How ‘I’ is AI without data?

By Chris Lees

Last year was a big one for generative AI and Machine Learning, with massive leaps forward in the field. It’s pretty clear that good data management and maturity are vital if we are to capitalise on ever-developing AI capabilities in 2024 and beyond.

In December, Solomon Klappholz’s article ‘A lack of data maturity could hamper enterprise AI ambitions in 2024’ featured on the ITPro site. Klappholz posits that since particularly generative AI requires mature data capabilities, organisations are being held back by their shortcomings in this area. He cites a recent Carruthers and Jackson study which reported that in 2023 the number of organisations that had no data strategy at all had fallen by just 2% (from 29% to 27%) since 2022. It’s surprising and somewhat alarming that this number is so menial. Perhaps even more worryingly, that same study showed no change in the (rather large) number of organisations – 56% – describing their data policies as either clunky or non-existent.

Surely everyone should be focusing on data strategies to ensure they develop in line with the technologies becoming available? It’s plain to see that most businesses haven’t cracked this problem yet.

In his Forbes article ‘2023: The Year Generative AI Transformed Enterprise Data Management’, published in December, Robert Kramer takes a more technology-centred view, looking at the evolution of various technologies that might support AI. He explores how these are increasingly being integrated into enterprise data management platforms and solutions, and how they have changed to become better-suited for AI.

In the context of building out capacities, Kramer introduces the interesting concept of a ‘data lake house’ – a hybrid of a data lake and a data warehouse – to try to meet some of the demands of AI in terms of requiring large amounts of data with a mixture of structured and unstructured data.

Kramer’s conclusion aligns with a quote from a recent speech by Salesforce cofounder and CTO Parker Harris: “In a way, this AI revolution is actually a data revolution, because the AI revolution wouldn’t exist without the power of all that data.”

Fundamentally, no data means no AI.

Kramer summarises that since data is playing an increasingly key role in business, there is also a growing necessity for effective data management strategies going forward.

This is certainly consistent with what our Data Clan team have seen in recent client engagements and aligns directly with our Data Management Maturity Assessment. Ultimately, AI – whether generative or machine learning – is all based upon the analysis of existing information. If that data is poor quality, AI can’t do anything to improve that. It can find information, summarize it, and even infer in some cases from the content. But its output will only ever reflect the data it’s given. In other words, the AI is only as good as the data you feed it.

AI could be used to help identify outliers, but the reasons these anomalies occur are polarized, as are the appropriate responses to them. Incorrect data should be removed or corrected, whereas the result of something interesting or unexpected should be a focal point for investigation. Trusting an AI or machine algorithm to correctly make that decision for us is risky given the current state of the art.

Kramer also mentions data governance, but he conflates governance and security. Although he explains how security aspects are being assisted by AI (in threat detection, for example), he doesn’t really address how AI can help with data governance. That belies the fact, today at least, that data governance, as with many elements of data management, still rests with people. Machines are not governing their own data, people are still doing that; so we need keep sight of what people need to do, what processes need to be in place, what kind of culture is needed, what education and awareness there needs to be around data, because AI will not solve those problems for us.

We have before us an enormous opportunity to start addressing some of the key data management competencies that organisations need to lay the groundwork for the utilisation of AI. These preparations will also enable them to operate much more efficiently: to appropriately apply lean principles, to drive out waste, to create the space and capacity to think, and to make better decisions. All of these are possible without AI, but are built on the same foundation of good data management.

There is no better time than now to invest in data management.