SfSkills Domains Docs GitHub

Home / agentforce / agentforce-sales-ai-setup

agentforce-sales-ai-setup

Setup and configuration of Einstein for Sales AI features: Opportunity Scoring, Pipeline Inspection AI insights, Einstein email insights and composition, and Forecasting AI — prerequisites, license checks, feature sequencing, data readiness validation. NOT for configuring or troubleshooting Einstein Activity Capture and Relationship Insights in a live org — use agentforce/einstein-copilot-for-sales. NOT for building an Agentforce agent — use agentforce/agentforce-agent-creation. NOT for its topic boundary design — use agentforce/agent-topic-design.

Open the skill folder on GitHub

Agentforce Sales AI Setup

This skill activates when a practitioner needs to set up or configure the Einstein for Sales AI feature stack from scratch: Opportunity Scoring, Pipeline Inspection AI insights, Einstein email insights and composition, and Forecasting AI. It focuses on prerequisites, sequencing, license tiers, and data readiness — not feature troubleshooting after setup. It does NOT cover core Agentforce agent creation, subagent design (subagents were called topics before April 2026), Einstein Trust Layer setup, or Einstein Activity Capture configuration — use the dedicated skills for those areas.

The most common source of Einstein for Sales failures is skipping prerequisites: enablement steps have hard dependencies that Salesforce does not always surface clearly in Setup UI, and missing a dependency silently disables downstream features.

Before Starting

Gather this context before working on anything in this domain:

Core Concepts

Opportunity Scoring Prerequisites and Model Training

Einstein Opportunity Scoring trains a machine learning model on your org's historical closed-won and closed-lost opportunities. The model outputs a 0–99 score on open opportunities plus score factors (top positive and negative drivers).

Data gate: at least 200 closed-WON opportunities AND at least 200 closed-LOST opportunities in the last 24 months, each with a lifespan of at least 2 days — two separate 200-record floors, not one combined 200. Salesforce evaluates this gate at training time. If data is insufficient, the model status shows "Insufficient Data" in Setup > Einstein > Opportunity Scoring. The feature toggle appears active, but no scores are generated — this is the most common "Einstein is enabled but scores are not showing" scenario.

Training timeline: Initial model training completes within 24–72 hours for qualifying orgs. The model retrains weekly. Scores appear on Opportunity records only after the first successful training pass.

Sandbox behavior: Training does not occur in sandboxes — this is a documented Salesforce platform constraint, not a configuration error. Never tell a practitioner that sandbox score absence indicates a setup problem; it is expected behavior.

Pipeline Inspection AI Insights Dependency Chain

Pipeline Inspection is a Sales Cloud feature that shows deal movement signals, engagement data, and predictive AI insights for forecast managers. The AI insights layer within Pipeline Inspection has a strict prerequisite chain:

  1. Einstein for Sales license must be active.
  2. Opportunity Scoring model must reach "Active" (trained) status.
  3. Collaborative Forecasting must be enabled — this is a hard dependency that is not enforced or communicated in the Pipeline Inspection enablement UI.

If Collaborative Forecasting is disabled, Pipeline Inspection shows deal activity and engagement data but the AI-driven insights column (predictive deal change signals, commit risk flags) does not appear. Admins frequently conclude Pipeline Inspection AI is broken when the real fix is enabling Collaborative Forecasting.

License Tiers: Einstein for Sales vs. Einstein Generative AI

Einstein for Sales (the add-on or Einstein 1 Sales edition) includes:

Einstein email composition — where a sales rep clicks "Generate Email" and a generative AI model drafts the email body — is a distinct feature requiring the Einstein Generative AI license (also called Einstein GPT in older documentation). This is a separate license tier that must be separately provisioned in Setup > Company Information > Feature Licenses.

Key mistake: Do not assume that enabling "Einstein for Sales" enables generative email composition. Assigning the Einstein for Sales permission set to a rep will not unlock AI email drafting if the org does not have the Einstein Generative AI add-on.

Common Patterns

Sequential Feature Enablement

When to use: Any new Einstein for Sales deployment where multiple features need to be enabled.

How it works: Follow this strict enablement order to prevent missing dependencies:

  1. Verify Einstein for Sales license is provisioned (Setup > Company Information > Feature Licenses).
  2. Enable Opportunity Scoring (Setup > Einstein > Sales > Opportunity Scoring). Wait for model training to complete (24–72 hours).
  3. If email composition is needed: verify Einstein Generative AI license is provisioned, then enable Einstein Email (Setup > Einstein > Sales > Email).
  4. Enable Collaborative Forecasting if not already active (Setup > Forecasts Settings).
  5. Enable Pipeline Inspection (Setup > Pipeline Inspection). AI insights now appear because Opportunity Scoring is Active and Collaborative Forecasting is enabled.
  6. Assign Einstein for Sales permission set to relevant users.

Why not enable all at once: Enabling Pipeline Inspection before Opportunity Scoring is Active results in the AI insights column never populating. Enabling before Collaborative Forecasting is on results in a silently partial Pipeline Inspection with no error message.

Pre-Enablement Data Readiness Check

When to use: Before enabling Opportunity Scoring in any org, especially recently deployed or thinly-loaded orgs.

How it works: Run a SOQL query to check opportunity data volume before enabling the feature:

SELECT COUNT(Id)
FROM Opportunity
WHERE IsClosed = true
AND CloseDate = LAST_N_DAYS:730
GROUP BY IsWon

Group by IsWon — a single total hides the failure mode. You need 200+ in the true bucket and 200+ in the false bucket. If either bucket is below 200, the Opportunity Scoring model will not train. Options: (1) wait until the org accumulates more data, (2) import historical closed opportunities from a legacy system, or (3) document the limitation and set expectations with stakeholders that scores will not appear until the threshold is reached.

Why not rely on Setup UI feedback: The Setup UI enables the feature regardless of data volume. The "Insufficient Data" status only appears after the training pass fails — not before enablement.

Decision Guidance

SituationRecommended ApproachReason
Fewer than 200 closed-won or fewer than 200 closed-lost opportunities in last 2 yearsDo not enable Opportunity Scoring yetModel will not train; scores will never appear; creates false "broken" perception
Enabling Pipeline Inspection AI insightsEnable Collaborative Forecasting firstHard dependency; no AI insights appear without it
Rep needs AI-drafted email compositionVerify Einstein Generative AI license separatelyDistinct from Einstein for Sales add-on; not included by default
Testing Einstein for Sales in sandboxValidate enablement flow only; do not expect scoresOpportunity Scoring model does not train in sandboxes
Org has Einstein for Sales but scores are missing after 72 hoursCheck Setup > Einstein > Opportunity Scoring > Model Status"Insufficient Data" means data gate not met; "Training" means wait; "Active" means check page layout

Step-by-step instructions for an AI agent or practitioner configuring Einstein for Sales from scratch:

  1. Verify licenses — In Setup > Company Information > Feature Licenses, confirm Einstein for Sales (or Einstein 1 Sales edition) is provisioned and seats are available. If email composition is in scope, confirm Einstein Generative AI license is also present. Document which licenses are confirmed before proceeding.
  2. Check Opportunity data readiness — Run SELECT COUNT(Id) FROM Opportunity WHERE IsClosed = true AND CloseDate = LAST_N_DAYS:730 in Developer Console or Workbench. If count is below 200, stop and align with the org owner on a data readiness timeline before enabling Opportunity Scoring.
  3. Enable Opportunity Scoring — Setup > Einstein > Sales > Opportunity Scoring > toggle on. Note the timestamp. Monitor Setup > Einstein > Opportunity Scoring for model status every 24 hours until status shows "Active". Do not enable Pipeline Inspection until this step is complete and model is Active.
  4. Enable Collaborative Forecasting — Setup > Forecasts Settings > Enable Forecasting. Configure at least one forecast type. This is required for Pipeline Inspection AI insights.
  5. Enable Pipeline Inspection — Setup > Pipeline Inspection > Enable. Confirm that AI insights column is visible to forecast managers. If not visible, recheck Opportunity Scoring model status and Collaborative Forecasting state.
  6. Enable Einstein Email (if licensed) — If Einstein Generative AI license is confirmed, go to Setup > Einstein for Sales > Email > Enable. Assign the Einstein for Sales Email permission set to relevant users. Test AI email generation from an Opportunity or Contact record.
  7. Assign permission sets and validate — Assign Einstein for Sales permission set to pilot users. Verify Opportunity Score field appears on Opportunity page layout. Verify Pipeline Inspection is accessible from the Sales Cloud navigation. Collect feedback from 2–3 reps within 72 hours of model going Active.

Review Checklist

Run through these before marking work in this area complete:

Salesforce-Specific Gotchas

Non-obvious platform behaviors that cause real production problems:

  1. Opportunity Scoring silently skips training with insufficient data — If the org has fewer than 200 closed-won or fewer than 200 closed-lost opportunities in the last 24 months (each floor is checked separately), the Opportunity Scoring model enters "Insufficient Data" status. The feature appears enabled in Setup, but no scores ever appear on records. There is no proactive warning — admins only discover the issue after checking model status.
  2. Pipeline Inspection AI insights require Collaborative Forecasting, not just Einstein for Sales — Enabling Pipeline Inspection without Collaborative Forecasting active results in a functional Pipeline Inspection view with deal movement data but no AI-driven insights. There is no error or warning in the UI. This dependency is documented in Salesforce Help but not surfaced in the enablement wizard.
  3. Einstein email composition is a separate license, not included in Einstein for Sales — Enabling Einstein for Sales and assigning its permission set to reps does not unlock AI email drafting. The generative email feature requires the Einstein Generative AI add-on license (distinct SKU). Orgs on Einstein 1 Sales may have it bundled, but standalone Einstein for Sales add-on customers must purchase it separately.

Output Artifacts

ArtifactDescription
License verification checklistConfirms Einstein for Sales and (if applicable) Einstein Generative AI licenses are provisioned before setup begins
Opportunity data readiness query resultCOUNT of closed opportunities in the last 24 months split by IsWon; both buckets must reach 200 to pass the Opportunity Scoring gate
Enablement sequence logOrdered record of which features were enabled, in what sequence, and model status timestamps
Permission set assignment confirmationList of users assigned the Einstein for Sales permission set and (if applicable) Einstein Email permission set

Package files