How to Not Handle Data: AI, Archaeology and Other Mixed Metaphors
Agentic AI. Multi-modal AI. Explainable AI. Conversational AI. Generative AI. The hype is everywhere.
Globally, 78% of organisations reported using AI in 2024, a substantive increase from 2023’s 55%.[i] Estimates put private investment levels in the hundreds of billions of dollars,[ii] and even governments are joining in.
AI job postings are up, AI skills penetration continues apace, and news stories abound with ever more eye-popping statistics. But in the headlong rush for adoption and processing power, it’s worth remembering those pesky data fundamentals: GIGO, aka garbage in, garbage out.
Less sexy? Sure. Critical? Absolutely.
The Case for (Good) Data in AI
Bad data costs businesses. End of. Indeed, one study estimated the losses could be as much as 15-25% of a business’s revenue,[iii] something the ongoing introduction of AI will only amplify, if not done right.
Afterall, AI is the shiny new layer over the top of older data foundations, in that it leverages “algorithms, data, and computational power to recognize patterns, make decisions, and learn from experiences.”[iv] Therefore, if the underlying business data is flawed, so too will the AI outputs be.
Understanding and mitigating this below-the-waterline danger is essential for businesses seeking reliable and trustworthy AI integration in the long term. Experience would suggest very few organisations have a clean house when to comes to their data. And once the digging starts, it’s often found to be in a much poorer state than first thought. Cue a veritable archaeological excavation of issues accrued over time, all requiring careful cleansing and rebuild, along with strict controls to prevent future deterioration. And so, in the short term, the optimism of AI’s promises vs delivery, can sometimes be a little akin to jamming a soap dispenser stalk into a potato and expecting perfect results to immediately flow forth.
By contrast, acknowledging the need to tidy up and transform those datasets into something robust that AI can properly piggyback off? Game changing.
And all it needs is a little human intelligence in the mix.
The Role of the Business Analyst in AI Preparedness and Success
A BA can play a crucial role, by collaborating with stakeholders to:
1. Identify Data Requirements and Gaps
By understanding processes and types of data that underpin AI workflows, BAs can help ascertain areas holding stale or incomplete data, as well as developing remediation plans to target those gaps.
2. Standardise Data Formats
Any Lean Six Sigma BA will tell you that reduction of variables and a degree of standardisation is key to smooth running processes. Therefore, by helping a business define appropriate structures, formats, labels and syntax for the data in play, a BA ensures that different systems, departments or users can share data without compatibility issues. This is essential for holistic insights, especially where AI straddles multiple datasets and those datasets may have many different owners and inputters.
3. Data Cleanse
After the identification of issues, BAs can collaborate with a combination of IT and business stakeholders to remove duplicates, correct inaccuracies, and manage transforms, plus fill in any missing values.
4. Implement Data Controls
A BA can also help establish or improve data governance frameworks, by detailing data ownership going forwards, as well as the processes for maintaining the data to ensure continued quality into the future. In conclusion, the layering of AI into business processes has the potential to exacerbate the already high cost of poor data to organisations, especially as users become more abstracted from the source material. But a business analyst provides a means to help get things back on track, and ensure AI models are utilising clean, reliable, and relevant data, leading to better outcomes, decisions, and credibility. So, your business data can be more soap dispenser, less potato.
Marbral Advisory helps businesses harness success through better data management. To find out more, contact our team today: hello@marbraladvisory.com
References:
[i] https://hai.stanford.edu/assets/files/hai_ai_index_report_2025.pdf
[ii] https://hai.stanford.edu/ai-index/2025-ai-index-report
[iii] https://sloanreview.mit.edu/article/seizing-opportunity-in-data-quality/
[iv] https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-top-trends-in-tech
