Precisely, The world leader in data integrity, announced today the results of new world research. The study of the observability of the AI, created by Business Application Research Center (Barc) and sponsored by precisely, studied a diversified set of more than 250 data and stakeholders of AI in the world, by discovering critical information on the way in which organizations exploit observability to stimulate the results of confidence in AI and analysis.
As organizations evolve their use of AI, observability has become a fundamental requirement to ensure transparency, responsibility and performance between data ecosystems. Research reveals that if many companies have taken important measures to formalize observability programs, progress vary from one discipline to another. The differences in the maturity of the program, measurement practices and regional adoption trends indicate all areas that still require attention.
Read also: Interview with Aithority with Nicole Janssen, co-founder and co-PDG of Altaml
The observability of the AI is gaining ground, but the gaps remain
76% of organizations have formalized, implemented or optimized programs for data quality and observability of the data pipeline, demonstrating a strong commitment to create foundations of Trust in AI. While the observability of the AI / ML model is close to 70%, the responses indicate a wider range of maturity levels, many organizations always operating with incoherent or underdeveloped programs.
With regard to the measurement of success, 68% of respondents use qualitative and / or quantitative measures to assess their observability efforts, however, the remaining organizations are based on ad hoc measures or no measure, which poses a significant risk. Without clearly defined parameters and alignment with governance executives on a business scale, organizations may not be below their AI objectives.
Unstructured data emerge as a key orientation
Organizations extend their observability programs beyond structured tables to include semi-structured data (such as JSON or newspaper files) and unstructured data (such as text, images, video and sound). According to the study, 62% of organizations explore the use of semi-structured data, with 28% already actively, while 60% currently assess unstructured documents. These strong adoption trends indicate increasing recognition of the importance of various types of data – in particular as advanced use cases such as predictive automatic learning and generative AI depend on it. The observation of these data requires observability techniques different from that of the tables, including the careful tension and monitoring of object metadata.
Read also: The growing role of AI in attacks based on identity in 2024
North America leads to AI and the maturity of observability
Compared to EuropeNorth American companies report significantly higher AI adoption rates and the maturity of observability. On average, 88% of North American organizations have formalized observability programs in all disciplines, against only 47% Europe. North American companies also focus on regulatory compliance and data confidentiality, despite the absence of federal AI regulations comparable to the EU AI. In addition, North Americans grant a higher priority than Europeans on the precision of the model and twice as much, many of which have formal observability measures in place.
“Like AI and the emergence of agent use cases increase the risks and analysis awards, data teams consolidate their observability programs to strengthen governance and quality of data,” said Cameron OgdenMain vice -president – Product management instead precisely. “Research strengthens that observability is not just a good to have but a fundamental capacity to ensure the integrity of corporate data – in particular when it comes to fueling AI models for reliable and scalable results.”
Methodology
The observability survey for AI innovation was carried out by Barc and sponsored by precisely and Collibra. The study collected the responses of 264 data and stakeholders of AI in all industries, in particular IT, manufacturing, financial services and retail. Respondents included data engineers, data scientists, level C executives and business process owners, offering a complete vision of the adoption of observability, challenges and best practices.
(To share your ideas with us, please write to pingen@itechseries.com)