Iris Context: The Context Layer for AI

By:

Jul 15, 2025

Share on social

Bridging the gap between business users and meaningful, data-driven insights has long been both the promise (and the shortcoming) of traditional BI tools. With the rise of AI, we now have a real shot at delivering on that vision. But one question keeps coming up: Does AI need a semantic layer? And if so, what should that layer look like?

If you believe, as many incumbents do, that the primary value of AI is to provide a conversational interface for existing BI tools, then the answer seems obvious: AI needs the semantic layer those tools already provide. In this view, large language models simply sit atop the existing BI stack, replacing dashboards and spreadsheets with a chat interface.

At Iris, we believe this view drastically underestimates the true power and potential of AI. It’s like building television shows as if they were just radio programs with pictures: a missed opportunity to rethink what’s possible. Limiting AI to the constraints of traditional BI tooling means overlooking its ability to transform how we reason with data altogether.

Because the real opportunity of AI isn’t just to answer the same questions today’s dashboards do — only in a chat window. It’s to unlock human-level reasoning and accuracy in the process of generating insights. That requires a fundamentally different kind of semantic layer — one purpose-built for AI to reason, not just to report.

BI Semantic Layer vs AI Context Layer

Imagine a business user wants to analyze repeat buying behavior but explore it across different definitions of a “repeat buyer” (e.g. 2+ purchases in 30 days, 3+ purchases lifetime, etc).

With today’s BI tools, this kind of flexibility usually requires a data modeler to edit the semantic layer, write custom SQL, or build new dashboards — a slow and expensive loop.

AI changes that. But to unlock its potential, we need a new kind of foundation: a context layer designed for reasoning, not reporting.

Iris Context: Built for AI

Iris Context is a new kind of semantic layer. One purpose-built for AI to reason about the business while providing the guardrails necessary for accuracy, consistency, and control.

‍

BI Semantic Layer	AI Context Layer
Metrics and dimensions Built around fixed definitions of measures and dimensions. Good for reporting, but not for nuanced questions.	Real-world concepts Captures higher-level business logic — like “repeat buyer” or “churn risk” — that AI can reason over and redefine dynamically.
Rigid Adding new logic often means rebuilding or refactoring parts of the model.	Flexible and adaptive Handles new questions and variations by composing from reusable concepts, without breaking consistency.
Built for dashboards and reporting Optimized for aggregations, filters, and charting — not open-ended exploration.	Built for reasoning Designed to support multi-step, contextual reasoning — the kind humans do when answering real business questions.
Manual Every metric, dimension, and logic change needs a human to define and maintain it.	Automated and self-learning Improves over time by learning from interactions, feedback, and new business context.

‍

Unlike traditional semantic models, which are rigid and manual, Iris Context is automated, expressive, and dynamic.

Iris Context is automated, expressive, and dynamic

It has three core components:

1. Data Context Layer

This layer automatically learns the structure and logic of your data schema and enforces it deterministically. It ensures that joins happen correctly, prevents common issues like fan traps and chasm traps, and guarantees that AI-generated queries always align with the rules of your data model.

2. Business Configuration Layer

Every business has foundational rules that must always be applied - no exceptions. For example, a retail company might require strict adherence to the retail calendar. Another company may need to enforce row-level access based on user roles. Iris allows you to configure these business-specific rules explicitly — and enforces them consistently, deterministically, and with 100% guarantee, regardless of the question being asked.

3. Business Logic Layer

This layer captures the nuanced, often messy rules that vary by business context — the ones traditional BI tools struggle with.
Examples:

“Unless otherwise specified, a repeat purchaser buys again within 1 week.”
“In San Francisco, a 'hired worker' means an offer is signed. In Dallas, it means the first full week has been completed.”

These rules would normally require complex modeling and manual maintenance. With Iris Context and our patent-pending logic framework, these rules can be configured easily and applied consistently across all AI-generated answers.

Results: Accuracy and Reasoning at Scale

Iris Context delivers on the two hardest challenges in AI-driven analytics: accuracy you can trust and reasoning that mirrors a top analyst.

Observe.AI needed insights with zero tolerance for error. Iris Context was key to delivering that high bar on accuracy:

We wanted insights with uncompromising accuracy standards. Iris stood out — it beat everything else we tried, including Snowflake Cortex.
— Vache Moroyan, Chief Product Officer, Observe.AI

At Veronica Beard, executives ask complex, high-context questions and get answers in seconds. Iris understands concepts like “repeat purchaser” and adapts logic dynamically based on business context.

With Iris, answers that took a day or two now come back in under a minute. It understands our business in a way generic tools can’t.
— Max Lagresele, Sr. Director of Data, Veronica Beard