Sales-Con-201 · Domain 5 of 5

Sales-Con-201 Domain 5: Predictive and Generative AI

Predictive and Generative AI covers which AI features Agentforce Sales offers, what each needs to work, how they help sellers, and how Salesforce's Trusted AI principles apply to sales use cases.

Exam weight estimate
13%
Official objectives estimate
3
Estimated questions estimate
~8
Suggested study time estimate
7–9 hrs

About 11 min read Reviewed October 2026

Overview

Predictive and Generative AI is the smallest of the five sections at 13% and has three objectives. It replaced older objectives that mentioned Sales Cloud Einstein in passing and now treats AI as a topic in its own right, reflecting the move to the Agentforce Sales name.

The generative objective covers tools that create content or act for sellers: sales emails drafted from prompt templates, record and call summaries, Agentforce employee agents in Lightning and Slack, and autonomous agents such as Agentforce Sales Development for lead nurturing and Agentforce Sales Coach for role-play practice. The predictive objective covers scores and insights built from historical data, including Einstein Lead Scoring, Opportunity Scoring, deal insights and forecasting predictions. The third objective asks you to spot ethical risks, such as bias in scoring or customer data exposure, and apply Salesforce's Trusted AI principles and the Einstein Trust Layer.

How it's tested

Scenarios describe a sales team that wants to save time, prioritize better or reach more prospects, sometimes with a concern about privacy or fairness. You decide whether a predictive or generative feature fits, what it needs to work and which principle or Trust Layer control answers the concern. Watch for answers that use AI where a simple rule would do.

  • Predict or generate

    A seller need is described, and you decide whether it calls for a score or insight, or for drafted content or an agent.

  • Data readiness

    A model gives poor or generic results, and you identify missing history, fields or setup as the cause.

  • Responsible use

    A fairness, privacy or accuracy worry is raised, and you apply a Trusted AI principle or Trust Layer feature.

Reasoning it through

A sales manager wants reps to spend mornings on the leads most likely to convert. Drafting more emails with generative AI would not decide which leads come first. Einstein Lead Scoring analyzes past conversions and ranks open leads with the top factors behind each score, so reps can sort list views by score. If the org lacks enough conversion history, Einstein falls back to a global model until local data builds up.

Objectives 3

Official objectives from the Salesforce outline, in outline order.

5.1 Understand what generative AI tools are available in Agentforce for Sales.Content-generating and agentic features for sellers.
What it means
Generative tools produce new text or take actions for sellers using large language models grounded in CRM data. In Agentforce Sales these include drafted sales emails, summaries of records and calls, conversational employee agents in Lightning and Slack, and autonomous agents for prospecting and coaching.
What to understand
Sales emails use Prompt Builder sales email templates grounded in lead, contact and opportunity data, and the rep reviews before sending. Record summaries condense activity and status for an account or opportunity. Call summaries and insights come from Einstein Conversation Insights recordings. Agentforce Sales Development engages and nurtures inbound leads autonomously and hands qualified leads to reps, while Agentforce Sales Coach lets reps practice pitches in role-play and receive feedback. Each requires the right licenses, permission sets and data access.
How it shows up
A seller wants help writing, summarizing, prospecting or practicing, and you choose the generative tool that fits.
Common confusion
Expecting generative features to rank or score records, which is the job of predictive models.
5.2 Understand what predictive AI tools are available in Sales Cloud.Scores, insights and predictions built from historical data.
What it means
Predictive tools analyze patterns in past records to estimate future outcomes. Sales examples include lead scores, opportunity scores, deal and account insights, activity-based relationship signals and forecast predictions.
What to understand
Einstein Lead Scoring ranks leads by likelihood to convert and shows the factors behind the score. Einstein Opportunity Scoring rates the chance an open deal closes and highlights positive and negative factors. Both need enough historical data; otherwise Einstein uses a global model built from aggregated, anonymized data. Pipeline Inspection surfaces deal insights such as slipping close dates, and Einstein Forecasting predicts period outcomes. Data quality strongly affects predictions, so incomplete fields and stale records weaken them.
How it shows up
A prioritization or forecasting need is described, and you pick the predictive feature and its data prerequisite.
Common confusion
Assuming a score is a business rule that can be edited directly, rather than a model output driven by data.
5.3 Identify ethical challenges of AI and apply Salesforce's Trusted AI Principles to sales scenarios.Responsible use of AI with sales data and customers.
What it means
AI in sales raises risks such as biased scoring, inaccurate generated content, exposure of personal data and over-reliance on automated decisions. Salesforce answers these with its Trusted AI principles, its guidelines for trusted generative AI and the Einstein Trust Layer.
What to understand
The Trusted AI principles are Responsible, Accountable, Transparent, Empowering and Inclusive. The generative AI guidelines stress accuracy, safety, honesty, empowerment and sustainability. In practice, exclude sensitive attributes such as age or gender from models when they could produce unfair outcomes, keep a person reviewing customer-facing content, explain why a score was given, and disclose when customers interact with an AI agent. The Trust Layer adds secure data retrieval, data masking, zero data retention with model providers, toxicity detection and an audit trail.
How it shows up
A fairness, transparency or privacy concern appears in a sales AI rollout, and you choose the principle or control that addresses it.
Common confusion
Treating the Einstein Trust Layer as a fix for biased training data, which needs data and model review.

High-yield concepts

The terms, rules and formulas this domain turns into questions most often.

  • Einstein Lead Scoring

    Predicts lead conversion likelihood with top factors.

    Remember
    Uses a global model when local history is thin.
    Tested as
    Lead prioritization scenarios.
    Trap
    Expecting it to write outreach.

  • Einstein Opportunity Scoring

    Rates the chance an open opportunity closes.

    Remember
    Shows positive and negative factors.
    Tested as
    Deal prioritization questions.
    Trap
    Treating the score as stage probability.

  • Global model

    Model trained on aggregated, anonymized data across customers.

    Remember
    Used until the org has enough history.
    Tested as
    New org or low data scenarios.
    Trap
    Assuming scoring cannot start without local data.

  • Sales emails

    Generated drafts from prompt templates grounded in CRM data.

    Remember
    Rep reviews before sending.
    Tested as
    Writing assistance needs.
    Trap
    Sending generated drafts with no review.

  • Record summary

    Generated overview of an account or opportunity.

    Remember
    Saves time before meetings.
    Tested as
    Preparation and handoff scenarios.
    Trap
    Confusing with a report.

  • Agentforce Sales Development

    Autonomous agent that engages and nurtures leads.

    Remember
    Hands qualified leads to human reps.
    Tested as
    Inbound lead volume scenarios.
    Trap
    Expecting it to close deals.

  • Agentforce Sales Coach

    Agent that runs role-play practice and gives feedback.

    Remember
    Grounded in deal context.
    Tested as
    Rep enablement questions.
    Trap
    Confusing with Conversation Insights.

  • Einstein Conversation Insights

    Analyzes recorded calls for mentions and next steps.

    Remember
    Feeds call summaries and coaching.
    Tested as
    Call review scenarios.
    Trap
    Expecting it to place calls.

  • Trusted AI principles

    Responsible, Accountable, Transparent, Empowering, Inclusive.

    Remember
    Guide ethical design decisions.
    Tested as
    Ethics scenario questions.
    Trap
    Mixing them up with the generative AI guidelines.

  • Einstein Trust Layer

    Security and governance between CRM data and models.

    Remember
    Masking, zero retention, toxicity checks, audit trail.
    Tested as
    Privacy concern questions.
    Trap
    Thinking it removes bias.

  • Human in the loop

    A person reviews AI output before action.

    Remember
    Supports accuracy and accountability.
    Tested as
    Customer-facing content scenarios.
    Trap
    Fully automating sensitive messages.

Side by side

Concepts candidates mix up, compared on the point that decides the answer.

Predictive versus generative

ConceptMeaningKey differenceCommon trap
Predictive AI Estimates outcomes from historical patterns. Produces scores and insights. Using it to draft content.
Generative AI Creates text or takes actions with language models. Produces drafts, summaries and conversations. Using it to rank records.
Agentic AI Agents that reason and act through topics and actions. Carries out multistep tasks. Removing human handoff points.

Ethics controls

ConceptMeaningKey differenceCommon trap
Exclude sensitive fields Remove protected attributes from model inputs. Reduces unfair outcomes. Assuming the Trust Layer handles this.
Data masking Hides personal data before it reaches the model. Privacy protection. Expecting it to fix inaccurate output.
Human review Person checks AI output. Accuracy and accountability. Skipping review for customer messages.
Transparency Explaining scores and disclosing AI use. Builds trust. Hiding that an agent is AI.

Exam traps

Each trap is an answer that looks right. Learn the clue that should change your mind.

  1. Generative for ranking

    What candidates pick
    Use generated summaries to decide which leads reps call first.
    Why it looks right
    Generative AI sounds more advanced.
    Why it's wrong
    Ranking needs a predictive model trained on conversions, such as Einstein Lead Scoring.

    Clue to watch for Reps need to prioritize records.

  2. No history, no scores

    What candidates pick
    Delay scoring until the org builds years of data.
    Why it looks right
    Models need training data.
    Why it's wrong
    Einstein can use a global model until the org has enough history, so scoring can start sooner.

    Clue to watch for A new org wants scoring now.

  3. Trust Layer fixes bias

    What candidates pick
    Rely on the Trust Layer to make scores fair.
    Why it looks right
    It is the AI safety feature.
    Why it's wrong
    The Trust Layer protects data and filters toxicity; fairness needs review of model inputs and outcomes.

    Clue to watch for Concern about unfair treatment of a group.

  4. Fully automated outreach

    What candidates pick
    Let generated emails send automatically to save time.
    Why it looks right
    Maximizes efficiency.
    Why it's wrong
    Accuracy and accountability call for a person to review customer-facing content, especially early on.

    Clue to watch for Emails go to customers or prospects.

  5. Agent closes deals

    What candidates pick
    Expect Agentforce Sales Development to negotiate and close opportunities.
    Why it looks right
    It works autonomously.
    Why it's wrong
    It engages and qualifies leads, then hands them to human sellers for the deal.

    Clue to watch for Handoff of qualified leads.

  6. Edit the score

    What candidates pick
    Adjust a lead score field manually to reflect rep opinion.
    Why it looks right
    Reps know their leads.
    Why it's wrong
    Scores are model outputs; improving data and inputs changes them, while manual edits break trust in the model.

    Clue to watch for Reps disagree with scores.

Answer framework

Use these steps when a scenario asks you to recommend or govern AI for a sales team.

  1. Decide whether the need is to predict and prioritize, or to create content and take action, since that splits predictive from generative features.
  2. Identify the specific feature that matches the seller's task, such as lead scoring, opportunity scoring, sales emails, summaries or an Agentforce agent.
  3. Check prerequisites including licenses, permission sets, enabled features and enough clean historical data for any model.
  4. Look for ethical or privacy risks such as sensitive attributes, customer-facing output or personal data, and apply a Trusted AI principle.
  5. Add the governing control, for example human review, transparency to customers, data masking or audit trail monitoring, before rollout.

Practice question

A bank's sales team plans to use Einstein Lead Scoring. The compliance officer worries that scores could disadvantage applicants by age because birth date is stored on the lead. Which action best applies Salesforce's Trusted AI principles?

Want more like this? The Sales-Con-201 practice test has questions for every domain.

Checklist

0 of 11

Tick each one you can do without notes. Your ticks stay in this browser.

Study strategy

Plan seven to nine hours. Start by making a two-column list: predictive features on one side, generative and agentic features on the other, with one line on what each does and what it needs. Complete the Trailhead modules on Einstein for Sales, Agentforce for Sales and responsible AI, and read Salesforce's Trusted AI principles and generative AI guidelines.

If you have access to an org with Agentforce or Einstein features, enable lead scoring, review the factors shown on a lead, and generate a sales email from a template to see grounding at work. If not, use product demos and help articles to trace how each feature appears to a rep.

Then focus on ethics scenarios. For each practice question, name the risk first, such as bias, privacy, accuracy or over-automation, and match it to a principle and a control. Because product names change often, study what a feature does rather than relying on its label alone.

FAQ

How large is the Predictive and Generative AI section of the Agentforce Sales Consultant exam?

Predictive and Generative AI is 13% of the exam, the smallest section, which suggests about eight of the 60 scored questions as a planning estimate.

Do I need the Agentforce Specialist certification first?

No. The only prerequisite is the Platform Administrator credential. This section tests feature fit and responsible use rather than building agents or prompt templates in depth.

Which AI product names should I recognize?

Know Einstein Lead Scoring, Einstein Opportunity Scoring, Einstein Conversation Insights, sales emails, record summaries, Agentforce Sales Development and Agentforce Sales Coach, plus the Einstein Trust Layer.

What are Salesforce's Trusted AI principles?

They are Responsible, Accountable, Transparent, Empowering and Inclusive. Salesforce also publishes generative AI guidelines covering accuracy, safety, honesty, empowerment and sustainability.

Why might lead scores look generic in a new org?

Without enough conversion history, Einstein uses a global model built on aggregated data. Scores become org specific once enough local history exists.

Does the exam test prompt writing?

Only at a high level. You may need to know that sales emails come from prompt templates grounded in CRM data, but detailed Prompt Builder design belongs to the Agentforce Specialist exam.

Other CRM & Business Applications certifications

Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.

Test your Domain 5 knowledge

You have covered generative and agentic sales tools, predictive scoring and insights, and Trusted AI principles. Practice questions that separate prediction from generation and test responsible use.

Take the Sales-Con-201 practice test Check your Sales-Con-201 readiness

Keep going