AI创新的可观察性：采用趋势、需求和最佳实践

Observability for AI Innovation Adoption Trends, Requirements and Best Practices BARC research study Research sponsored by: Authors Kevin Petrie | VP of Researchat BARC US Timm Grosser | Senior AnalystData & Analytics at BARC Kevin Petrie is the VP of Research at BARC,where he leads the data management practi-ce and writes about topics such as AI, data in-tegration and data governance. For 30 yearsKevin has deciphered what technology meansto practitioners, as an industry analyst, in-structor, marketer, services leader, and techjournalist. Timm Grosser is a Senior Analyst Data &Analytics at BARC with a focus on data strate-gy, data governance and data management.His core expertise is the definition and im-plementation of data & analytics strategy,organization, architecture and software se-lection. Timm is a popular speaker at conferences and seminars and has authorednumerous BARC studies and articles. Kevin launched a data analytics services team for EMC Pivotal in the Ameri-cas and EMEA, and ran field training at the data integration software providerAttunity (now part of Qlik). A frequent public speaker and co-author of twobooks about data management, Kevin most loves teaching data and AI leadersabout evolving strategies, tools and techniques to capitalize on the value ofdata. Contact Contact Mail: tgrosser@barc.com Mail: kpetrie@barc.com www.barc.com www.barc.com Social Media:linkedin.com/timm-grosser Social Media:linkedin.com/kevinpetrietech Executive Summary����������������������������������������������������������� 5 Survey Results�������������������������������������������������������������������� 8State of Observability...................................................................9How to Tackle Challenges in Data Observability....................12Growing Importance of Observing Unstructured Data.........14The Role of Observability Tools................................................17Teamwork matters.....................................................................20The Next Wave – Does GenAI Change the Game?.................22 Methodology������������������������������������������������������������������� 27 Company Profile������������������������������������������������������������� 29About BARC................................................................................30 Sponsor Profiles�������������������������������������������������������������� 31About Collibra............................................................................32About Precisely..........................................................................33 Foreword We conducted this survey with a wide range of respondents, fromdata engineers and data scientists to CxOs and business processowners. Our population includes stakeholders from a variety ofbusiness functions and industries, most notably IT and manufac-turing respectively. Company sizes are evenly distributed: nearly1/3 have fewer than 500 employees and nearly 1/3 have 5,000or more. We also represent North America and Europe in equalportions, with a small fraction of respondents coming from Asia. As artificial intelligence (AI) raises the risks and rewards of ana-lytics, organizations recognize the imperative for transparent,trustworthy inputs and outputs. So there is no better time for thisreport, which surveys 264 data and AI stakeholders across indus-tries about why, where and how they implement observability. We examine three distinct observability disciplines: data quality,data pipeline and AI/ML model. In each case observability refers tothe measurement, monitoring and optimization of these elements.We find that most organizations now have formalized programsfor data, pipeline and model observability. Organizations priori-tize privacy, auditability, and compliance in their effort to fosterResponsible AI. Challenges do persist, of course, thanks to short-ages of skills, collaboration and process automation that hinderfull adoption of observability. As always, you can count on BARC to identify and explain mean-ingful differences between these various subgroups. Enjoy! Another headline is the extension of observability beyond basic ta-bles. Organizations now gather metadata to observe text, images,videos and other semi- or unstructured data objects in supportof generative AI (GenAI). As often happens, regional differencespersist: North American firms lead in AI adoption and observabi-lity maturity, with a stronger emphasis on regulatory compliance,model accuracy and consistent program metrics. Timm Grosser and Kevin PetrieMarch 2025 Executive Summary Executive Summary As AI raises the risks and rewards of analytics, data teams are solidifyingtheir observability programs to strengthen data governance. Engineers,managers and executives now contribute to formal programs for data, pipe-line and model observability. While much work remains, cross-functionalteams seek to improve privacy, trust, transparency, regulatory c