By Steven Orpwood

July 2024

 

Rubbish in, rubbish out, or tidy house, tidy mind

 

In a world where AI looms large, and in a push to stay ahead of the game, organisations are scrambling to craft their AI strategies.

These blueprints will determine how they will apply AI, the constraints placed upon it, and how they will assess its impact over time. Whilst this is an essential activity, it only deals with one aspect of the use of AI, that of governance, and not the issue of quality – of AI tools producing incorrect or inaccurate data or information, regardless of whether the AI is being used in a generative or deep search mode. 

 

The phrase “rubbish in, rubbish out” comes to mind. If you train your AI model using poor data (which could be for example out of date, irrelevant or inappropriate for general use), you will get poor results, and unless the AI can learn from these mistakes, the results will continue to be poor. So, if we take a step back, rather than focusing on AI compliance alone and accepting that this will not solve quality problems, we should be spending time resolving organisational data issues: eliminating poor data and duplicates etc and understanding and organising it, so that when used, the AI is generating content from ‘clean’ data, thereby giving it a greater chance of producing high quality output. 

 

At Aim we use our data cleansing tool, dataBelt® to index our data, eliminate duplicates and erroneous entries, and ensure that what we feed into our AI tools can produce good outputs with a good correlation to the question posed.

 

 

If you would like to find out more about our data cleansing and AI tools, and data management processes, please get in touch.