Artificial Intelligence is changing how organisations interact with business information.
Instead of searching through reports, navigating dashboards or manually comparing different sources, users can increasingly ask a question and receive an answer in seconds.
But there is an important limitation: AI can only work with the information available to it.
If that information is incomplete, inconsistent, outdated or poorly governed, the resulting answer may be unreliable regardless of how sophisticated the technology is.
This is why the growth of AI does not make traditional data management less important. In many ways, it makes it more important than ever.
This article is the third in our Beyond the Dashboard series, exploring how business intelligence, reporting and Artificial Intelligence are changing the way organisations interact with their data.
Better Technology Does Not Fix Bad Data
It is easy to focus on the capabilities of AI itself. Can it answer questions in natural language? Can it summarise a report? Can it identify trends? Can it explain why a KPI changed?
Those capabilities are increasingly impressive, but the quality of the answer still depends on the quality of the foundation information.
In a scenario, you could ask an AI assistant: “What was our most profitable product last quarter?” You would think this a straightforward question. Before an organisation can trust the answer, the system needs to understand several things including:
- Which definition of profit should it use?
- Which data source is authoritative?
- Are all business units calculating the measure consistently?
- Is the latest quarter fully loaded into the system?
- Does the user asking the question have permission to see all of the underlying information?
If those foundations are unclear, AI does not remove the problem. It can simply make the problem easier to access.
One Question, Several Answers
Many organisations already experience this challenge without AI.
A finance team may calculate a KPI one way, while an operational team uses a slightly different definition. Two reports may appear to show the same measure but draw information from different systems or use different calculation logic.
The result is familiar: two people enter a meeting with two different versions of the truth.
Introducing AI into that environment does not automatically resolve the disagreement. If anything, it raises a more difficult question: “Which version should the AI trust?”
For AI-assisted analytics to be genuinely useful, organisations need confidence in the definitions, calculations and data sources that sit behind the answer. Otherwise, the interface may become easier while the underlying inconsistency remains.
Governance Is About More Than Restricting Access
Data governance can sometimes sound like a technical or compliance exercise, focused primarily on permissions and control.
Those things are important, but good governance goes much further. It helps establish which information should be trusted, how important metrics are defined and who is responsible for maintaining them. It can also provide clarity around where information came from, when it was updated and how it has been transformed before reaching the user.
In a traditional reporting environment, inconsistencies may become obvious because users are comparing reports or questioning different figures.
With AI, the interaction can feel much more direct. A user asks a question and receives a response. That makes the governance behind the answer even more important, because the user may not immediately see all of the calculations and assumptions involved in producing it.
Trusted Metrics Become Even More Important
Consider a commonly used measure such as revenue, margin, customer churn or headcount.
Each sounds relatively simple until different parts of the organisation define it differently. Does revenue include intercompany transactions? Is margin measured before or after certain costs? When does a customer officially count as having churned? Does headcount include contractors?
A dashboard can sometimes expose these differences because users know which report or model they are looking at. A conversational interface may remove some of that visible context.
That means organisations need confidence that the AI is working from approved business definitions rather than selecting whichever version of a measure happens to be easiest to find. The more natural the interaction becomes, the more important those foundations are.
Security Still Applies When the Interface Changes
There is another important aspect of governance: access.
AI should not become a shortcut around existing security controls. A user asking a question in natural language should only be able to access information they would otherwise be permitted to see. The fact that someone can phrase a question differently should not allow them to retrieve confidential financial, HR or customer information.
This becomes particularly important as analytics environments connect with a wider range of systems and AI tools.
The user experience may change, but the underlying security model still needs to apply. That means organisations need to think carefully about permissions, data access and how AI interacts with governed business systems.
AI Needs Context, Not Just Data
Having accurate data is important, but accuracy alone may not be enough. AI also needs context.
A number may be technically correct but difficult to interpret without understanding what it represents. A decline in revenue may look concerning until it is compared with seasonality. An unusually high cost may make sense when linked to a planned investment. A KPI may appear to have improved because its calculation changed halfway through the year.
This is where well-managed business intelligence environments become particularly valuable. They do not simply store numbers. They provide definitions, relationships, calculations and business context that help users understand what those numbers mean.
When AI works with that governed information, it has a much stronger foundation for producing useful answers.
The Risk of a Convincing Answer
One of the challenges with AI is that an answer can sound authoritative even when the underlying information is weak.
That does not necessarily mean the technology itself has failed. If an AI system is given outdated information, inconsistent definitions or incomplete data, it may produce a perfectly logical response based on inaccurate foundations.
This creates a different type of risk. Users may be less likely to challenge an answer because it is presented with confidence and clarity.
That is why organisations should not judge AI purely on how impressive the interface appears. They also need to understand what information sits behind it and whether that information can be trusted.
AI does not remove the need to question the answer. It increases the importance of knowing where that answer came from.
Start with the Foundation
There is understandable pressure on organisations to experiment with AI and identify practical use cases.
But introducing AI into reporting and analytics should not begin with the question: “Which AI tool should we deploy?”
A more useful starting point may be: “Is our information ready for AI?”
That means understanding where key data sits, which sources are authoritative, whether important metrics are consistently defined and whether appropriate security and governance are already in place.
Organisations do not need perfect data before they can begin experimenting with AI. But they do need to understand the limitations of the information they are asking AI to work with.
Without that understanding, there is a danger of accelerating the production of answers without improving the quality of the decisions behind them.
From Trusted Data to Trusted Answers
The future of business intelligence is likely to involve a combination of reports, dashboards, natural-language interfaces and increasingly intelligent analytics.
But the foundation remains familiar: trusted data, consistent definitions, appropriate security and clear business context.
These may not be the most exciting parts of an AI strategy, but they are what allow the more exciting capabilities to be used with confidence.
Because ultimately, the quality of an AI-generated answer depends on more than the model producing it. It depends on the information behind it.
AI is only as reliable as the data it has been given to work with.
Catch Up on the Beyond the Dashboard Series
If you missed the earlier articles in the series, you can catch up here:
Want to Explore What This Looks Like in Practice?
If this topic is relevant to your organisation, we’ll be exploring it further in our upcoming webinar, Beyond the Dashboard: The Future of Cognos Analytics, Reporting and AI.
Join HAYNE Solutions and IBM for a practical look at how reporting, trusted data and AI-assisted analytics are evolving in IBM Cognos Analytics, including a live demonstration of the latest capabilities.
Next in the Beyond the Dashboard series:
From Reports to AI Agents: How Business Intelligence Is Changing
In the next article, we’ll explore how business intelligence is evolving from traditional reports and dashboards towards more conversational and AI-assisted ways of working.
