Advances in technology can sometimes seem counterintuitive, such as the shift towards on-premises AI currently happening across healthcare, financial services and other segments that handle large volumes of sensitive data. On-premises data has been overlooked recently due to our preoccupation with the cloud, but the truth is the world’s most critical data still resides on-premises, and the scale of on-premises data is only going to increase with AI.
On-premises data is growing at a phenomenal rate because it offers organisations a tried and trusted method to access, manage and govern data. It also provides a level of security that appeals to organisations that deal with sensitive information. Which is why it has always been the primary option for organisations that operate in high-compliance environments, such as healthcare, financial services and the public sector. This preference isn’t due to inertia or a failure to embrace the shift to the cloud, it’s because these types of organisations want to be able to manage their data in the controlled and predictable manner that on-premises facilitate.
There’s also a lot to be said about keeping AI capabilities on-premises, because it ensures data quality and performance in a more secure environment. It gives organisations the platform to incubate new ideas, develop new AI models, and manage intense AI workloads, all built on the firm foundation of on-premises data. Organisations are also beginning to realise that on-premises data is an accelerant to the speed of change, lowering the bar for the adoption of AI.
Making the link between AI and data analytics
Essentially, organisations want to be able to access their data wherever it resides. Increasingly, as AI takes hold, they will also want to find a way to run AI on-premises, using data and models that are already on-premise. Not only does this meet with data compliance expectations and the needs of their industries, but it also makes sound business sense. The need to adopt, develop and scale AI capabilities isn’t going to go away anytime soon. It will continue to grow at the same rate as the datasets that support the new AI models and applications that businesses develop.
That’s why it’s important businesses can leverage managed data solutions that combine the elasticity of the cloud with the security that on-premises provides. Fortunately, these tools already exist. They have been powering data analytics for many years and can easily be applied to managing AI workloads.
Data analytics and AI have a lot in common; they both rely on data and require a firm data foundation to be successful. Big data can’t function without access to data lakes, data warehouses and databases that span organisations, across different technologies and environments. While AI models can’t learn from data they can’t access or reach. This means that organisations can build on their existing data analytics strategies to build a strong foundation for AI.
Data governance and securing AI
In addition, knowing who truly owns your data will become increasingly important. AI brings new opportunities, but it also brings new rules and new risks. This means that data security and governance are about to become increasingly important for enterprises as they scale their AI capabilities.
The ability to operate AI in a hybrid environment will make true AI data governance possible. In addition, data sovereignty will continue to grow as a requirement for many organizations. To satisfy demand and to allow organizations to operate AI compliantly and internationally, data governance will need to grow at the same pace that AI scales.
For many organisations, this will not only make the difference between compliant and non-compliant AI. It will be the difference between green-lighting projects that can move forward and those that may become stalled, or never even get off the ground, because of compliance and regulatory hurdles. Much of this will depend on the rules around data access. Some of which can be addressed by ensuring that AI datasets and assets are air-gapped to prevent unnecessary access between the cloud and on-premises data solutions.
Moving towards on-premises AI
To coin a famous Mark Twain phrase, ‘the death of on-premises data has been greatly exaggerated’. Far from dying, it’s growing faster than ever, acting as the baseline for critical applications and business processes. However, we have reached an important tipping point where many enterprises might have AI as a roadmap item but haven’t been able to progress because they lack the AI data architecture needed to support their plans.
The truth is that very few organisations use analytics or AI in isolation, therefore it makes sense to adopt a single data foundation that can power both ventures. This isn’t just good data engineering, it’s good business management. It builds a solid foundation for AI today, tomorrow and well into the future.
