Observability for AI innovation study​​

​​

Latest research from global data integrity leader Precisely and the Business Application Research Centre (BARC) shows that observability programme maturity is uneven across data quality, data pipelines, and AI/ML models, as unstructured data adoption in organizations continues to grow.

According to the Observability for AI Innovation report whilst 76% of organizations have formalized,  implemented, or optimized programmes for both data quality and data pipeline observability demonstrating a strong commitment to building trusted AI foundations, when it comes to AI/ML model observability the responses indicate a broader range of maturity levels, with many organizations still operating with inconsistent or underdeveloped programmes.

When it comes to measuring success, 68% of respondents use qualitative and/or quantitative metrics to assess their observability efforts, however, the remaining organizations rely on ad-hoc or no measurement at all, posing a significant risk.  Without clearly defined metrics and alignment with enterprise-wide governance frameworks, organizations risk falling short of their AI objectives.
 
Unstructured data is emerging as a key focus with 62% of respondents saying that their organization is exploring the use of semi-structured data, with 28% saying they are already actively using it, while 60% of organizations are currently evaluating unstructured documents.  These strong adoption trends signal growing recognition of the importance of diverse data types – particularly as advanced use cases like predictive machine learning and Generative AI depend on them.  Observing this data requires different observability techniques than tables, including the careful appending and tracking of object metadata.
 
“As AI and the emergence of agentic use cases raise the risks and rewards of analytics, data teams are solidifying their observability programmes to strengthen data governance and quality,” comments Cameron Ogden, Senior Vice President – Product Management, Precisely.  “The research reinforces that observability is not simply a nice-to-have but is a foundational capability for ensuring the integrity of enterprise data – particularly when it comes to fuelling AI models for trustworthy and scalable outcomes.”

Compared to Europe, North American firms report significantly higher AI adoption rates and observability maturity with an average of 88% of North American organizations having formalized observability programmes, compared to just 47% in Europe.