A global survey of 2,050 senior executives from organizations with more than 500 employees finds that security and resilience (85%), data localization (74%) and data ownership and control (72%) have emerged as the top three data management challenges organizations face in the age of artificial intelligence (AI).
Conducted by EnterpriseDB, a provider of a distribution of an open source Postgres database, the report also suggests that only 13% of respondents work for organizations that are currently achieving those goals, with that subset of organizations already achieving a 5X return on investment (ROI) from their artificial intelligence (AI) projects.
In fact, organizations that have invested in data management and sovereignty are now starting to distance themselves from organizations that are not as committed to gaining control over their data, says Michael Gale, chief marketing officer for EDB.
On the plus side, however, a full 95% of respondents said their organization does plan to take control of their own AI and data platforms by 2028.
In general, more organizations are starting to realize they will need much greater control over their data to successfully operationalize AI, says Gale. The issue is that much of that data is still highly distributed within large enterprise IT environments, he adds. “It’s just massively fragmented,” says Gale.
It’s not clear to what degree the 13% that currently exercise the level of control over their data required to successfully operationalize are benefiting from best practices they may have always followed, because they, for example, operate in a highly regulated industry. The one thing that is certain is that as organizations embrace AI, adopting those best practices is now an absolute requirement. The challenge is that it can take several years to define and implement those best practices, especially in organizations that might have petabytes of data.
At the same time, many of those same organizations are now also trying to adjust to data sovereignty regulations that are forcing them to store data within the bounds of a specific national border, notes Gale.
Ultimately, AI requires organizations to be able to expose the right data at the right time to a large language model (LLM) that may be hosted in the cloud or in an on-premises IT environment. Putting the infrastructure in place to achieve that goal, however, requires a significant amount of investment in IT infrastructure that needs to be factored into any return on investment (ROI) calculus for AI that organizations are making.
In theory, AI might eventually drive more organizations to rationalize the number of databases and tools they currently have deployed, especially as databases such as Postgres add the ability to process vector data in a way that makes it simpler to build and deploy AI applications that expose additional data to an LLM. Otherwise, the usefulness of an LLM is going to be limited to the last time it was retrained.
Hopefully, most enterprise IT organizations appreciate the scope of the data management and engineering challenges that lie ahead in the age of AI. The challenge and the opportunity are finding the will and resolve to address them before rivals establish an insurmountable AI advantage.
