Request a demo
Salesforce CRMNovember 20, 2026

What is Salesforce Einstein? - How Einstein AI Works, Key Features, Einstein GPT, 2026 Updates

Salesforce Einstein is the family of AI features inside the Salesforce clouds, from lead scores to drafted replies. Its scores learn from the company's own records. In 2026, Salesforce sells much of this AI under the Agentforce name.

13 min read
What is Salesforce Einstein? - How Einstein AI Works, Key Features, Einstein GPT, 2026 Updates
Rakesh Kurhade
Written byRakesh KurhadeCRM Campaign Specialist
Rakesh Kurhade
Reviewed byRakesh KurhadeCRM Campaign Specialist
Rakesh Kurhade
Fact-checked byRakesh KurhadeCRM Campaign Specialist
Last reviewed

Salesforce Einstein is the artificial intelligence (AI) built into Salesforce, the popular cloud customer relationship management (CRM) platform.1 Salesforce introduced it in September 2016 and built it into the records and workflows that teams already use.1 Generative AI followed in March 2023 with Einstein GPT.2 In 2024, its assistant, Einstein Copilot, took the Agentforce name.3

Today, the Einstein name sits mostly on the AI features inside the Salesforce clouds.4 They score leads and deals, sort incoming cases, and draft emails and summaries for the team.4 5 6 7 The scoring features learn from the company's own past records, so the data decides how useful they are.6 7

What is Salesforce Einstein?

Salesforce Einstein is a family of AI features that run inside the Salesforce apps and work on the company's own CRM data.6 8 It is commonly known as Einstein AI.

Einstein is sold in parts. Salesforce licenses it under several names, such as Sales Cloud Einstein, Einstein for Service, and Einstein for Platform.4 That's why two companies on Salesforce can have very different Einstein features in their orgs.

Below are the three parts of Einstein:

  • Predictive AI for scores, predictions, and recommendations
  • Generative AI for emails, replies, and summaries
  • Builder Tools, such as Prompt Builder and Einstein Prediction Builder, for admins

Please note that Salesforce has renamed several Einstein products since 2023. One feature can carry two names in Salesforce's own documents.4 5

Einstein is usually set up by an admin or a consulting partner. Some partners also resell the licenses, which can tilt the advice toward buying more AI. Vegacompass is a CRM-only consultancy and a Salesforce Consulting Partner that earns nothing from Salesforce licenses or AI credits. So the Einstein advice we give starts with your process and your records.

What happened to Einstein GPT?

Einstein GPT was Salesforce's first generative AI product for CRM, launched on 7 March 2023.2 Salesforce called it "the world's first generative AI CRM technology."2 It combined OpenAI's ChatGPT technology with Salesforce's own AI models.2

Einstein GPT for Sales was built to draft emails and help reps prepare for the next interaction. Einstein GPT for Service was built to draft chat replies and write knowledge articles from past case notes.2

Six months later, Salesforce announced Einstein Copilot, an AI assistant built into every Salesforce app.9 In September 2024, Einstein Copilot became part of Agentforce, Salesforce's platform for AI agents.3 Salesforce's own note on the 2023 release now says that Einstein GPT has evolved into Agentforce.2

Please note that the Einstein GPT name still appears in Salesforce's license notices of May 2026.4 It covers features such as sales emails, call summaries, work summaries, and service replies.4

How does Salesforce Einstein AI work?

Salesforce Einstein AI combines two kinds of AI, predictive and generative. Agentforce adds a third kind, agentic AI, on the same platform.5

Below is how each kind works:

  • Predictive AI: Learns from the company's past records, such as converted leads and won deals. It then scores each new record and shows the factors behind the score.6 7
  • Generative AI: Writes new text from a prompt, such as a sales email or a case summary. In Prompt Builder, admins turn prompts into templates filled with data from the record or Data 360.8

Every Prompt Builder template passes through the Einstein Trust Layer before it reaches the large language model (LLM).10 11

Below is what the Einstein Trust Layer does:

  • Grounding: Adds the company's CRM data to the prompt, so the answer fits the record.
  • Data Masking: Hides sensitive data, such as social security numbers, before the prompt is sent.
  • Toxicity Detection: Checks what the model writes for harmful content.
  • Audit Trail: Logs the AI activity and the users' feedback.
  • Zero Data Retention: Commits model providers, such as OpenAI, to keep none of the data.11

Key Features of Salesforce Einstein

Below are the main things Einstein does, with one example each:

  • Predictive Scoring: Ranks leads and deals by how well they match the company's past wins, and shows the top factors behind each score. Einstein Lead Scoring is one example.6 7
  • Predictions: Estimates an outcome before it happens. For example, Einstein Forecasting predicts the team's revenue for the period from past opportunities, account history, and each owner's win rate.12
  • Recommendations: Suggests the next step to the person on the record, such as an offer, an action, or a knowledge article. Next Best Action does this with data insights and business rules.5
  • Classification and Routing: Reads a new case, predicts fields such as its priority or type, and sends it to the right queue.5
  • Drafts and Summaries: Writes emails, chat replies, and call or case summaries for a person to review before sending.4 5
  • Activity and Call Capture: Logs emails and meetings on the related records, and turns call recordings into key moments and action items.13 14
  • Builder Tools: Prompt Builder for an org's own prompts, and Einstein Prediction Builder for its own predictions, both with clicks rather than code.8 15

Salesforce Einstein Across Different Salesforce Clouds

Each Salesforce cloud has its own Einstein features, built around the records that its team works with.

Einstein in Sales Cloud

Sales Cloud's Einstein features help reps decide which leads and deals to work first.6 7

Below are the main features:

  • Einstein Lead Scoring and Einstein Opportunity Scoring for ranking leads and deals
  • Einstein Forecasting for revenue predictions in the current forecast period12
  • Einstein Activity Capture for logging emails and meetings13
  • Conversation Intelligence, formerly Einstein Conversation Insights, for call recordings16
  • Sales Emails and Call Summaries, written with generative AI4

Einstein in Service Cloud

In Service Cloud, each Einstein feature handles a different stage of a case.17 Trailhead now lists them under Agentforce Service, as the next generation of Einstein for Service.5

Below are the features by stage:

  • Case Classification and Case Routing for new cases
  • Article Recommendations for open cases
  • Service Replies and Work Summaries, drafted for the rep to review
  • Knowledge Creation for drafting new knowledge articles5

Please note that Einstein Bots has been in maintenance mode since 17 August 2026. Salesforce now recommends Agentforce for new AI agents.18

Einstein in Marketing Cloud

In Marketing Cloud Engagement, Einstein makes its predictions from each contact's engagement data.19

Below are the main Einstein features for marketers:

  • Einstein Engagement Scoring for segmenting email and mobile audiences20
  • Einstein Send Time and Frequency Optimization for when and how often to message each contact20
  • Einstein Content Selection and Einstein Copy Insights for the content and copy of each email19
  • Einstein Behavior Scoring, a 0 to 100 score for each prospect's engagement in Account Engagement21

Einstein in Commerce Cloud

In Commerce Cloud, Einstein is built into B2C Commerce and learns from each store's own commerce data.22

Below are the shopper and merchandiser features:

  • Einstein Product Recommendations, personal to each shopper22
  • Einstein Predictive Sort for ordering product grids and search results23
  • Einstein Search Recommendations and Einstein Search Dictionaries for site search22
  • Einstein Commerce Insights for basket analysis and product bundles23

Is Salesforce Einstein included in Salesforce CRM?

Partly. Salesforce says some editions already include AI out of the box, and the others can add it at a price.24

Below is how Salesforce sells it in November 2026:

  • Editions: Sales Cloud and Service Cloud are sold in five paid editions, from Starter Suite and Pro Suite to Core, Advanced, and Max. In Sales Cloud, AI can be added to Core and above.25 26
  • Add-ons: The main AI add-ons are Agentforce for Sales and Agentforce for Service. The pricing pages name features such as Lead Scoring and Service Replies without the Einstein prefix.25 26
  • Credits: Core, Advanced, and Max come with Flex Credits for each org, and more can be bought. Agents are billed through credits, resolutions, or per-user licenses.25 27

Please note that Salesforce's license notices still use the Einstein names.4 Before buying an add-on, ask the Salesforce account team which Einstein features your edition already includes.

Salesforce Einstein vs. Agentforce

Salesforce Einstein and Agentforce are easy to mix up, because both are built into the same Salesforce Platform.27

Einstein works inside the record a person has open, with a score, a suggested article, or a drafted reply. The person still makes the call.5 6

Agentforce agents take a request, break it into smaller tasks, and then act within the guardrails set for them.27 Agentforce Assistant, formerly Einstein Copilot, helps employees in the flow of their work.4 27

The difference is simple: Einstein advises the person on the record, while an agent handles the request itself.

To learn more about Salesforce's AI agents, please read our article: Guide to Salesforce Agentforce: Working, Agent Types, Key Features, Use Cases, Benefits.

Salesforce Einstein News and Updates in 2026

In 2026, Salesforce's own pages show the Einstein name giving way to Agentforce and AIforce.

Below are the main updates, as of November 2026:

  • May 2026: Salesforce's license notices, dated 22 May, list Einstein Copilot and Einstein for Flow under their Agentforce names.4
  • August 2026: Einstein Bots went into maintenance mode on 17 August. Existing bots stay supported but get no new features.18
  • September 2026: Salesforce unveiled AIforce at Dreamforce on 15 September. It lets people work with Salesforce data from other tools, such as Slack. The announcement does not mention Einstein.28
  • October 2026: Salesforce's main AI page, updated on 22 October, leads with AIforce and does not name Einstein.29
  • November 2026: The updated Sales Cloud and Service Cloud pricing pages list their AI features without the Einstein prefix.25 26

Note that Salesforce's AI names may change again with the next release. Check the current release notes before you plan an Einstein rollout.

How to get value from Salesforce Einstein

At Vegacompass, we say Einstein is the second step of any Salesforce AI project, never the first. Its scores learn from your own records, so duplicates and empty fields skew them.6 7

Below is the order of work we follow:

  • Records, Process, and Permissions: Clean up the records with a CRM data cleanup, agree on what each stage means, and set who can see each score.
  • Einstein Features: Pilot one feature your edition already includes with one team, judged by one number agreed in advance.
  • Custom Agents: Build an Agentforce agent only where no built-in feature or plain workflow rule covers the job.

Our AI CRM consulting starts with a readiness review of the records, process, and permissions, generally two weeks long. Its output is one written diagnosis.

On the other hand, a team new to Salesforce has little history for the scores to learn from. So, drafts and summaries make the better first pilot.

FAQs

Q. Does Salesforce Einstein use ChatGPT?

Partly. Einstein GPT launched in 2023 on OpenAI's ChatGPT technology, combined with Salesforce's own models.2 Today, Prompt Builder offers models from OpenAI, Anthropic, Google, and others.30 A company can also bring its own model through AI Models, formerly Einstein Studio.30

The scores are different, as they learn from the company's own leads and deals.6 7

Q. How much data does Salesforce Einstein need?

It depends on the feature. Einstein Lead Scoring needs at least 1,000 leads from the last 200 days, with 120 converted.31 Einstein Opportunity Scoring needs at least 200 won and 200 lost opportunities from the last 24 months.31 Below those numbers, Einstein can score with a global model, built on anonymous data from many Salesforce customers.31 32

Q. Is Salesforce Einstein being replaced by Agentforce?

Not entirely. Einstein Copilot became part of Agentforce in 2024, and Salesforce recommends Agentforce over Einstein Bots for new AI agents.3 18 However, the scoring features still carry the Einstein name in Salesforce Help.31 Salesforce's Agentforce page also describes Einstein as natively embedded in the Salesforce Platform.27

Vegacompass: Salesforce Consulting Partner for Einstein and CRM AI

Vegacompass has been a CRM-only consultancy for 11+ years, since 2015. In that time, our team has delivered 120+ CRM projects. As a Salesforce Consulting Partner, we configure Einstein without a stake in the licenses or credits it runs on.

Our consultants work on Salesforce, Microsoft Dynamics 365, HubSpot, ServiceNow, and open-source CRM, from strategy to delivery. On Salesforce, Einstein and Agentforce come in after the records and the process are in order.

We offer the following CRM consulting services:

  • CRM Sales Strategy
  • CRM Customer Service Strategy
  • CRM Audit
  • CRM Data Cleanup
  • Salesforce Implementation
  • CRM Integration
  • AI CRM Consulting

Get in touch

Please tell us which Einstein or Agentforce features your org runs today, and where the scores or drafts fall short. Your Salesforce edition helps too.

We start with a 30-minute call, led by a senior consultant. If a readiness review follows, the senior consultant who runs it stays through the rollout. Our delivery team works from Pune, Palo Alto, Austin, and Singapore.

Talk to a CRM AI consultant today to find out which Einstein features your records are ready for.

Authorship

Rakesh Kurhade (Written by)

Rakesh is a technical content writer. For the past 8+ years, he has been writing practical, plain-language guides on CRM, revenue operations, and AI in CRM for sales, marketing, and RevOps leaders at mid-size and enterprise companies.

References

  1. (September 19, 2016). Salesforce Introduces Salesforce Einstein - Artificial Intelligence for Everyone. Salesforce Newsroom. https://www.salesforce.com/news/press-releases/2016/09/19/salesforce-introduces-salesforce-einstein-artificial-intelligence-for-everyone/
  2. (March 7, 2023). Salesforce Announces Einstein GPT, the World's First Generative AI for CRM. Salesforce Newsroom. https://www.salesforce.com/news/press-releases/2023/03/07/einstein-generative-ai/
  3. (September 12, 2024). Salesforce Unveils Agentforce - What AI Was Meant to Be. Salesforce Newsroom. https://www.salesforce.com/news/press-releases/2024/09/12/agentforce-announcement/
  4. (May 22, 2026). Einstein Platform Notices and License Information. Salesforce. https://www.salesforce.com/en-us/wp-content/uploads/sites/4/documents/legal/misc/einstein-platform-notices-and-license-information.pdf
  5. (2026). Deliver Smarter Service with Artificial Intelligence (AI Solutions for Service: Quick Look). Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/einstein-for-service-quick-look/deliver-smarter-service-with-artificial-intelligence
  6. (2026). Target the Best Leads. Salesforce Trailhead. https://trailhead.salesforce.com/modules/ai_sales/units/ai_sales_leads
  7. (2026). Close More Deals. Salesforce Trailhead. https://trailhead.salesforce.com/modules/ai_sales/units/ai_sales_deals
  8. (2026). Prompt Builder. Salesforce. https://www.salesforce.com/ap/artificial-intelligence/prompt-builder/
  9. (September 12, 2023). Salesforce Announces the New Einstein 1 Platform - Driving Productivity and Trusted Customer Experiences Powered by Data, AI, and CRM. Salesforce Newsroom. https://www.salesforce.com/news/press-releases/2023/09/12/salesforce-platform-news-dreamforce/
  10. (April 2024). Ground Your Prompt Templates with Data Using Flow or Apex. Salesforce Developers Blog. https://developer.salesforce.com/blogs/2024/04/ground-your-prompt-templates-with-data-using-flow-or-apex
  11. (2026). Trust Layer (Agentforce Developer Guide). Salesforce Developers. https://developer.salesforce.com/docs/ai/agentforce/guide/trust.html
  12. (2026). View and Set Sales Forecasts. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/sales-forecasting/view-and-set-sales-forecasts
  13. (2026). Set Up the Gmail Integration and Einstein Activity Capture. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/gmail_integration/gmail_integration_unit_2
  14. (2026). Find Out How Conversation Intelligence Works. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/einstein-call-coaching/optimize-sales-performance-with-einstein-call-coaching
  15. (2026). Use the Einstein Platform. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/get_smart_einstein_feat/use-einstein-platform
  16. (2026). Set Up Conversation Intelligence for Sales. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/set-up-einstein-conversation-insights/set-up-einstein-conversation-insights-for-sales
  17. (2026). Use Artificial Intelligence to Improve Service. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/einstein-for-service-rollout-strategies/use-artificial-intelligence-to-improve-service
  18. (2026). First, What's a Bot? Salesforce Help. https://help.salesforce.com/s/articleView?id=sf.bots_service_whats_a_bot.htm&language=en_US&type=5
  19. (2026). Activate Einstein Features in Marketing Cloud Engagement. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/marketing-cloud-einstein-activation-quick-look/activate-einstein-features-in-marketing-cloud
  20. (2026). Use Einstein Features in Marketing Cloud Engagement. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/trails/use-einstein-features-in-marketing-cloud
  21. (2026). Explore Einstein Features for Account Engagement. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/account-engagement-for-targeted-audiences/explore-einstein-features-for-account-engagement
  22. (2026). Learn About Einstein for Commerce. Salesforce Trailhead. https://trailhead.salesforce.com/en/modules/cc_cccapability/units/cc_cccapability_cceinstein
  23. (2026). Improve Search Recommendations with AI. Salesforce Trailhead. https://trailhead.salesforce.com/content/learn/modules/agentforce-commerce-for-b2c-merchandisers-new/improve-search-recommendations-with-ai
  24. (November 2, 2026). Customer Service AI Pricing. Salesforce. https://www.salesforce.com/products/service-cloud-einstein-pricing/
  25. (November 18, 2026). Salesforce Sales Pricing (India). Salesforce. https://salesforce.com/in/sales/pricing
  26. (November 4, 2026). Service Cloud Pricing (Canada). Salesforce. https://www.salesforce.com/ca/service/pricing/
  27. (November 4, 2026). Agentforce: The AI Agent Platform. Salesforce. https://www.salesforce.com/agentforce/
  28. (September 15, 2026). Salesforce Unveils AIforce, Bringing the Full Power of Its Platform to Any Interface. Salesforce Newsroom. https://www.salesforce.com/news/stories/aiforce-announcement/
  29. (October 22, 2026). AIforce: Salesforce AI, Wherever You Already Work. Salesforce. https://www.salesforce.com/artificial-intelligence/
  30. (2026). Large Language Model Support. Salesforce Help. https://help.salesforce.com/s/articleView?language=en_US&id=ai.generative_ai_large_language_model_support.htm&type=5
  31. (2026). Data Requirements for Sales Cloud Einstein. Salesforce Help. https://help.salesforce.com/s/articleView?id=einstein_sales_data_requirements.htm&language=en_US&type=5
  32. (2026). Considerations for Setting Up Einstein Opportunity Scoring. Salesforce Help. https://help.salesforce.com/s/articleView?language=en_US&id=ai.einstein_sales_setup_opportunity_scoring_considerations.htm&type=5
ai in salesforcesalesforce agentforcesalesforce aisalesforce einsteineinstein aieinstein gpt
Rakesh Kurhade
Rakesh Kurhade

Rakesh is a technical content writer. For the past 8+ years, he has been writing practical, plain-language guides on CRM, revenue operations, and AI in CRM for sales, marketing, and RevOps leaders at mid-size and enterprise companies.

More from Rakesh Kurhade →
Beyond the reading

This article comes from client work. See the services behind it.

CRM strategy consulting and delivery across Salesforce, MS Dynamics, HubSpot, ServiceNow and open-source CRM.

Explore our servicesSee case studies

Leave a comment

Have a question about this, or seen it play out differently? Add your take.

Never published.

Your name and comment are published once reviewed.

Keep reading

All articles →
What is Salesforce Einstein? Einstein AI, Einstein GPT, News