TL;DR — Key Takeaways
-
AI Maturity Isn’t Translating Into ROI: While 68% of organizations consider their AI adoption mature or maturing, only 28% can quantify a return on investment.
-
Data Readiness Remains a Major Obstacle: Just 4% of organizations have more than 70% of their unstructured data ready for analysis in a lakehouse.
-
Data Management Challenges Are Slowing AI Adoption: Organizations struggle with finding relevant data, cleaning outdated information and establishing governance and compliance.
A survey of 500 technical decision makers finds that while more than two-thirds (68%) feel their organization is either maturing (48%) or has matured (20%) in terms of adoption of artificial intelligence (AI), only a little more than a quarter (28%) are able to quantify a return on investment (ROI).
Conducted by Komprise, a provider of a data management platform, the survey also identifies customer service as the highest area for quantified ROI at 38%, followed by employee productivity and decision speed (28%), revenue growth (26%) and operational savings (21%).
While how ROI is measured will naturally vary from one organization to the next, in many instances AI is rapidly becoming table stakes to remain competitive. As such, it’s often difficult to quantify an ROI simply because, in the absence of AI, it’s increasingly likely an organization will find it is losing market share to rivals that are more efficient.
The one thing that all organizations share is they are struggling to adopt AI because they are unable to expose the right data to the right AI model at the right time. For example, the survey finds only 4% have more than 70% of their unstructured data ready, with schema, for analysis in a lakehouse and only 24% have more than half ready and available.
Nearly half (49%) struggle to find the right project data and clean redundant, obsolete, and trivial (ROT) data. Other challenges include metadata extraction/enrichment to create unstructured-data schema (42%), governance and compliance (41%), visibility into what data exists and where it is located (37%) and presenting or copying dispersed file data to AI platforms (31%), the survey finds.
Additionally, only 40% have or plan to deploy an enterprise-wide file-data connector and classification program. Only 7% of respondents said their organization has 70% or more of their data cleaned, prepared and classified for AI, while the majority (57%) have 50% or less ready. Only 29% go beyond classification to remove stale data and identify the right data for each AI use case.
The survey makes it apparent that many organizations in the AI era are now finally addressing long-standing data management issues and challenges, says Komprise COO Krishna Subramanian. Among those challenges is finding a way to apply AI to where data resides because so little of it resides in a centralized data lake, she adds. In fact, 46% of respondents said leaving structured and unstructured data in place but still being able to query it from analytics, lakehouse, and AI tools is very important for dashboards and/or AI pipelines.
Organizations are also starting to realize that in the absence of an ability to manage data, the cost of the tokens consumed rises, adds Subramanian. “Data equals tokenomics,” she says.
No two organizations are, of course, on the exact same AI journey, but some ability to centralize the management of the data exposed to AI models and agents is clearly going to be crucial. The survey finds that the most widely used data lakehouses are Microsoft Fabric (49%), Google BigQuery (32%), Amazon Redshift (28%), Databricks (27%) and Snowflake (22%). Regardless of approach, however, the less relevant and accurate the data exposed to AI models is, the harder it becomes to justify an ROI.
