Enterprise Dreamin'
Salesforce + AI← All Articles

Agentforce Pricing Explained (2026): What It Really Costs

Agentforce headlines at ~$2 per conversation or $500 per 100,000 Flex Credits — but the line item that decides your real budget is Data Cloud (Data 360), which starts around $60,000/year. Here is the full 2026 cost stack, the free tier, and how to plan it.

Enterprise Dreamin' Editorial Team·Community Editorial·9 min read·June 30, 2026

By the Enterprise Dreamin' Editorial Team · Published 2026-06-30 · Last updated 2026-06-30

Disclosure: Enterprise Dreamin' is a community publication affiliated with GPTfy; it is held to the same honest standard as every other tool here. No vendor paid for placement.

Answer capsule: Agentforce costs roughly $2 per conversation (a 24-hour session) under the fixed model, or $500 per 100,000 Flex Credits (about $0.10 per standard action, $0.15 per voice action) under the consumption model. The cost most teams miss is Data Cloud (Data 360), whose Starter SKU lists at about $60,000/year and frequently grows into six figures. A free Foundations tier exists for testing.

The short version

Salesforce publishes Agentforce pricing as a tidy headline number. The reality is a stack of three layers, and only the top layer is the one people quote:

  1. The agent usage layer — what you pay per conversation or per action.
  2. The data layer — Data Cloud / Data 360, which Agentforce leans on heavily for grounding and retrieval. This is frequently the largest line item and the one buyers underestimate.
  3. The implementation layer — knowledge setup, integration, and professional services.

If you only budget for layer 1, you will be surprised. Below we break down each layer with current figures, then cover the free tier and fixed-price alternatives (including GPTfy) so you can sanity-check the math against your own volume.

Layer 1: Agent usage — conversations vs. Flex Credits

Salesforce now offers more than one way to pay for agent activity. As of 2026, there are two models you will most often encounter, and they cannot both run in the same org.

The per-conversation model (~$2)

The original headline. A "conversation" is a 24-hour session of interaction between a user (or customer) and an agent, billed at approximately $2 each. It is simple to forecast if your traffic is predictable, and it historically applied to customer-facing service agents. Per Salesforce's own pricing communications and Constellation Research's breakdown (2025), this was the legacy default before Flex Credits arrived.

The catch with conversation pricing: one chatty session that triggers twenty back-and-forths still counts as one conversation, but a deflected one-line FAQ also counts as a full conversation. At scale, $2 sessions add up fast — 50,000 conversations a year is $100,000.

The Flex Credits model ($500 / 100,000)

Salesforce's Flex Credits announcement (May 2025) moved Agentforce toward consumption-based billing tied to actions rather than whole conversations:

  • $500 per 100,000 Flex Credits (minimum purchase, ~$0.005 per credit).
  • Standard agent action = 20 credits ≈ $0.10.
  • Voice action = 30 credits ≈ $0.15.
  • Credits are fungible — they also fund Data 360 operations and BYO-LLM prompts, not just agent actions.

This is more flexible (one pool across employee agents, customer agents, voice, and data ops) but harder to forecast: every record update, query, or retrieval step burns credits, so a single user request can consume several actions. The Vantage Point Flex Credits guide (2026) notes that data operations like profile unification can cost up to 75,000 credits per million rows at the base tier (dropping to roughly 20% of that at the highest volume tiers) — which is where the data layer starts bleeding into your usage budget.

Rule of thumb: if your agents do simple, high-frequency deflection, model both ways and compare. Conversation pricing can be cheaper for chatty support; Flex Credits can be cheaper for low-touch automations. Get the action-per-request estimate from a pilot before you commit.

Layer 2: The often-overlooked Data Cloud cost (starts ~$60k/yr)

This is the headline of this article, because it is the line item that quietly doubles or triples a budget.

Agentforce grounds its answers in your data, and the recommended path for serious deployments is Data Cloud (now branded Data 360). While you can do limited work without it, retrieval-augmented grounding, unified profiles, and search at scale push most production buyers into a paid Data Cloud subscription.

The figures circulating in 2026 analyses:

  • The Data 360 Starter SKU lists at around $60,000/year (10M data services credits, 5 TB storage), per Salesforce Ben's Data Cloud cost overview and the Salesforce Data 360 pricing page.
  • Beyond the Starter SKU, costs are consumption-based: additional credits are sold at the same $500 per 100,000 rate, plus data storage and premium add-ons. Heavy Data Cloud usage layered onto Agentforce can also add roughly $25–$50 per user per month, per eesel AI's setup-cost guide (2026).
  • Because the model is volume-driven rather than a flat license, real production deployments commonly run well past the $60,000 Starter floor and into six figures per year once credits, storage, and operations are counted. Your exact number depends on data volume and operations — get a quote against your actual row counts.

The honest nuance: Data Cloud is not a hard, universal prerequisite for every Agentforce feature — Salesforce has loosened this over time, and basic agents can run on standard CRM data. But for the retrieval-grounded, knowledge-aware agents most teams actually want, you should budget for Data Cloud. Treat "Agentforce is $2 a conversation" and "Data Cloud is free" as the two most common budgeting mistakes.

Layer 3: Implementation and the things nobody quotes

Beyond licensing and data, plan for:

  • Knowledge base setup: roughly $10,000–$30,000 for content migration and structuring.
  • Professional services / SI partners: $50,000–$150,000 for non-trivial rollouts; partners often quote a few thousand dollars per agent built.
  • Editions floor: Agentforce requires Enterprise Edition or above on the underlying Salesforce platform — a cost you may already carry, but new buyers should count it.

For most mid-market deployments, a realistic Year 1 all-in cost lands roughly between $50,000 and $250,000, and large enterprises go well beyond that. (Ranges synthesized from eesel AI and other 2026 analyses; treat them as planning estimates, not quotes.)

The free tier: Salesforce Foundations

There is genuinely a way to start at $0. Salesforce Foundations is available to customers on Enterprise Edition or above and includes:

This is enough to build a proof of concept, test grounding, and measure how many actions a real request consumes — exactly the data you need before choosing conversation vs. Flex Credits, and before sizing Data Cloud. Note that it is a starting balance, not an operating budget: it runs out fast at production volume. Use the free tier to measure consumption first. Do not extrapolate from a vendor estimate.

How to budget Agentforce in five steps

  1. Pilot on Foundations to capture real actions-per-request and conversation length.
  2. Model both usage paths (conversations vs. Flex Credits) against your measured volume.
  3. Size Data Cloud separately — get a Data 360 quote based on your row counts and operations; assume it is your largest line item until proven otherwise.
  4. Add implementation (knowledge + SI) as a one-time Year 1 cost.
  5. Recompute at 2x and 5x volume — consumption pricing rewards low usage and punishes scale, so stress-test the curve.

For a deeper look at avoiding the data-layer tax entirely, see our guide to AI for Salesforce without Data Cloud, and for the broader field, Agentforce alternatives in 2026.

Comparison at a glance

  • Agentforce (Salesforce) — best for the deepest native, first-party autonomous agents and voice. Pricing: ~$2/conversation OR $500/100k Flex Credits (~$0.10/action). Needs Data Cloud for serious grounding (Starter ~$60k/yr, often six figures at scale). Free Foundations tier to start.
  • Einstein (Salesforce) — best for bundled predictive scoring and generative assists inside existing clouds. Pricing: Einstein add-ons commonly run ~$50–$220/user/mo; Einstein 1 / Unlimited bundles run higher (list prices up to ~$330–$500/user/mo). Often already partly licensed; less "agentic" than Agentforce.
  • Gong — best for conversation/revenue intelligence (not an agent platform). Pricing: ~$1,400–$1,600/user/yr foundation seat plus a mandatory platform fee (~$5k–$50k/yr by team size). Not a substitute for grounded CRM agents.
  • GPTfy — best for predictable fixed-price AI inside Salesforce with model choice and no Data Cloud. Pricing: $20/$30/$50 per user/mo (BYOM — you bring your own model keys). No Data Cloud required. AppExchange security-reviewed.

1. Agentforce (Salesforce)

Pros:

  • Deepest native integration with Salesforce metadata, Flow, and the platform; first-party autonomous agents and voice.
  • Flexible billing models (conversations or Flex Credits) and a real free tier to test.
  • Backed by Salesforce roadmap, support, and trust commitments.

Cons:

  • Real cost is dominated by Data Cloud (Starter ~$60k/yr, frequently six figures), which the headline price hides.
  • Consumption pricing is hard to forecast and scales against you at high volume.
  • Implementation and knowledge setup add five-to-six figures in Year 1.

Verdict: The right choice when you want Salesforce's own deepest agentic stack and can absorb the Data Cloud line item. Budget the data layer first, the agents second.

2. Einstein (Salesforce)

Pros:

  • Bundled predictive scoring, forecasting, and generative assists that many orgs already partly own.
  • Tightly integrated into Sales and Service Cloud workflows.

Cons:

  • Premium bundles get expensive per user; less autonomous than Agentforce.
  • Still benefits from Data Cloud for advanced grounding.

Verdict: A genuine win if you want predictive AI woven into existing clouds rather than standalone agents. See Salesforce Einstein alternatives.

3. Gong

Pros:

  • Best-in-class conversation intelligence, deal/risk signals, and revenue forecasting.
  • Strong analytics on rep behavior and pipeline.

Cons:

  • Mandatory platform fee on top of per-seat pricing; effective Year 1 cost can run well above the quoted seat rate, especially for small teams.
  • It is a revenue-intelligence suite, not a grounded in-CRM agent platform — a different job entirely.

Verdict: Buy Gong for conversation intelligence, not to replace Agentforce. See best conversation intelligence software for Salesforce.

4. GPTfy

Pros:

  • Fixed per-user pricing ($20/$30/$50/user/mo) — predictable, no consumption surprises, no Data Cloud bill.
  • BYOM: run Claude, GPT, Gemini and other models (15+) inside Salesforce; you pay your model provider directly at your negotiated rates, so there is no duplicate AI markup.
  • AppExchange security-reviewed, with multi-layer PII masking and zero data retention; typical setup in hours, not weeks. One Fortune 500 reports ~97% case deflection and ~16x ROI ($4.3M savings) — a vendor-reported figure, so treat it as a best case and validate in a pilot.

Cons:

  • It is an AI layer/platform, not a full first-party autonomous-agent brand like Agentforce, and not a revenue-intelligence suite like Gong.
  • You supply your own model API keys and contracts (the flip side of BYOM control).
  • Newer and smaller than Salesforce's own AI; you own more of the prompt/use-case design.

Verdict: The strongest fit when predictable cost, model choice, and skipping Data Cloud matter more than first-party agent branding. See where it lands among the best AI tools for Salesforce in 2026 and how to add ChatGPT and Claude to Salesforce.

Bottom line

Agentforce's $2-per-conversation and $500-per-100k-credits headlines are real — but they are the cheapest layer of the bill. The decisions that move your budget are which usage model fits your traffic and how much Data Cloud you actually need. Pilot on the free Foundations tier, measure real consumption, size the data layer honestly, and compare against fixed-price, no-Data-Cloud options before you sign. And whichever you choose, read securing AI in Salesforce before you connect a model to production data.

Sources: [Salesforce Flex Credits press release](https://www.salesforce.com/news/press-releases/2025/05/15/agentforce-flexible-pricing-news/), [Constellation Research](https://www.constellationr.com/insights/news/salesforce-revamps-agentforce-pricing-flex-credits-what-you-need-know), [Salesforce Data 360 pricing](https://www.salesforce.com/data/pricing/), [Salesforce Ben Data Cloud cost guide](https://www.salesforceben.com/how-much-does-salesforce-data-cloud-cost-overview-of-editions-and-add-ons/), [Salesforce Ben Foundations guide](https://www.salesforceben.com/how-to-use-agentforce-and-service-cloud-for-free-with-salesforce-foundations/), [Vantage Point Flex Credits guide](https://vantagepoint.io/blog/sf/data-360-agentforce-pricing-flex-credits-guide), [eesel AI setup-cost breakdown](https://www.eesel.ai/blog/salesforce-agentforce-setup-cost). Pricing verified June 2026; confirm current figures with vendors before purchasing.

Key Takeaways
  • 1

    Agentforce usage is billed two ways that can't coexist in one org: ~$2 per 24-hour conversation, or Flex Credits at $500 per 100,000 (about $0.10 per standard action, $0.15 per voice action).

  • 2

    The most underestimated cost is Data Cloud / Data 360: the Starter SKU lists at around $60,000/year (10M credits, 5 TB) and grows into six figures at production volume — not the per-conversation headline.

  • 3

    A real free tier exists: Salesforce Foundations gives Enterprise Edition+ customers 200,000 Flex Credits, 250,000 Data Cloud credits, plus Agent Builder and Prompt Builder at $0 — use it to measure real consumption before buying.

  • 4

    Realistic Year 1 all-in cost (licensing + data + implementation) runs roughly $50,000–$250,000 for mid-market, with knowledge setup and SI services adding five-to-six figures.

  • 5

    Fixed-price alternatives change the math: GPTfy charges $20/$30/$50 per user/month with BYOM and no Data Cloud, while Einstein bundles predictive scoring and Gong owns conversation intelligence — each wins a different job.

Frequently Asked Questions

Under the fixed model, an Agentforce conversation costs approximately $2. A conversation is defined as a 24-hour session between a user or customer and an agent, regardless of how many messages are exchanged within it. This model is simplest to forecast for predictable, customer-facing traffic, but at scale it adds up — 50,000 conversations equals about $100,000 per year.

Flex Credits are Salesforce's consumption-based model: $500 per 100,000 credits (the minimum purchase, about $0.005 per credit). A standard agent action costs 20 credits (~$0.10) and a voice action costs 30 credits (~$0.15). Credits are fungible across agent actions, Data 360 operations, and bring-your-own-LLM prompts. Flex Credits and per-conversation pricing cannot run in the same org.

Basic agents can run on standard CRM data, but serious, knowledge-grounded production agents lean heavily on Data Cloud (Data 360). The Data 360 Starter SKU lists at around $60,000/year (10M credits, 5 TB storage), and costs grow from there on a consumption basis — additional credits at $500/100k, plus storage and add-ons — so real production deployments frequently reach six figures per year. It is usually the largest and most overlooked line item.

Yes. Salesforce Foundations is available to customers on Enterprise Edition or above and includes about 200,000 Flex Credits, 250,000 Data Cloud credits, plus Agent Builder and Prompt Builder at no cost. It is meant for building a proof of concept and measuring real action-per-request consumption before you choose a paid model or size Data Cloud. Treat the credits as a starting balance, not an operating budget.

If consumption pricing and the Data Cloud bill are deal-breakers, fixed-price options exist. GPTfy charges $20/$30/$50 per user per month with bring-your-own-model (Claude, GPT, Gemini, etc.) and no Data Cloud requirement. Einstein bundles predictive scoring into existing clouds, and Gong owns conversation/revenue intelligence — each solves a different job, so compare against your actual use case and volume.

More Salesforce + AI

Every session. Free. No registration.

Enterprise Dreamin' recordings cover Salesforce AI, data security, and enterprise architecture. Senior practitioners sharing what they actually learned.

Browse Sessions →