Rubbish In, Rubbish Out: Why Data Quality Still Rules in the Age of AI

At IBM TechXchange 2025, Anna Gutowska presented a live demo of her new IBM tutorial, Build an SQL Agent using watsonx.ai.

It shows data scientists how to build an AI tool that automates the grunt work of writing SQL queries, freeing up time for the parts that actually require human thinking.

It’s clever stuff. But it also made us reflect on one stubborn truth:

👉 Data quality is still the biggest hurdle in analytics.

The Double Pareto Problem

In our experience, analytics projects rarely fail because of the technology.
They struggle because of the data.

It’s often said that 80% of the time goes into the data prep, leaving only 20% for analysis and insight.
But in reality? It’s more like 80% of the 80% delivering 20% of the 20%  what we like to call the Double Pareto Principle of analytics.

It sounds like a fancy Italian coffee, but what it really means is that 96% of the effort delivers only 4% of the value unless your data foundations are solid.

AI Can Help… But It Can’t Save Bad Data

AI agents can speed things up, generating queries, writing syntax, cleansing data, even predicting patterns.
But they’re only as good as the data they can access.

AI may be brilliant at accelerating work.
It’s also brilliant at being confidently wrong.

The truth is: Rubbish in = Rubbish out.

So before we look to AI to “fix” data issues, we have to address the root cause. Generally, disconnected systems, messy metadata, unclear ownership, and legacy processes that haven’t evolved as fast as the tools have.

Human Context Still Matters

Even the smartest models can’t understand business context, governance rules, or why that one spreadsheet named “final_FINAL_v3.xlsx” keeps breaking your dashboard.

That’s where human expertise and collaboration come in, turning automation into augmentation. When teams work with clean, well-structured data, AI stops being a novelty and starts being an accelerator.

Laying the Right Foundations

At Aramar, we see this across data transformation projects every day. Whether we’re implementing IBM watsonx.ai or improving existing analytics systems, the pattern is clear:

  • The best AI outcomes start with trusted data.

  • The fastest automations happen on structured foundations.

  • And the smartest insights come from people who know how to ask the right questions.

So yes, AI agents can transform analytics but only when the groundwork is done right.

Where Next?

If you’re experimenting with AI in your analytics, take a moment to ask: Is my data ready for AI?

If the answer is “not yet,” or “I’m not sure” let’s chat.

We help finance and data teams build strong, automated foundations for analytics. Preparing it so AI can do what it’s best at: accelerating insight.

📩 Email us to discuss your bespoke situation.