Data dictionaries have long been one of those “we should do this someday” items on every Salesforce project checklist, but the reality is that most organizations never get around to actually defining, entering, or maintaining them. Business analysts often champion the cause. Admins – especially those inheriting a mature org – understand the pain of missing documentation all too well. Yet, despite their importance, the vast majority of Salesforce orgs still lack robust data dictionaries that are current, accessible, and meaningful.
To put it in perspective: trust in data foundations – which includes metadata, context, and governance – remains a core blocker to delivering reliable insights in the modern AI era. A Salesforce State of Data & Analytics survey found that only 43% of data and analytics leaders have formal data governance frameworks in place – an environment in which comprehensive metadata documentation like data dictionaries would normally live.
This statistic underscores why we need to rethink data dictionaries, especially as Einstein Copilot evolved into Agentforce and Salesforce’s Data 360 portfolio will reshape how we work with data across the enterprise. The goal of this article is simple: to broaden your perspective and spark a discussion about how Salesforce data dictionaries need to evolve in an agentic world.
Before We Talk About What’s New, Let’s Be Real…
When people complain about “gaps in the data dictionary,” it’s worth stepping back and acknowledging some common realities:
- Salesforce standard objects and fields often have built-in documentation. You can find help text and field descriptions in the platform or in Salesforce documentation.
- Managed packages may document metadata internally or externally, but most teams don’t pull that context into the org itself.
- Data properties like Business Owner or Sensitivity Classification are important, do exist somewhere, often in spreadsheets, confluence pages, or external catalogs – they may just not surface natively in the org metadata.
- Large enterprises frequently invest in standalone data catalog tools (e.g., Collibra, Informatica, Alation). Those catalogs might be robust, but they live outside Salesforce.
In other words, not seeing field descriptions in Salesforce doesn’t mean they don’t exist. The problem is often about access and usability, not the absence of knowledge. Perhaps the most frustrating version of this truth is that even if your org metadata were perfectly described, much of that context is not exposed in the UI when you need it.
If you want to see a field’s business definition while building a formula, report, or validation rule, you have to leave the screen you’re on. That interruptive workflow is a big part of why data dictionary documentation rarely gets traction – until now.
Why You Should Re-Think Data Dictionaries in 2026
There were three major developments in 2025 that should fundamentally change how Salesforce practitioners think about metadata documentation.
1. Salesforce Admin and Setup Agents
Salesforce’s new admin and setup agents, showcased at Dreamforce ’25, turn metadata into actionable knowledge and knowledge into metadata, helping admins and builders create reports, formulas, automations, and queries with full awareness of the underlying model and documentation. These agents don’t just react to your prompts; they reason over metadata and context.
This means:
- Well-described objects and fields make these agents more powerful and trustworthy.
- If your data dictionary lives in an external catalog or document, you can integrate it as a trusted source so the agentic experience references it as context, closing the gap between documentation and action.
This means you have a choice: you can bulk update metadata properties, adhering to the current data dictionary fields, or you can configure trusted external sources (e.g., enterprise data catalogs) so your agents treat them as authoritative reference material. Either way, static spreadsheets are no longer the path forward.
Note: If you have not seen “The Future of Setup Powered by Agentforce” session at DF25, you can watch it on Salesforce+.
2. Tableau Next and the Rise of the Semantic Layer
The world of data and analytics is shifting fast.
Today, when Salesforce Data 360 ingests CRM data, it doesn’t bring all the metadata context with it, and Data Model Objects (DMOs) themselves lack deep data dictionary properties.
In the meantime, Data 360 governance features do not expect you to have assessed and tagged all fields for sensitivity; it delivers a scalable tagging and classification – a sensible choice given the volume of enterprise data.
Meanwhile, the Data 360 semantic data model (SDM), which Tableau Next relies upon, can be seen as another canonical representation, with business-facing definitions, metrics, and relationships. Though I’d like to see tighter integration with native metadata over time, I am glad the semantic layer has a richer metadata model to be used by agentic and analytic workflows.
Data dictionaries now matter at multiple layers – not just in Salesforce metadata, but across the analytics stack. If your definitions and classifications live only in a spreadsheet or internal wiki, they won’t help when you need them during a BI query or dashboard build.
3. Informatica and MuleSoft Becoming Part of the D360 Portfolio
This one is still emerging, but the implications are significant. With Informatica (a metadata and governance heavyweight) becoming part of Salesforce’s Data 360 portfolio alongside MuleSoft, data catalogs and metadata frameworks are becoming first-class citizens in the Salesforce ecosystem.
You can expect easier access to enterprise metadata inside Salesforce’s AI and UI layers.
Final Thoughts
In essence, here’s what you should take away:
- Data dictionaries still matter, perhaps more than ever. The move to agentic workflows and cross-platform analytics emphasizes context, not just data access.
- The specific format doesn’t matter as much as having clear definitions, ownership, and sensitivity classifications tied to your metadata.
- Assess the quality of your current data dictionary (many AppExchange solutions can help you do this quickly).
- If your solutions – such as Agentforce or Tableau – rely on accurate metadata, use your assessment to shore up gaps. A few PDFs and prompt-based LLM workflows can significantly accelerate this remediation.
Did you find this interesting, or do you disagree? Be sure to drop a comment below or connect with me on LinkedIn – I’d love to hear your thoughts. Let’s learn from one another and get better together.
Comments: