Snowflake and Salesforce launch open standard for standardised AI data
Inconsistent definitions of business data slow down AI systems across platforms. The Open Semantic Interchange initiative aims to solve the problem.
(Image: aboutstock/Shutterstock.com)
An alliance of numerous technology companies—led by Snowflake, Salesforce, and dbt Labs—has launched a new initiative to create a uniform standard for business data. According to the companies involved, Open Semantic Interchange (OSI) is intended to solve a central problem for AI applications: the inconsistent definition of business data across different systems.
(Image:Â iX)
The data2day programme on 4 and 5 November 2025 offers numerous practical insights into the topics of data architectures and pipelines, data quality and governance as well as agentic AI and data privacy.
According to the announcement, in addition to the three main initiators, Alation, Atlan, BlackRock, Blue Yonder, Cube, Elementum AI, Hex, Honeydew, Mistral AI, Omni, RelationalAI, Select Star, Sigma, and ThoughtSpot are also involved in the project. The declared aim is to create standardized semantics that put an end to the contradictory definitions of identical business terms in different systems.
The problem of the fragmented data landscape
The problems that arise in practice when an AI model has to retrieve and process data with contradictory definitions become particularly clear with terms such as “advertising expenditure,” “active customer,” or “gross profit margin.” Such terms are often defined differently in different IT systems. Retailers typically classify an “active customer” in their sales systems as someone who has made a purchase within the last 90 days, while marketing uses the same term for anyone who has interacted with content in the last month.
The OSI initiative now aims to create a vendor-neutral standard that allows different organizations to share data across their platforms with a common set of definitions. The aim is to ensure interoperability between different data analysis, AI and business intelligence applications and services.
Videos by heise
The initiative has not yet published the specification itself, but the OSI model is intended to define the representation of semantic metadata in the standardized YAML format. The semantic definition file should contain various attributes, including references to base tables, connection keys, and relationships. It could also include a range of metrics, synonyms, and customized instructions for AI. The working group also plans to set up an open-source repository containing both the specification itself and converters that can be contributed by various partners.
OSI initiative with vendor-neutral administration
The initiative is to be continued in the future under a vendor-neutral administration. According to the initiators, given the different market positions and strategic interests of the participating companies, this is the only way to ensure the long-term success of OSI and attract further partners for support.
More details and further information on the Open Semantic Interchange initiative can be found in the official announcement on the Snowflake blog.
(map)