Salesforce AIforce

Where Salesforce AIforce sits

Salesforce's AI story in 2026 is powerful, expensive, and confusingly named — and a review that skips the naming has already lost the reader. The platform historically sold as Einstein, then reorganized around Agentforce agents, was relaunched as AIforce at Dreamforce 2026 (the announcement went out on September 16, 2026), and then, on October 9, 2026, Marc Benioff posted on X that it is now SIforce — "The era of Super Intelligence is here." As of my checks there was still no formal product announcement behind the rename, just the CEO's post and press coverage; Salesforce Ben reported days later that the rename covers the agent platform only. If you are evaluating this product under the AIforce name, note that you are evaluating a name the vendor has already retired.

Naming aside, the substance of AIforce is genuinely different from the Einstein-era feature list. The pitch is that Salesforce comes to the user instead of the user coming to Salesforce: employees ask questions, update records, and trigger workflows from inside Claude, Slack, or a generated interface, all grounded in CRM data and business logic. That is an architectural claim, not a marketing one — the delivery vehicle is a prebuilt MCP server inside Claude (Claudeforce), deep Slack integration (Slackforce), and a Coworker agent inside Salesforce's own Lightning surface. Piloted by Deloitte, GitLab, and Legora before the beta opened to all customers, it is the first Salesforce AI release in a while that changes where work happens rather than adding a panel to the same screens.

What it actually does

At the capability level, AIforce layers three things over the Salesforce platform. First, agent access anywhere: Claudeforce puts Salesforce intelligence inside Anthropic's Claude via MCP, so a user can read and update CRM records from a Claude conversation; Slackforce does the same inside Slack channels, where work happens in the open and context carries between people and agents. Second, Agentforce Coworker, the in-Salesforce agent that can read across hundreds of records, pull in connected systems, reason over them, and act. Third, composable interfaces: users describe a screen and get a live UI of exactly the fields and insights they need, rather than navigating fixed layouts and fields. Underneath it all sit the Agentforce building blocks — Agent Builder, pre-built templates and actions, customer- and employee-facing agents, and Agentforce Voice.

The pricing is consumption-shaped and, unusually for Salesforce, published. At time of writing, conversational pricing is $2 per conversation. The Flex Credits model charges $500 per 100,000 credits, with each Agentforce action (updating a record, summarizing a case, running a custom flow) consuming 20 credits — $0.10 per action — and Agentforce Voice actions consuming 30 credits. Agentforce User Licenses run $5 per user per month and require Flex Credits on top. Employee-facing add-ons for Sales, Service, and Field Service are $125 per user per month with unmetered employee agent usage; Industries add-ons are $150; Max Edition starts at $550 per user per month and bundles 2.75 million Flex Credits per org per year. Enterprise-edition-and-above customers can start with 100,000 Flex Credits free via Salesforce Foundations. And everything sits on a required Salesforce CRM foundation — Service Cloud Enterprise, Unlimited, Performance, or Einstein 1 Edition.

My testing notes

Observation one: the pricing model punishes vague architectures, and I worked the numbers to see how much. Take a 10,000-interaction-per-month support workload: at $2 per conversation that is $20,000 a month flat. Under Flex Credits, a conversation that resolves through five platform actions costs five times twenty credits — $0.50 — so the same workload runs about $5,000 a month. But an agent that summarizes, updates three records, and triggers a flow per turn can cross the conversational price quickly, and an action-heavy agent can cost several times a simple Q&A bot. My observation from the published rate card is that the two models are not interchangeable: one prices interactions, the other prices work done. Salesforce's own worked examples on the pricing page (a Voice reservation flow at 120 credits across 300 calls per month comes to $180) confirm the arithmetic.

Observation two: the friction of the surrounding package is the real cost. Every add-on sits on a CRM edition purchase, the useful consumption numbers appear only after a sales conversation ("Contact us" is the button on nearly every tier), and the Digital Wallet — the tool that tells you what agents are actually consuming — is part of the package rather than a free observability layer. In my reading of the docs and pricing pages, a mid-size team cannot self-serve a realistic estimate: the unit economics are published, but the mapping from your workflows to actions is yours to model.

Observation three: the naming churn has a measurable tax on research and enablement. Three names in one calendar year (Agentforce, AIforce, SIforce) means docs, blog posts, partner decks, and job listings refer to different things with overlapping vocabulary — and the SIforce rename followed a political news cycle rather than a product change, per Business Insider's reporting and the press coverage around it. When I searched for AIforce implementation guidance a week after the rename, the top results mixed the three names with four older Einstein-era product names. A buyer's internal training materials will age badly through no fault of the buyer.

Limits

What AIforce does not do is float free of the Salesforce stack. If you are not a Salesforce CRM customer, this product does not exist for you; if you are on a lower edition, the add-ons are unreachable without an upgrade. There is no meaningful self-serve trial of the agent platform — you evaluate it through a sales process and a pilot. And the pace of change is a risk in itself: beta surfaces (Claude integration was in beta at announcement), a rename with no formal announcement behind it, and a rate card introduced in May 2025 and extended since all point to a pricing and branding structure still in motion.

Adoption signals deserve a mention. Salesforce Ben's ecosystem coverage has included the observation that while some 85% of customers have tried Agentforce, demand has not picked up in line with the trials — I cannot audit that claim, and it is community reporting rather than a vendor disclosure, but it matches the general shape of what I see: lots of pilots, thin public evidence of scaled production use. Enterprise AI agent projects fail on data quality and process design more often than on model capability, and no CRM vendor's pricing page will tell you that.

Who should actually shortlist it

The realistic buyer is an organization with Salesforce already in the center of its revenue stack, a Service or Sales Cloud edition high enough to host the add-ons, and a workload with enough repetitive record-touching work that agent economics beat headcount. The strongest fits I can see are high-volume service deflection (where the $2-per-conversation model is predictable), internal operations agents that update records and run flows (where per-action pricing rewards tight workflows), and sales teams whose reps will genuinely ask Claude or Slack for CRM context instead of opening a dashboard. The weakest fits are teams that want AI without a Salesforce commitment, small deployments that cannot absorb a sales-led procurement cycle, and anyone whose workflows are too ambiguous to count actions — because if you cannot count the actions, you cannot forecast the bill, and consumption pricing without forecasting is how budgets get surprised.

Verdict

I rate Salesforce AIforce (SIforce) 3.5 out of 5. The rubric: core functionality depth is the highest in this review batch — agents grounded in real CRM data, acting across Slack, Claude, and Salesforce with published per-action economics — but value is conditional on workload shape and CRM commitment, usability is gated by a sales process and an edition ladder, and reliability signals are clouded by three rebrands in a year and mixed adoption reporting. For organizations already standardized on Salesforce with the budget for consumption pricing and the internal discipline to architect agent workflows around actions, this is the deepest AI layer available inside that ecosystem. For everyone else, the honest summary is that you are buying a platform relationship, not a tool — and the price of admission is the platform itself.

Sources

• https://www.salesforce.com/ap/news/press-releases/2026/09/16/ph-salesforce-unveils-aiforce-bringing-the-full-power-of-its-platform-to-any-interface
• https://www.salesforce.com/au/news/stories/aiforce-announcement/
• https://www.salesforce.com/service/ai/agentforce-for-service-pricing
• https://www.salesforce.com/au/service/ai/agentforce-for-service-pricing
• https://www.salesforce.com/in/news/press-releases/2025/05/15/agentforce-flexible-pricing-news/
• https://www.salesforceben.com/salesforce-renames-aiforce-to-siforce-after-trump-order
• https://www.salesforceben.com/salesforce-evp-addresses-ecosystem-comments-about-new-agentforce-pricing
• https://www.salesforceben.com/85-of-customers-have-tried-agentforce-but-demand-hasnt-picked-up-why/
• https://thenextweb.com/news/salesforce-aiforce-siforce-trump-super-intelligence
• https://timesofindia.indiatimes.com/technology/tech-news/salesforce-ceo-marc-benioff-renames-companys-biggest-product-hours-after-the-us-governments-enemy-post-adopts-label-that-white-house-wants-everyone-to-use/articleshow/134864245.cms
• https://www.chitizagarwal.com/blog/aiforce-siforce-what-changes