineage: The ability to trace how data transforms from raw source to business insight---essential for regulated industries and for building stakeholder trust

Providing Context:

• Business relationships: Understanding how metrics connect to strategy and outcomes (e.g., why monthly active users matter differently at different stages of company growth)

• Historical knowledge: Patterns, trends, and institutional memory---the kind of "we tried that in 2021 and here's what happened" knowledge that often lives only in people's heads

• Strategic alignment: Connecting day-to-day metrics to long-term objectives, so teams are optimising for the right things

Without metric consistency, organisations face conflicting numbers and misaligned decisions. Without contextual knowledge, they get accurate measurements of the wrong things---or miss the bigger picture entirely.

The key is striking the right balance between three often-competing needs:

• Governance and trust: Essential for large-scale organisations, regulated industries, and anywhere that decisions have significant financial or operational consequences

• User autonomy and agility: Critical for teams that need to move quickly, test hypotheses, and explore data without waiting in a queue

• Contextual intelligence: Necessary for AI to provide recommendations that are not just accurate, but actually useful

Effective foundations act like expert advisors, not rigid rulebooks: they provide both precise definitions and rich context while leaving room for exploration and discovery.

3. Beyond the Semantic Layer: A Complete Knowledge Ecosystem

You may have heard the term "semantic layer"---a technical term for the translation layer that sits between raw data and the people (or AI tools) that want to use it. Think of it as the glossary and grammar rules for your organisation's data language. One major limitation in industry discussions is the narrow focus on semantic layers alone.

While metrics layers are essential, they represent only half of what AI needs to deliver truly intelligent insights.

The Metrics Layer handles the "what" and "how":

• What constitutes revenue, churn, or customer lifetime value

• How to calculate these metrics consistently across systems

• Where to source the underlying data

The Knowledge Graph provides the "why" and "so what":

• Why certain metrics matter for specific business objectives

• How different business concepts and entities relate to each other

• What actions typically follow from specific metric patterns

• Which external factors influence internal performance

However, at the heart of any successful foundation are well-defined metrics. While the knowledge graph provides essential context and relationships, the metrics layer serves as the foundational truth---the concrete business calculations that drive decisions. Everything else in the ecosystem exists to support, contextualise, and enhance these critical metric definitions. This metric-centric approach ensures that all analytics work flows from standardised, trusted business calculations while being enriched by institutional knowledge and business context.

A successful AI-native strategy requires both layers working in harmony. The metrics layer ensures consistency and accuracy; the knowledge graph enables intelligence and insight.

4. The Role of AI: From Tool to Intelligent Agent

A common question: "Can AI bypass these data foundations entirely and just figure things out on its own?"

Short answer: No. But the real question is more nuanced: how can AI leverage both precise definitions and rich context to become truly intelligent---more like a knowledgeable business partner than a search engine?

AI without structured foundations fails in predictable ways. Here are three real-world scenarios:

Scenario 1: The Revenue Confusion

Without standardised metric definitions, an AI assistant might calculate "Monthly Recurring Revenue" differently across conversations---sometimes including one-time fees, sometimes excluding them, sometimes using different time windows. Even worse, without contextual knowledge, it can't explain why MRR matters differently for a subscription-based software company versus a transactional retail business, or how seasonal patterns should influence interpretation.

Scenario 2: The Customer Health Mystery

An AI tool analysing customer health might interpret "active user" inconsistently across time periods. But even with consistent definitions, without knowledge graph context, it misses crucial relationships: that enterprise customers have different usage patterns than smaller businesses, or that seasonal companies show predictable activity cycles that aren't actually indicators of churn risk. Acting on that misinterpretation could lead to unnecessary---and costly---interventions.

Scenario 3: The Performance Paradox

Marketing teams using AI to analyse campaigns might see consistent metric calculations but conflicting strategic recommendations. Without knowledge of current business priorities, competitive landscape, or historical performance patterns, AI can't distinguish between a successful brand awareness campaign (low conversion, high reach---exactly as intended) and a failed lead generation effort with the same surface metrics but a very different goal.

The solution requires both layers working together: the metrics layer ensures accurate, consistent calculations that prevent "moody" AI outputs that change from day to day. The knowledge graph enables context-aware reasoning that connects metrics to business outcomes. The result is that AI becomes less like a calculator and more like an experienced analyst who knows both the numbers and the business.

In practice, this creates a partnership where:

• AI acts as an intelligent assistant, translating questions into both metric queries and contextual analysis

• The metrics layer provides reliable, governed calculations

• The knowledge graph supplies business context, relationships, and strategic alignment

• Users receive insights that are both accurate and actionable

Over time, AI can also help auto-suggest new metric definitions, identify missing relationships in the knowledge graph, and surface unexpected connections between business concepts---making the entire ecosystem smarter as it learns.

Level up data-driven decision making

Make metric analysis easy for everyone.

Get Started Now

5. Market Direction: Building Complete Knowledge Ecosystems

The consensus among data leaders is clear: organisations need more than semantic layers---they need intelligent knowledge ecosystems. Recent industry studies show that organisations with both governed metrics and rich contextual knowledge report 40% faster time-to-insight and 60% fewer data-driven decision conflicts compared to those relying on metrics alone. That's not a minor improvement---it's a fundamentally different way of operating.

Key trends shaping this space include:

• Integrated Architectures: Metrics layers and knowledge graphs working as unified systems rather than separate tools---so users don't have to toggle between a "definitions database" and a "context repository"

• Dynamic Relationships: Knowledge graphs that automatically discover and update business relationships as data and strategies evolve---reducing the manual maintenance burden on data teams

• Semantic APIs: Programmatic access to both metric definitions and contextual relationships, enabling AI tools to tap into business knowledge in real time (an API is essentially a standardised way for software systems to talk to each other)

• Conversational Intelligence: Natural language interfaces that let business users ask questions in plain English---and receive answers grounded in both precise calculations and business context

• Federated Knowledge: Decentralised domain expertise combined with central governance---business units contribute both metrics and institutional knowledge relevant to their area, while a central team maintains consistency and quality

Organisations implementing these complete knowledge ecosystems see measurable improvements: analytics teams report responding to business requests three times faster, while business users achieve 50% greater self-sufficiency in generating insights. More importantly, AI recommendations become significantly more relevant and actionable---not just technically correct, but genuinely useful for decision-making.

This new generation of AI-native infrastructure transforms data teams from metric maintainers---people who answer one-off questions---to knowledge architects who build systems that empower everyone in the organisation.

6. Redefining the User--AI Relationship

To understand why this evolution matters, it helps to contrast how analytics has worked historically versus where it's heading:

Traditional BI (Business Intelligence):

Users request a report. Data teams build it. Users wait---sometimes days, sometimes weeks. By the time the answer arrives, the question may have changed.

Metric-only AI:

Users ask a question. AI provides accurate numbers quickly. But users still need to interpret what those numbers mean in context---and that interpretation requires knowledge that the AI doesn't have.

Knowledge-enhanced AI:

Users ask a question. AI provides both accurate metrics AND business context, relationships, and strategic implications. Users receive actionable insights---not just data, but guidance on what to do with it.

The result:

• Faster insights: No manual modelling required for each new request

• Smarter insights: AI understands both the numbers and the business context behind them

• Empowered users: Self-serve analytics with built-in business intelligence---so anyone can get a meaningful answer, not just technical specialists

• Reduced friction: Data teams spend their time architecting knowledge ecosystems rather than answering one-off questions

This evolution elevates AI from a sophisticated calculator to a knowledgeable business advisor, while positioning data professionals as the architects of organisational intelligence---people who build the systems that make everyone else smarter.

7. Building an AI-Native Knowledge Ecosystem

How can organisations actually build these foundations? Here's a practical roadmap---whether you're starting from scratch or evolving what you already have.

Step 1: Audit your existing foundations

• Identify gaps in both metric definitions and business context---where do teams disagree on definitions? Where does critical knowledge live only in someone's head?

• Catalogue institutional knowledge that isn't yet documented or structured

• Map relationships between metrics, processes, and business outcomes

Step 2: Design your metrics layer foundation

• Break complex, monolithic data models into reusable metric components that teams can mix and match

• Store definitions in a central, versioned repository so changes are tracked and reversible

• Establish clear governance for metric creation and updates---who can define a new KPI? Who approves changes?

Step 3: Build your knowledge graph

• Model relationships between entities (customers, products, markets), processes, and strategic objectives

• Capture business rules, constraints, and decision logic---the unwritten rules that experienced employees know

• Document historical patterns and institutional knowledge in structured, machine-readable formats

Step 4: Create unified APIs

• Expose both metric definitions and contextual relationships via integrated APIs so AI tools can access them together

• Enable AI to access both precise calculations and business intelligence in a single query

• Support natural language querying across both layers

Step 5: Implement AI-native capabilities

• Use natural language processing (NLP)---the technology that allows computers to understand human language---to interpret user questions against your complete knowledge base

• Implement AI-driven suggestions for new metrics and missing relationships in the knowledge graph

• Enable conversational analytics: let users ask questions the way they'd ask a colleague

Step 6: Establish knowledge governance

• Empower domain experts (not just data teams) to contribute both metrics and business knowledge from their area

• Create processes for validating and updating contextual information as the business evolves

• Ensure knowledge graph accuracy through regular audits and feedback loops---treat it like a living document, not a static archive

These steps ensure your foundation provides both the reliability of governed metrics and the intelligence of rich business context. You don't have to do everything at once---starting with even a few well-defined metrics and a handful of documented business relationships can dramatically improve the quality and speed of insights across your organisation.

8. Conclusion and Next Steps

The debate around AI and data foundations often misses the bigger picture. It's not just about maintaining consistent metrics---it's about building intelligent systems that understand both the numbers and the business context behind them. Getting this right is what separates organisations that use AI to generate noise from those that use it to generate genuine, actionable insight.

By combining metrics layers with knowledge graphs in an AI-native architecture, organisations can:

• Achieve consistent, trusted metrics at scale---so everyone is working from the same shared reality

• Provide AI with the business context it needs for intelligent, relevant recommendations

• Empower users with insights that are both accurate and actionable---not just numbers, but guidance

• Transform data teams from metric maintainers to knowledge architects who build systems that make the whole organisation smarter

Your next steps:

• Evaluate your current foundations for both metric gaps and missing business context---where are the disagreements? What knowledge is at risk of walking out the door?

• Pilot an AI integration that leverages both precise definitions and rich relationships, even at a small scale

• Begin capturing institutional knowledge in structured, machine-readable formats---start with what matters most to current decisions

• Collaborate across teams to identify the contextual knowledge that makes your business unique and that no off-the-shelf tool could ever know on its own

Ready to kickstart your AI-native knowledge ecosystem? Discover how Klipfolio PowerMetrics provides both flexible, governed metric foundations and the contextual intelligence needed to transform your analytics practice.

Learn more about PowerMetrics and start your free trial today.