Salesforce Einstein GPT Licensing:
What’s Included vs What Costs Extra

1. The Einstein Licensing Confusion

Salesforce has rebranded its AI capabilities multiple times in three years: Einstein, Einstein GPT, Einstein Copilot, Einstein 1, and now Agentforce. Each rebrand introduced new licensing constructs without fully retiring the old ones. The result is an AI licensing model that even Salesforce’s own account executives struggle to explain clearly — and that creates significant risk for enterprise procurement teams making purchasing decisions.

The core confusion: some Einstein features are included free with your existing Salesforce edition, some require paid permission set licences (PSLs), some require consumption-based credits, and some require an entirely separate product (Data Cloud) as a prerequisite. Salesforce’s marketing materials don’t always make these distinctions clear, and sales presentations frequently blur the line between what’s included and what’s an upsell.

This guide cuts through the branding by organising Salesforce’s AI capabilities into three distinct generations, each with a different licensing model. Understanding which generation a feature belongs to tells you immediately whether it’s included, requires a PSL, or runs on consumption credits. Once you have this framework, the licensing complexity collapses into a navigable structure.

2. Generation 1: Predictive Einstein (Included)

The first generation of Einstein capabilities is predictive AI — machine learning models trained on your Salesforce data to score, rank, and forecast. These features have been available since 2016–2018 and are now included at no additional cost with Enterprise edition and above for most Salesforce clouds.

What’s Included Free

Einstein Lead Scoring: Automatically scores leads based on conversion likelihood, using historical data from your Salesforce org. Included with Sales Cloud Enterprise+. No PSL required. The model trains automatically and updates scores daily. This alone saves sales operations teams weeks of manual lead prioritisation work.

Einstein Opportunity Scoring: Scores open opportunities by win probability. Included with Sales Cloud Enterprise+. Uses signals like stage velocity, activity history, and deal size relative to historical patterns. Surfaced directly in the opportunity record and in pipeline views.

Einstein Case Classification: Predicts case fields (type, reason, priority) based on the case subject and description. Included with Service Cloud Enterprise+. Reduces manual data entry for service agents and improves routing accuracy.

Einstein Prediction Builder: A no-code tool for building custom prediction models on any Salesforce object. Available with Enterprise+. You can create predictions like “likelihood this account will churn” or “probability this renewal will close on time” without writing code or engaging data science resources.

Einstein Discovery (basic): Automated data analysis that surfaces patterns and insights from your Salesforce data. The basic version is included with Enterprise edition; the advanced version with prescriptive recommendations requires a paid licence.

These features represent genuine, production-ready AI that most enterprises underutilise. Before purchasing any paid Einstein upgrade, ensure your teams have fully adopted the included predictive features — they deliver measurable value at zero incremental cost.

3. Generation 2: Einstein GPT / Generative AI (Paid PSLs)

The second generation, introduced in 2023–2024, brought generative AI capabilities powered by large language models. These features create content — drafting emails, summarising cases, generating knowledge articles, writing call summaries — rather than scoring or classifying existing records. Generative AI features universally require paid licensing, either through PSLs or through the Einstein 1 edition.

Key Generative AI Features and Their Licensing

Einstein GPT for Sales (Sales Email Generation, Call Summaries, Account Research): Drafts personalised outreach emails, generates post-call summaries, and creates account research briefs. Requires the Einstein for Sales PSL at approximately $75/user/month (list), or is included in the Einstein 1 Sales edition ($500/user/month). Available for Enterprise edition and above.

Einstein GPT for Service (Reply Recommendations, Case Summaries, Knowledge Generation): Drafts customer replies, summarises lengthy case histories, and generates knowledge articles from resolved cases. Requires the Einstein for Service PSL at approximately $75/user/month, or is included in Einstein 1 Service edition.

Einstein Copilot (Conversational AI Assistant): A natural-language interface within Salesforce that lets users ask questions, trigger actions, and navigate data through conversation. Available as part of the Einstein 1 editions or with specific Copilot PSLs. Copilot has been rebranded as Agentforce Assistant in the Spring ’26 release but the licensing model remains the same.

Einstein GPT for Developers (Code Generation): Generates Apex code, SOQL queries, and Lightning component code from natural-language descriptions. Requires specific developer-tier licensing and is targeted at Salesforce development teams rather than end users.

4. Generation 3: Agentforce (Consumption Credits)

The third and newest generation, launched in late 2024 and expanding rapidly through 2025–2026, introduces autonomous AI agents that act independently rather than assisting human users. Agentforce agents can qualify leads, resolve service cases, schedule meetings, and execute multi-step workflows without human intervention. The licensing model is fundamentally different from generations 1 and 2: it is consumption-based, priced in Flex Credits.

How Flex Credits Work

Each AI agent action consumes credits at $0.10 per credit. An “action” is a single AI operation: a conversation turn, a record lookup, a field update, an email send, or a knowledge retrieval. A typical autonomous case resolution might consume 5–10 credits ($0.50–$1.00). A multi-turn lead qualification conversation might consume 15–25 credits ($1.50–$2.50).

Enterprises purchase Flex Credits in committed volumes (monthly or annual minimums) with overage rates that are typically 10–20% above the committed rate. The minimum Flex Credit commitment varies by deal but typically starts at $10,000–$25,000/month for enterprise agreements. Unused credits may or may not roll over depending on your negotiated terms — always negotiate rollover provisions.

The consumption model creates budgeting uncertainty that per-user licensing doesn’t. A 200-agent service team that deploys Agentforce for case deflection might consume $15,000/month in credits during a quiet period and $45,000/month during a seasonal spike. Without consumption caps and overage protections, costs can escalate rapidly. This is the single most important commercial risk in the current Salesforce AI licensing landscape.

5. Edition-by-Edition: What AI Is Included Where

The following breakdown maps which Einstein capabilities are available in each major Salesforce edition, helping you determine whether your current licence already includes the AI features you need — or whether an upgrade is required.

The critical insight: even the $500/user/month Einstein 1 edition does not include Agentforce consumption credits. Flex Credits are always an additional, consumption-based cost regardless of your edition. Einstein 1 includes the generative AI features that would otherwise require $75–$100/user PSLs, plus Data Cloud — but autonomous AI agents are priced separately. This is the most commonly misunderstood aspect of Salesforce’s AI pricing.

6. The Full Included vs Extra Matrix

The following matrix provides the definitive reference for which AI capabilities are included and which require additional licensing across the four primary editions.

This matrix reveals the commercial logic behind the Einstein 1 edition: it bundles $150–$200/user/month of PSLs into a single $500 SKU that also includes Data Cloud. For organisations deploying 3+ generative AI features across the entire user base, Einstein 1 can be more cost-effective than Enterprise + stacked PSLs. But for most enterprises, Enterprise edition with 1–2 selective PSLs deployed to a subset of users is significantly cheaper than upgrading the entire org to Einstein 1. The math only favours Einstein 1 when the majority of your users will actively use multiple generative AI features — which, based on our client data, is rare.

7. The Data Cloud Prerequisite

Salesforce Data Cloud has become the de facto prerequisite for getting the best results from generative AI and Agentforce. While Einstein’s predictive features (Gen 1) train on your native Salesforce data and require no additional data infrastructure, the generative and autonomous features (Gen 2 and Gen 3) perform significantly better when grounded in the unified customer profiles that Data Cloud provides.

Salesforce positions Data Cloud as “included” in various contexts, but the licensing reality is nuanced. Data Cloud is included with Einstein 1 editions — but with usage limits (volume of data ingested, number of segments, activation events) that enterprise deployments frequently exceed. Data Cloud is a separate product for all other editions, starting at approximately $108,000/year for the base tier and scaling based on consumption.

The practical question: do you need Data Cloud to use Einstein? For predictive features, no. For basic generative AI (email drafting, case summaries), the features function without Data Cloud but produce generic outputs that lack customer context. For Agentforce, Data Cloud is effectively required — autonomous agents need the unified customer profile to personalise interactions and make accurate decisions.

Budget Data Cloud into your AI cost model if you plan to deploy Agentforce or if you want generative AI that is genuinely personalised rather than template-driven. Negotiate Data Cloud as part of your overall Salesforce agreement, not as a standalone purchase. For the full commercial analysis, see our AI & Data Cloud Licensing guide.

8. Total Cost: Modelling Your Einstein Spend

The following scenarios illustrate the realistic total cost of Salesforce AI for three representative enterprise configurations, demonstrating how quickly costs escalate as you move beyond the included predictive features.

The pattern is clear: selective deployment on Enterprise edition is almost always more cost-effective than upgrading to Einstein 1, unless more than 70% of your user base genuinely needs and will use multiple generative AI features. Based on our advisory work, the typical enterprise AI adoption rate in Year 1 is 25–40% of licenced users. Paying for Einstein 1 across the entire org means paying premium pricing for AI features that the majority of your users won’t touch.

9. Negotiating Einstein and Agentforce Licensing

Einstein PSL Negotiation

Never purchase Einstein PSLs standalone. Bundle them into your base Sales Cloud or Service Cloud renewal. Standalone Einstein PSL purchases typically attract minimal discounting (10–15% off list). Bundled into a larger deal, we see 25–40% discounts on Einstein PSLs.

Deploy to the minimum viable user population. Salesforce will push for org-wide deployment (“all 500 users need Einstein for Sales to maximise AI adoption”). Resist. Start with the user population that will demonstrably use the features — typically sales leadership, RevOps, and top-performing reps — and expand based on measured adoption. Negotiate the right to add PSLs mid-term at the same discount rate.

Insist on PSL reduction rights. If you purchase 100 Einstein PSLs and adoption plateaus at 60, you need the contractual right to reduce at your annual anniversary. Without this, you’re locked into paying for unused AI licences for the duration of the term. Salesforce resists reduction rights on PSLs even more than on base licences — push back harder, referencing our Contract Terms FAQ.

Agentforce / Flex Credit Negotiation

Pilot before committing. Negotiate a 90-day pilot with a capped credit budget ($5,000–$15,000/month) before signing a long-term Flex Credit commitment. Use the pilot to measure actual credit consumption per workflow and extrapolate annual costs. Salesforce’s consumption estimates are frequently optimistic — real-world usage often exceeds projections by 30–50%.

Negotiate credit rollover. Standard Flex Credit terms may not include rollover of unused credits. Insist on full monthly rollover with an annual true-up, or negotiate quarterly rollover at minimum. Without rollover, you lose any credits not consumed within the period — creating a “use it or lose it” dynamic that incentivises wasteful consumption.

Include consumption caps. Negotiate a hard monthly cap above which Salesforce provides credits at no additional charge, or at a significantly discounted overage rate. Without caps, a seasonal spike in AI agent activity can generate an invoice that exceeds your budget by 50–100%. Caps protect against the most dangerous aspect of consumption pricing: unpredictability.

Tie Agentforce to seat reduction. If Agentforce agents are handling workloads previously performed by human users, your per-user seat count should decrease proportionally. Negotiate explicit seat-to-credit conversion provisions in your Flex Agreement. For every 10 seats you reduce, a defined credit allocation should be added to your Flex balance at a favourable conversion ratio. This ensures that Agentforce delivers genuine cost substitution rather than additive cost. For the complete negotiation framework, see our Salesforce Renewal pillar guide and competitive pricing playbook.

10. The Build vs Buy Decision

Before committing to Salesforce’s AI licensing stack, evaluate whether alternative approaches deliver comparable value at lower cost. The emergence of powerful third-party AI tools and enterprise LLM APIs means Salesforce is no longer the only path to AI-powered CRM.

Third-party AI tools like Gong (conversation intelligence), Clari (revenue intelligence), Outreach (AI-powered engagement), and Drift (conversational AI) provide specialised capabilities that often exceed Salesforce’s native features in their specific domains. These tools integrate with Salesforce via APIs and can be more cost-effective than Einstein PSLs for specific use cases.

Enterprise LLM APIs (OpenAI, Anthropic, Google) can power custom AI workflows within Salesforce through Apex callouts and Flow integrations. A custom-built email drafting tool using an LLM API costs significantly less per interaction than a $75/user/month Einstein for Sales PSL, though it requires development investment and ongoing maintenance. For organisations with Salesforce development capacity, this approach can deliver 60–80% of Einstein’s generative capabilities at 20–30% of the licensing cost.

Our recommendation: Use the included predictive Einstein features (free). Evaluate third-party alternatives for each specific generative AI use case before purchasing Einstein PSLs. Pilot Agentforce before committing significant credit budgets. And negotiate everything as part of a single bundled deal at renewal, using competitive alternatives as leverage. The enterprises that achieve the best AI outcomes and the best AI economics are those that deploy selectively, measure relentlessly, and never accept Salesforce’s AI pricing at face value.

For deals involving AI licensing above $200K annually, consider engaging independent advisory support. Our advisory practice includes Einstein-specific benchmarking, PSL utilisation audits, and Flex Credit negotiation as part of every Salesforce renewal engagement. View our case studies for documented engagement outcomes.