Einstein Platform
Einstein Platform is the umbrella Setup area in Salesforce that hosts the configuration surfaces for the org's Einstein AI capabilities: Prediction Builder, Einstein Discovery, Einstein Bots, Einstein for Service, Einstein for Sales, Einstein Vision, Einstein Language, and the newer Agentforce family.
Definition
Einstein Platform is the umbrella Setup area in Salesforce that hosts the configuration surfaces for the org's Einstein AI capabilities: Prediction Builder, Einstein Discovery, Einstein Bots, Einstein for Service, Einstein for Sales, Einstein Vision, Einstein Language, and the newer Agentforce family. It is less a product than an organizing label; admins navigate to Setup, Einstein Platform to find per-feature setup pages, monitor usage and quota, and review the Einstein Activity log that captures every model call across features.
The Einstein Platform area predates the Atlas Reasoning Engine and the modern Agentforce capabilities. Salesforce has been adding new AI capabilities under the Einstein brand for years; the Platform label is the consolidation point that anchors them in Setup. The newer Einstein Setup guided experience overlaps significantly with the Platform area; most teams use Einstein Setup for initial enablement and the Platform area for ongoing configuration of specific features.
Why Einstein Platform is a label, not a product
What the Setup area actually contains
Setup, Einstein Platform expands to a tree of per-feature configuration pages: Einstein Discovery, Prediction Builder, Einstein Bots, Einstein for Service (Article Recommendations, Case Classification, Case Wrap-Up, Reply Recommendations), Einstein for Sales (Lead Scoring, Opportunity Scoring, Forecasting, Activity Capture), Einstein Search, Einstein Vision, Einstein Language, Einstein Trust Layer, and Agentforce surfaces (Agent Builder, Agent Actions, Data Libraries, Testing Center). The tree is long. Newer features land here per release. Admins coming to the area for one feature often discover three others they did not know were available.
The relationship to Einstein Setup
Einstein Setup is a newer guided-experience surface that overlaps with Einstein Platform. Where Platform is a Setup-tree organizing label, Setup is a curated activation wizard that suggests which features to enable based on the org's data and usage. The two surfaces share the underlying configuration; changing a setting in Einstein Setup updates the same record as changing it in the Platform area. Use Einstein Setup for first-time enablement and feature discovery; use the Platform area for ongoing per-feature tuning where the wizard would be too abstract.
Usage, quota, and the Einstein Activity log
Einstein Platform surfaces a usage report per feature, showing model calls per day, latency, and quota consumption. The Einstein Activity log captures every individual model call across features with the prompt, response, masking applied, and audit metadata. The log is the compliance backbone for using LLM and predictive models on regulated data. Most orgs do not look at the log until something goes wrong; building a habit of weekly log sampling catches issues (drift, masking errors, unexpected feature usage) before they escalate.
Trust Layer integration
The Einstein Trust Layer is configured from inside Einstein Platform. The Trust Layer enforces PII masking, data residency, no-training-on-customer-data, and audit logging across every feature in the Platform area. Trust Layer settings are org-wide, not per-feature, which is intentional; an org's compliance posture should not differ across Einstein features. Verify Trust Layer settings during initial enablement and re-verify after any compliance review or regional expansion.
Licensing and what is included by edition
Einstein features are licensed differently by edition. Predictive Einstein features (Lead Scoring, Opportunity Scoring, Case Classification) are bundled with Enterprise+ editions of the relevant cloud at no extra cost in most cases. Generative AI features (Einstein GPT, Prompt Builder, Agentforce) are licensed separately through Agentforce SKUs and Einstein-for-X SKUs. The Einstein Platform area shows which features the org has access to based on licenses; features the org has not licensed appear grayed out with an upgrade prompt. Reading the area first before assuming features are available is a cheap habit.
Permissions, governance, and the admin pattern
Access to Einstein Platform features is gated by per-feature permission sets. The Einstein Admin permission set grants broad configuration access; per-feature permissions (Einstein NBA, Einstein Case Classification, etc.) grant more granular access. A common governance pattern: one or two admins hold the Einstein Admin permission set and manage configuration; broader team members hold the per-feature permissions they need to use the features. Avoid handing Einstein Admin to everyone who asks; the Platform area exposes settings that change behavior across the org, and unaudited changes there cause downstream surprises.
Where the Platform area is going
Salesforce continues to add features to Einstein Platform per release. The trajectory is more Agentforce, more generative AI, more LLM-grounded features alongside the predictive features that built the brand. The Platform label as an organizing principle works whether the underlying tech is a small classifier (Lead Scoring) or a large language model (Agentforce). Teams should expect the area to expand and should plan their configuration discipline to scale with it rather than treating each new feature as a one-off setup task.
How to navigate Einstein Platform without missing features
Most teams underuse Einstein Platform because they enter through one specific feature page and never explore the rest of the tree. Spending an hour walking the full tree once per release catches features that ship without marketing fanfare and would otherwise sit unused.
- Open Setup, Einstein Platform, and walk the tree
Click into every node. Note which features are enabled, disabled, or unavailable for your edition. The walk takes 30 to 45 minutes and produces a written inventory of what is in scope.
- Confirm Trust Layer settings
Open the Einstein Trust Layer node. Verify PII masking matches the data types your org handles, data residency matches your region requirements, and audit logging is enabled. These are org-wide settings; getting them right matters once, not per feature.
- Audit existing per-feature configuration
For each enabled feature, open its configuration page. Confirm settings are current, fields are still relevant, and the feature owner is still on the team. Features whose owner has left often run on stale config.
- Sample the Einstein Activity log
Pull a week of the Activity log. Confirm prompts and responses look sensible, masking applied where expected, no unexpected feature firing high-volume calls. Build the weekly sampling habit before it becomes urgent.
- Pair with Einstein Setup for new feature discovery
Einstein Setup suggests features to enable based on org data. Run through the wizard quarterly to surface candidates you might miss in the per-feature Platform tree.
- Document the Einstein feature inventory
Write the inventory down. Owner, enabled date, last review date per feature. The document is the registry that prevents features accumulating without anyone responsible for them.
- Schedule a quarterly platform review
Quarterly, walk the tree again, sample the log, update the inventory, retire features no one uses. The cadence is the discipline that keeps the Platform area from becoming a graveyard.
Which Einstein features are toggled on for the org. Drives licensing and Trust Layer scope.
Org-wide PII masking, residency, and audit logging settings. Apply to every Einstein feature.
Granular access control per feature (Einstein Case Classification, NBA, Agentforce Builder).
How much detail (prompt content, masking applied) the log captures per call. Verbose for development, summary for production.
Team-maintained registry of enabled features, owners, and review dates. The discipline that prevents accumulation.
- Features ship per release without big marketing pushes. Walking the Platform tree quarterly is the only reliable way to know what is available without missing capabilities.
- Trust Layer settings are org-wide. Changing them affects every Einstein feature. Coordinate before tweaking.
- Per-feature permission sets are easy to misconfigure. Run an audit of who has Einstein Admin and remove broad access from anyone who only needs one feature.
- The Einstein Activity log fills quickly on high-volume features. Set retention policy explicitly rather than letting it become a multi-year compliance footprint.
- Features whose owner has left the team continue to run on stale config. The inventory document is the discipline that catches them.
Trust & references
Cross-checked against the following references.
- Salesforce Einstein overviewSalesforce
- Salesforce AI strategySalesforce
Straight from the source - Salesforce's reference material on Einstein Platform.
- Einstein Platform OverviewSalesforce Help
- Einstein Trust LayerSalesforce Help
- Einstein SetupSalesforce Help
Hands-on resources to go deeper on Einstein Platform.
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What does Einstein Platform need to work effectively?
Q2. What is a key benefit of Einstein Platform for business users?
Q3. How does the Einstein Trust Layer relate to Einstein Platform?
Discussion
Loading discussion…