Business intelligence
« Back to Glossary Index
Business intelligence is the set of processes, tools and practices that turn raw operational data into reporting and analysis people use to run the business. It covers gathering data from source systems, modeling it into a consistent structure, and presenting it through dashboards, reports and self-service exploration.
BI is sometimes described as legacy in a market focused on AI. That is a misreading: the definitions, models and governance that BI builds are exactly what AI systems depend on to produce trustworthy answers.
What Is Business Intelligence?
Business intelligence answers what happened and, with reasonable tooling, why. It gives the organization a shared factual basis: how many customers, what revenue, which products, which regions, compared with when.
That shared basis is the real deliverable, and it is easy to undervalue until it is absent. An organization where finance, sales and operations each produce a different revenue figure spends its meetings reconciling rather than deciding. BI’s core function is to make the number agreed, so the conversation can move to what to do about it.
In scope, BI covers data integration from source systems, modeling into analysis-friendly structures, a definition layer holding business measures, and the delivery layer of dashboards and reports. The delivery layer gets the attention and the modeling layer does the work.
BI vs. Analytics vs. Data Science
The terms overlap and are used loosely, but the distinction is practical.
Business intelligence reports on what happened using known, modeled data. The questions are largely anticipated, and the output is structured for repeated consumption.
Analytics is broader and more exploratory — investigating why something happened, testing hypotheses, examining relationships that were not designed into the model.
Data science builds models that predict or classify, producing estimates about things not directly observed rather than summarizing what is.
They are layered rather than competing. Data science built on data whose BI layer is inconsistent inherits every inconsistency — which is why organizations that skip BI foundations in favor of AI frequently find their models cannot be trusted for reasons that have nothing to do with the models.
How a BI Stack Fits Together
Source systems — ERP, CRM, operational databases, external feeds — hold the data as it is created.
Integration and pipelines move that data into an analytical environment, applying transformation and quality rules along the way.
Storage and modeling — a warehouse or lakehouse where data is structured for analysis rather than for transaction processing. Star schemas remain common because they are easy to reason about and query efficiently.
Semantic layer defines business measures once — what active customer means, how margin is calculated — so every report inherits the same definition rather than reimplementing it.
Delivery through dashboards, reports, self-service tools and increasingly embedded or conversational interfaces.
The semantic layer is where consistency is won or lost. Organizations that skip it end up with the same measure defined slightly differently in dozens of reports, and no reliable way to find out which is correct. Our business analytics work usually starts there.
Types of BI Content
Operational reports support day-to-day work — order status, exception lists, daily volumes. Usually detailed and consumed at row level.
Management dashboards track performance against target for a function or region, designed to be scanned quickly for what needs attention.
Executive reporting summarizes organizational performance, with few measures and heavy emphasis on trend and variance.
Self-service exploration lets analysts and managers answer their own questions from governed datasets.
Regulatory and statutory reporting carries defined formats and audit requirements, and demands far tighter control than internal reporting — the area where enterprise reporting discipline matters most.
Applying uniform standards across all five is a common and expensive mistake. Executive reporting needs certification and change control; exploratory analysis does not, and imposing it there suppresses the exploration that makes BI valuable.
What Makes BI Trusted or Ignored
Most BI estates fail on trust and usage rather than on technology. The recurring causes are consistent.
Disagreeing numbers. Two reports showing different values for the same measure. Once this happens in a leadership meeting, every subsequent figure is questioned.
Uncontrolled proliferation. Content is created and never retired, so users cannot tell which of several similar dashboards is current or authoritative.
Stale data with no freshness indicator. Users have no way to know whether they are looking at this morning’s figures or last week’s.
Built to a specification rather than a question. Reports that satisfy a written requirement but do not answer what the requester actually wanted.
The corrective measures are unglamorous: define measures centrally, certify and label authoritative content, show data freshness on every dashboard, assign ownership, and retire what nobody opens.
How BI Is Changing
Three shifts are reshaping BI without displacing its purpose.
Cloud platforms have removed the capacity constraints that shaped older design, so pre-aggregation exists for cost and speed reasons rather than because detail could not be stored.
The semantic layer has become independent of any single BI tool, letting definitions be shared across tools and increasingly consumed by AI systems rather than re-implemented in each.
Natural-language interfaces allow questions to be asked in plain language. This shifts the constraint rather than removing it — an AI answering questions about the business is only as reliable as the definitions it queries. Far from making BI obsolete, it makes the modeling and governance layer more consequential than it was when only analysts saw it.
Frequently Asked Questions
Business intelligence is the set of processes, tools and practices that turn raw operational data into reporting and analysis the business uses — covering integration from source systems, modeling into consistent structures, and delivery through dashboards and reports.
BI reports on what happened using modeled data and largely anticipated questions. Analytics is broader and more exploratory, investigating why something happened and testing relationships that were not designed into the model.
BI summarizes what is observed. Data science builds models that predict or classify things not directly observed. They are layered — models built on an inconsistent BI foundation inherit that inconsistency regardless of how good the modeling is.
Source systems, integration pipelines, a warehouse or lakehouse where data is modeled for analysis, a semantic layer defining business measures once, and a delivery layer of dashboards, reports and self-service or conversational interfaces.
A semantic layer defines business measures centrally — what active customer means, how margin is calculated — so every report inherits the same definition. Without it, the same measure ends up defined slightly differently across dozens of reports.
Usually on trust rather than technology: reports that disagree on the same measure, content that proliferates without retirement, stale data with no freshness indicator, and reports built to a written specification rather than to the question the requester actually had.
No. Natural-language and AI interfaces change how questions are asked, but an AI answering business questions is only as reliable as the definitions it queries. That makes the BI modeling and governance layer more consequential, not less.
Operational reports for day-to-day work, management dashboards tracking performance against target, executive summaries focused on trend and variance, self-service exploration on governed datasets, and regulatory reporting with defined formats and audit requirements.
