Einstein Copilot is now Agentforce. Salesforce renamed its AI assistant in January 2025, and its Agentforce page lists "Agentforce (formerly Einstein Copilot)". Salesforce says Einstein GPT, announced in March 2023, has also evolved into Agentforce. In November 2026 you turn it on through Agentforce Studio, configure it in Agent Builder and pay through Flex Credits or per-user licenses.
We first published this article in August 2024, when Einstein Copilot was new. It was rewritten in November 2026 to show what each Copilot feature is called today and how to use it. For a full introduction to the platform, read what Agentforce is and how it works, or see our Agentforce services.
What happened to Einstein Copilot?
Einstein Copilot was Salesforce's conversational AI assistant for employees. It sat inside Salesforce, understood plain-language requests, and answered or acted using your CRM data through a library of actions. Salesforce changed its name twice:
- Einstein Copilot was the original name. Salesforce's own Einstein Copilot article was first published on March 6, 2024.
- Agentforce Assistant was the interim name. Salesforce retitled its Copilot article "What Is Agentforce Assistant? (Formerly Einstein Copilot)".
- Agentforce is the current name. Salesforce's AI copilot page says the tool previously known as Einstein Copilot was renamed to Agentforce as of January 2025, and the editor's note on the Einstein Copilot article says Agentforce Assistant has been upgraded to Agentforce.
The change was more than a rename. Copilot answered when an employee asked. Agentforce is a platform for building agents that can also work on their own: Salesforce's Agentforce page (checked November 11, 2026) describes it as an AI agent platform for autonomous support at enterprise scale and says over 18,000 companies run on it.
What happened to Einstein GPT?
Salesforce announced Einstein GPT on March 7, 2023, calling it the world's first generative AI CRM technology. It launched in closed pilot, promised AI-created content across every sales, service, marketing, commerce and IT interaction, and combined Salesforce's own Einstein models with OpenAI's models, with the option to bring an external model.
Einstein GPT never became a product you buy today. The editor's note on that announcement now says Einstein GPT has evolved into Agentforce. The ideas it introduced, such as grounding AI in CRM data and letting customers choose the model, carried through Einstein Copilot into Agentforce.
What is the Einstein Copilot to Agentforce timeline?
| Date | What changed | Source |
|---|---|---|
| March 7, 2023 | Einstein GPT announced as a closed pilot | Salesforce press release |
| March 6, 2024 | Salesforce publishes its Einstein Copilot article | Salesforce Einstein Copilot page |
| Not dated by Salesforce | Interim name Agentforce Assistant (formerly Einstein Copilot) | Salesforce Einstein Copilot page |
| January 2025 | Einstein Copilot renamed Agentforce | Salesforce AI copilot page |
| October 14, 2025 | OpenAI models available in Agentforce, selectable as the default model for the Atlas Reasoning Engine and Prompt Builder | Salesforce press release |
| November 2026 | Agentforce page lists "Agentforce (formerly Einstein Copilot)" among its agent types; setup runs through Agentforce Studio | Salesforce Agentforce page; Trailhead |
For what Salesforce announced at its most recent event, see our Dreamforce 2026 recap.
What is each Einstein Copilot feature called now?
If you read older documentation or configured Copilot in 2024, this table maps the old names to the 2026 ones.
| Einstein Copilot (2024) | Agentforce (November 2026) |
|---|---|
| Einstein Copilot, the assistant | Agentforce, listed by Salesforce as "Agentforce (formerly Einstein Copilot)" |
| Copilot Builder, a library of actions users chose from to customize the assistant | Agent Builder, which Salesforce describes as configuring Agentforce, subagents, actions and instructions |
| Library of standard and custom actions | Actions grouped into subagents, reviewed under Agentforce Assets (for example the General CRM subagent) |
| Prompt Builder | Prompt Builder, still the low-code tool for reusable prompt templates |
| Bring your own model | Model choice, including the option to select OpenAI as the default model for the Atlas Reasoning Engine and Prompt Builder |
| Copilot's reasoning engine | The Atlas Reasoning Engine, which breaks a request into smaller tasks and checks each step |
| No equivalent | Agent Script, for precise control over how an agent behaves |
| Einstein Trust Layer | Einstein Trust Layer, same name, with one agent-specific caveat (see below) |
Salesforce also offers Agentforce Coworker, which Trailhead describes as an AI teammate built into Salesforce Global CRM Search that reasons across your data and routes requests to specialized agents, in Lightning and the Salesforce mobile app. Salesforce does not describe it as Copilot's direct successor, but it covers much of the same ground for employees.
How do you turn on Agentforce in place of Einstein Copilot?
The 2024 steps (turn on Generative AI, toggle Einstein Copilot, assign a Copilot permission set) no longer match Setup. Trailhead's current quick start uses this path:
- In Setup, search Quick Find for Salesforce Go.
- In the Search features box, select Agentforce Studio.
- Click Get Started, then Turn On. This enables Einstein Setup and Agentforce Agents. Confirm the popup.
- Check that the Turn on Agentforce Studio toggle is on, then refresh your browser.
- Open Agentforce Assets to review the default subagents, such as General CRM, and their actions.
- Configure the agent in Agent Builder, give users access through permission sets, and test before you roll it out.
Before step one, check your licenses and data. Our Agentforce readiness checklist covers data quality, permissions and use case selection.
What can Agentforce do that Einstein Copilot did?
The use cases from 2024 still apply. What changed is that agents can carry a task further without a person driving each step.
| Team | With Einstein Copilot (2024) | With Agentforce (2026) |
|---|---|---|
| Sales | Summarize an account, draft a follow-up email | The same, plus agents that work inbound leads and book meetings |
| Service | Find answers from knowledge for a service rep | Customer-facing agents that resolve requests and hand off to people with context |
| Marketing | Draft campaign content and surveys | Draft and adapt campaign content grounded in customer data |
| Operations and IT | Answer questions about records | Employee agents that answer questions and automate routine tasks |
For the sales side in detail, see Agentforce for sales. For service teams, read Agentforce for customer service. For internal help desks, see Agentforce employee agents for IT and HR.
How much does Agentforce cost compared with the old Copilot model?
Salesforce's Agentforce pricing page (read November 11, 2026, and marked as subject to change) lists these options:
| Option | Listed price |
|---|---|
| Salesforce Foundations | $0 |
| Flex Credits | $500 per 100,000 credits |
| Conversations | $2 per conversation |
| Agentforce User License (requires Flex Credits) | $5 per user per month |
| Agentforce Coworker User License | $50 per user per month |
| Agentforce add-ons (Sales, Service, Field Service and Operations) | $125 per user per month |
| Agentforce Industries add-ons | $150 per user per month |
| Max Edition (includes the Agentforce add-on and 2.75 million Flex Credits per org per year) | From $550 per user per month |
The page also notes that Flex Credits and Conversations cannot be used in the same org. Our Agentforce pricing guide explains which model fits which use case, and the Agentforce cost calculator gives a first estimate for your volumes.
Is Agentforce as secure as Einstein Copilot?
Agentforce runs on the same Einstein Trust Layer that protected Copilot. Trailhead lists secure data retrieval, dynamic grounding, data masking, zero data retention, toxic language detection and an audit trail. One difference matters if you relied on Copilot's field masking: Trailhead states that data masking for LLMs is currently disabled for agents (read November 11, 2026).
That puts more weight on design. Decide which records the agent user can see, which actions it can call and when it must hand off to a person. Our guide to Agentforce governance, security and testing covers the controls to put in place.
What should you do with an existing Einstein Copilot setup?
- Inventory what you built. List custom actions, prompt templates, permission sets and the users who rely on them.
- Map each item to its Agentforce home. Actions move into subagents, prompt templates stay in Prompt Builder, and instructions belong in Agent Builder or Agent Script.
- Check licensing. Confirm which Agentforce pricing model your contract uses before you widen access.
- Retest with real requests. Replay the questions employees asked Copilot and check every answer and action.
- Review data access. With LLM data masking disabled for agents, confirm the agent user only sees the fields it needs.
- Update training and names. Replace Copilot references in internal guides so users know what to open.
Agents are only as good as the data they read. If customer data sits in several systems, see how Data 360 grounds Agentforce. For the wider family of Einstein features, including which ones are still current, read our guide to Salesforce Einstein features.
How can Techila help you move from Einstein Copilot to Agentforce?
Techila Global Services is a Salesforce Summit Partner headquartered in Frisco, Texas, with 15+ Agentforce programs delivered and 500+ Salesforce certifications across the team. Our architect-led Agentforce implementation team reviews existing Einstein Copilot configurations, moves them to Agentforce, and tests agents against real requests before rollout. Talk to a Salesforce architect about moving from Einstein Copilot to Agentforce. We reply within 1 business day.
Frequently asked questions
What happened to Einstein Copilot?
Salesforce renamed Einstein Copilot to Agentforce as of January 2025. For a time Salesforce pages called it Agentforce Assistant, and Salesforce now says Agentforce Assistant has been upgraded to Agentforce. Its Agentforce page, checked November 11, 2026, lists "Agentforce (formerly Einstein Copilot)" among its agent types, as the agent that supports employees with specific tasks.
Is Einstein GPT still available?
Not under that name. Salesforce announced Einstein GPT on March 7, 2023 as a generative AI CRM technology in closed pilot. Salesforce's own editor's note on that announcement now says Einstein GPT has evolved into Agentforce. The generative features it promised, such as drafting emails and summaries, are delivered today through Agentforce and Prompt Builder.
Is Agentforce Assistant the same as Einstein Copilot?
Yes. Agentforce Assistant was the interim name Salesforce used for the product formerly called Einstein Copilot. Salesforce's Einstein Copilot article now carries the note that Agentforce Assistant has been upgraded to Agentforce. If you see any of the three names in documentation, they refer to the same conversational assistant for employees inside Salesforce.
How do I turn on the Einstein Copilot replacement in my org?
Trailhead's current steps are: in Setup, open Salesforce Go, select Agentforce Studio, click Get Started and then Turn On, which enables Einstein Setup and Agentforce Agents. Confirm, check that the Turn on Agentforce Studio toggle is on and refresh. Then review the default subagents and actions under Agentforce Assets. After that, configure the agent in Agent Builder.
How much does Agentforce cost in 2026?
As of November 11, 2026, Salesforce lists Flex Credits at $500 per 100,000 credits, $2 per conversation, Agentforce add-ons at $125 per user per month, an Agentforce User License at $5 per user per month (requires Flex Credits) and Salesforce Foundations at $0. Salesforce marks these prices as subject to change, so confirm with your account team.
Do I need to rebuild what we configured in Einstein Copilot?
Usually not from scratch. Actions, prompt templates and permissions carry the same ideas into Agentforce, but the building blocks now have new names: subagents, actions, instructions and Agent Script. Review each custom action, test it in Agent Builder, and retire prompts that no longer match how your teams work.
