Data intelligence: the key to successful AI

Jim Liddle at Nasuni discusses why smart data is the key to smart AI and explains how data intelligence enables successful and sustainable AI implementations

 

It’s fair to say that AI was one of the hottest topics of 2023 thanks to the explosive growth of Generative AI tools. Following this surge, enterprises are under increasing pressure from their boards to successfully implement AI to gain an advantage over their competitors.

 

However, while the pace of AI innovation is moving quickly, when it comes to implementing AI in a business, it should be approached as a marathon, not a sprint. 

 

Why? In the mad AI panic, deeper data infrastructure issues are surfacing within organisations that have been slowly mounting in recent years, and it’s causing AI projects to fail.

 

For example, S&P Global’s 2023 Global Trends in AI report, based on 1500 enterprises and research bodies worldwide, found that data management was the most commonly cited technological barrier to AI/ML deployments, ahead of security and compute performance. 

 

Instead of forging ahead with doomed AI implementations in 2024, it’s time for businesses to take stock and instead focus attention towards data intelligence and ‘AI readiness’.

 

In essence, organisations must realise the data quality, governance, access, and storage requirements of AI. Only then can they consider pursuing digital transformation or improvement programmes that will successfully provide their desired competitive edge.

 

Data is the essence of AI

Shockingly, a huge number of businesses are dithering over their data, storing significant volumes simply because they don’t understand how to use it, or if they actually need it. And this same data is being used to train AI models.

 

When organisations don’t know if their data is accurate and up-to-date, available on-demand or archived, properly classified and ‘searchable’, compliant, or if it contains sensitive information like personal identifiable information (PII) or protected health information (PHI), how can we expect to train a smart and capable AI model with it, while being compliant?

 

Visibility into organisational data is a crucial and often-overlooked first step towards successfully utilising AI, and the key to this is to focus efforts on data intelligence, with the support of a hybrid cloud platform.

 

Why AI will rely on data intelligence

Data intelligence is all about intelligence and insight into the data that a business has and an understanding of how to leverage it. IDC defined six fundamental questions that data intelligence helps organisations to answer:

• Who is using what data?
• Where is data, and where did it come from (lineage and provenance)?
• When is data being accessed, and when was it last updated?
• Why do we have data? Why do we need to keep (or discard) data?
• How is data being used, or perhaps more specifically – how should data be used?
• Relationships – what relationships are inherent within data and with data consumers?

Identifying the answers to these questions is critical for any business to enable them to make better decisions and to provide the avenue for successful technology implementations, especially with AI - data intelligence and AI go hand in hand.

 

Why? We can only control AI and realise its full potential in driving productivity, efficiencies and cost reductions, through the information we feed it; as such we want to avoid irrelevant, old, duplicate or sensitive information. Not only will this skew the potential of AI, but it could lead to sharing critical business information with the wrong people.

 

AI doesn’t know if it’s the CEO or a wider team member asking it a question, it only knows the data it’s been given.

 

Only by curating the data in a business and becoming intelligent about what data can deliver insights, what is required for compliance, and what data you can cut, will organisations become truly ready to expose their data to an AI system and reap the benefits from doing so.

 

Smarter data will make AI more eco-friendly

In addition to the challenge of implementing AI successfully, the energy and computing requirements for advancing AI systems pose environmental sustainability implications. The resources needed to run large AI models are significantly increasing power consumption, and therefore the carbon footprint of companies and their data centres alike. 

 

Data intelligence can help to balance this out. As businesses refine their data sets through consolidation and cleansing processes, they’ll reduce the old and duplicate data that would be fed into AI models and gain a better understanding of proprietary data and how it can be utilised by AI.  

 

Refined, rich data sets will be able to feed smaller, smarter AI models that require less power. These better-trained models will outperform the larger models that are based on raw compute power, too.

 

Succeed or fail: Getting data in order is key

AI cannot succeed in a vacuum. It must be utilised as a part of the wider data intelligence umbrella. While data projects have already been put into motion for sales, customer support, and similar low-hanging initiatives by many businesses through the implementation of data analytics, machine learning and AI, a common sticking point comes when trying to integrate this same technology into more sophisticated, high-value applications.

 

By utilising resources to achieve better data intelligence this year, businesses will be able to finally realise the power of AI for these higher-value initiatives. Ultimately, IT leaders will gain a single source of the truth from previously disparate data assets, to boost their companies’ innovation capabilities and responsiveness to customers.

 

And taking a sustainable approach to implementing these technologies, both in terms of speed and considering environmental impact, will enable companies to deliver the competitive advantage that boardrooms are looking to AI for.

 

---

 

Jim Liddle is Chief Innovation Officer at hybrid cloud company Nasuni

 

Main image courtesy of iStockPhoto.com