Before expanding artificial intelligence investments, companies should first assess their data foundation, according to Fivetran and dbt Labs. A set of eight questions is intended to help organizations take stock of their data architecture.
Summer is traditionally a time for check-ups – on the car before a road trip, or on the bicycle before the next tour. According to Fivetran and dbt Labs, a similar review is worthwhile for corporate data infrastructure amid the current debate over artificial intelligence. Many companies are currently focused on AI applications, yet according to the vendor, projects fail less often because of the models used than because of a data foundation that is inconsistent, outdated, or difficult to use flexibly. Tobias Knieper, Senior Manager EMEA Marketing at Fivetran + dbt Labs, has compiled eight questions intended to help companies conduct an initial assessment of their data architecture.
The first question concerns data consistency: do all teams work with the same, reliable data, or do multiple copies circulate? According to Knieper, differing data versions lead to conflicting analyses and unnecessary administrative overhead. Modern data architectures should therefore aim to provide data once and make it reusable across teams and applications.
The second question addresses the openness of the data platform. Companies should examine whether their data genuinely belongs to them or is controlled by the platform on which it is stored. Open data formats and interoperable architectures – referred to in the industry as Open Data Infrastructure (ODI) – are intended, according to Fivetran, to reduce the risk of vendor lock-in. The company has published a so-called “Data Access Scorecard” that aims to assess how easily data can be used and transferred from common cloud applications.
The third and fourth questions relate to data freshness and the speed of integration. Companies need current and complete data for reliable AI results; delayed availability reduces the value of analyses and automation. The onboarding of new data sources – from CRM, ERP, or SaaS systems – should, in Fivetran’s assessment, happen quickly and with the help of pre-built connectors, so that integration projects do not become a bottleneck for innovation.
The fifth question focuses on the effort required to maintain data pipelines: according to Fivetran, data engineers still spend considerable time on API changes and error-prone pipelines – time that is then unavailable for analytics or AI projects. Standardized, automated integration is intended to reduce this maintenance burden.
The sixth question addresses scalability: generative AI and agent-based applications generate substantially higher data and query loads than traditional BI systems. Open architectures with decoupled compute and storage, according to Fivetran, can help companies better control infrastructure costs as workloads grow.
Seventh, Knieper points to the importance of a semantic layer: AI agents require not only data access but also the context to correctly interpret tables, fields, and business terms. An abstraction layer between raw data and AI models can improve the quality of AI-supported analyses.
The eighth question concerns data traceability. With regard to the GDPR, the EU AI Act, and internal governance requirements, data lineage is increasingly becoming a mandatory practice. Companies should be able to demonstrate at any time where their data originates and how it has been processed.
In Fivetran and dbt Labs’ assessment, the success of AI projects does not depend solely on the capability of the models deployed, but significantly on an open, flexible data infrastructure designed for future requirements. The concept of Open Data Infrastructure, the vendor notes, is therefore gaining traction across industries, as it allows companies to use data independently of individual vendors, integrate new technologies promptly, and safeguard innovation over the long term. Fivetran has published further information on the eight questions, along with the Data Access Scorecard, on its website.
Dr. Jakob Jung is Editor-in-Chief of Security Storage and Channel Germany. He has been working in IT journalism for more than 20 years. His career includes Computer Reseller News, Heise Resale, Informationweek, Techtarget (storage and data center) and ChannelBiz. He also freelances for numerous IT publications, including Computerwoche, Channelpartner, IT-Business, Storage-Insider and ZDnet. His main topics are channel, storage, security, data center, ERP and CRM.
Contact via Mail: jakob.jung@security-storage-und-channel-germany.de