What We Typically Find
When we start working with a new client, we almost always find one of two things.
The first is a business still running largely on Excel. Data is pulled from various systems, dropped into spreadsheets, and passed around. Different teams have different versions. Nobody is quite sure which one is right. The weekly reporting cycle eats hours that should be spent on something more valuable.
The second and increasingly common scenario is a business that has already invested in Power BI. They’ve made a genuine effort to move away from spreadsheets. But what’s emerged over time is a collection of individual reports and dashboards, each built for a specific purpose, each sitting on top of its own small data model. Sales built their dashboard one way. Finance built theirs another. HR did their own thing.
On the surface it looks like progress. Underneath, the same fragmentation problem remains. It’s just harder to see.
You Already Have the Foundation
AI is only as good as the data it is working from. Ad-hoc tools struggle because they are starting from raw, ungoverned data: spreadsheets that may or may not be current, definitions that vary between teams, numbers without context. The output looks confident, but the inputs cannot be trusted.
If your organisation has already invested in BI, especially in a properly built Power BI Semantic Model, you have exactly what AI needs to be useful. You have clean, structured, governed data. Core metrics like revenue, utilisation, or wins and losses are defined consistently. Business entities like customer, product, and location are classified the same way across every report. The hard work has already been done.
This is the foundation that turns AI from a parlour trick into a genuine business capability. And it is what makes a data agent the natural next step.e. It does not happen when the BI genuinely makes their job easier.
Every dashboard is solving the same problem from scratch, and solving it slightly differently each time.
The root cause in both cases is the same: there is no shared, governed layer that defines what the data actually means across the business. No agreed definition of what a “customer” is. No single calculation for “revenue.” No common understanding of how the numbers should be structured before they reach a report.
The Problems This Creates
Without a proper semantic data model underpinning your data, you end up with a set of problems most senior leaders will recognise immediately:
- You can’t scale. Adding a new data source, a new market, or a new business function means unpicking existing models and rebuilding them. There is no structure that absorbs growth naturally.
- Numbers that don’t match. Finance says one thing. Sales says another. Both are technically pulling from the same source, but different calculations, filters, or definitions produce different results. Trust in the data erodes, and decisions get deferred while people investigate discrepancies.
- Reports that can’t be brought together. When each dashboard sits on its own data model, pulling information from different sources into one place becomes unreliable. Figures don’t align, time periods don’t match, and questions that cross two reports rarely have a clean answer.
- Every new report is a new project. With no reusable foundation, every new reporting need requires starting from scratch. Reconnecting to data sources – a problem often solved by implementing data integration services – agreeing on definitions, rebuilding calculations that already exist somewhere else. Development time is wasted and inconsistency grows with every report added.
- Good data, poor adoption. When people don’t trust the numbers, they stop using the dashboards and go back to spreadsheets. BI investment goes underutilised. Not because the tools are wrong, but because the foundations were not built to earn trust.
What Good Looks Like
The answer is a semantic layer. This is a single, governed definition of your core business concepts, metrics, and relationships that sits between your raw data and your reports. It defines what things mean, how measures are calculated, and how different parts of the business relate to each other. Every report and dashboard draws from it, so consistency becomes automatic rather than something that has to be policed manually.
The good news is that you do not need to build a full data warehouse to get started. A well-built Power BI Semantic Model is an excellent first step. It gives you that shared, governed layer on top of your existing data sources and delivers an immediate improvement in consistency and trust across your reporting. For many organisations, this alone is transformative.
As your data environment grows more complex, the natural next step is to introduce a structured data warehouse underneath. This brings your source data together, cleans it, and organises it into layers: raw ingestion at the bottom, cleaned and harmonised data in the middle, and a gold layer of business-ready, trusted datasets at the top. Your semantic model then draws from that gold layer, inheriting all of that structure and quality. The principles are the same whether you start with just the semantic model or build the full stack from day one.
In practical terms, this means:
- Key business metrics, whether that is revenue, wins and losses, utilisation, or headcount, are calculated the same way everywhere, with agreed rules about what is included and excluded.
- Core business entities like customer, product, and location are classified and structured consistently, so every report is working from the same underlying definitions.
- When a business measure changes, it is updated once in one place. Every report and dashboard that uses it reflects the change automatically.
- Adding a new data source or a new report means extending the model, not rebuilding it.
The result is a business where people trust what they’re looking at, where the answer to a question is the same regardless of who pulls the report, and where data becomes a genuine foundation for decision-making rather than a source of debate.
Making It Happen
Building a semantic model well starts with the business, not the technology. Before anything is configured, the right questions need to be asked. Take revenue as an example. Does it include tax? Is it recognised at the point of invoice or the point of payment? Does it include work that has started but where contracts have not yet been signed? These are questions most organisations assume everyone has already agreed on. They usually have not.
These conversations are often the first time an organisation has explicitly agreed on its own definitions. They surface assumptions that have been baked into different spreadsheets and reports for years. They’re almost always more valuable than anyone expects. Not just as a technical input, but as an alignment exercise for the business itself.
From there, the build is straightforward. Start with the semantic model, defining what the data means and how it’s calculated. Reports and dashboards are built on top of that model, not on top of raw data or each other. If and when a data warehouse makes sense, it slots in underneath, with the gold layer providing the clean, business-ready foundation the semantic model draws from.
Done properly, the initial build doesn’t take as long as people assume. A focused semantic model covering the core metrics and entities that matter most to the business can be in place within weeks. From there it grows with the business, absorbing new sources and new questions without losing the consistency established at the start.
For organisations that have already invested in Power BI, this doesn’t mean throwing away what’s been built. It means introducing the shared structure underneath it that makes everything more trustworthy and scalable, and significantly reducing the time and cost of every report built on top of it going forward.
Frequently Asked Questions
A semantic model is a single, governed layer that defines your core business concepts, metrics, and relationships. It sits between your raw data and your reports, ensuring that data means the same thing across your business.
For organisations that’ve already invested in Power BI, implementing a semantic model doesn’t mean throwing away what’s been built. It simply means introducing a shared structure underneath those reports to make everything cohesive and scalable.
A prime example is a Power BI Semantic Model, which maps data sources into clear business terms.
