No doubt many of you have seen the growing number of references to ‘Semantics’ and ‘Context’ in articles and videos, LinkedIn of course, and in documents and presentations.
These are commonly mentioned as being an important component in the context of AI applications and for data or metadata management.
Even noted analyst Mike Ferguson said on a recent call with us that “context was the most overused word of 2026”.
I decided that I would add my thoughts to the conversation, especially in the context of how to find and inject useful information, or metadata, from SAP and other ERP applications into a semantic or context platform.
Before I do that, I am going to try to clearly distinguish between the meaning of semantics and context.
For me, the semantics are how terms are defined, agreed, and how they are related to specific data and each other. So, for example, ‘Customer’ or ‘Net Revenue’ or even ‘Quarter’ should be able to be understood even if they appear in different systems across an enterprise. Once set these are fixed so that they provide consistency across projects and use cases. They also provide a translation, or mapping facility between raw data and users or systems that consume data.
This is not actually a new concept. Some of the early analytics and business intelligence tools provided these. Examples include the Business Objects Universe, Microstrategy’s Semantic Graph amongst others.
On the other hand, context is about situational relevance. It answers the questions related to how those definitions and the related metadata or data are used for example, in business processes, analytics and integrations.
And now of course, both semantics and context are both critical for the effective and secure implementation of AI projects. This is because no matter how smart AI is, if it is not working with the right data in the right context then anything it does risks being wrong, with whatever consequences that brings.
One crucial component of semantics and by association, context is metadata. Naturally, this metadata comes from many sources. Metadata is a foundational component of any system. Not all metadata is created equal and as you might expect there are no accepted standards for metadata. So, for example the metadata which underpins an in house application built on an Oracle RDBMS is very different to that contained in an ERP application from say SAP, Microsoft, Oracle or Salesforce.
Metadata for ERP applications can be very large, complex, highly customized and difficult to work with. For example, the S/4HANA system we use for demonstrations has over 140,000 tables and 1.5 million attributes as well as other components such as CDS Views etc. Accessing this metadata in the System Catalog provides little of any use for data focused projects. There are no logical names for tables, attributes and other components and no relationships defined.
This means that trying to work with it for semantics or context is virtually impossible. The valuable metadata which contains logical names and relationships is stored elsewhere. As a result, data projects are often late and business confidence in data can be compromised. This also flows through into the challenges for AI if clarity about data and meaning is not available.
Customers use Safyr to help improve confidence in data and accelerate delivery of data projects. It also helps them to ensure that AI is working with the right elements in an ERP application.
Safyr provides easy access to the logical, as well as technical, metadata in ERP applications. This means that data analysts can explore the data model easily without needing specialist application knowledge. They can curate metadata which is relevant for a specific use case into subsets which can be shared with data governance, data catalog, metadata management, data modeling products and in other formats.
AI is smart, however data integrity is critical and if it has to work within an enterprise where complex applications are part of its process flow, then it is vital that that it is aware of their metadata in the context of its work.
Safyr gives AI a reliable foundation to work from when SAP or ERP metadata is part of the picture.