Salesforce Agentforce
Salesforce’s agentic AI platform — autonomous and assistive agents across Sales Cloud, Service Cloud, Marketing Cloud, Commerce Cloud and the Data Cloud, built on the Einstein Trust Layer and the Atlas Reasoning Engine — multi-model under the hood (Anthropic Claude inside the Salesforce VPC, OpenAI via Agentforce 360, Google Gemini, BYOLLM via Amazon Bedrock, Azure OpenAI and Google Vertex AI) and sold inside Salesforce’s CRM subscription stack into the existing 150,000+ customer base.
The Business
Salesforce Agentforce is Salesforce’s agentic AI platform — autonomous and assistive agents embedded across the Salesforce CRM workflow surfaces: Sales Cloud, Service Cloud, Marketing Cloud, Commerce Cloud and the Data Cloud. Agentforce launched in September 2024 at Dreamforce, was re-platformed onto the Atlas Reasoning Engine and re-bundled as Agentforce 360 at Dreamforce 2025, and is built on Salesforce’s long-running Einstein platform plus the Einstein Trust Layer for security, privacy and compliance grounding. The foundation-model stack is deliberately multi-supplier: Anthropic Claude is the preferred model inside the Salesforce VPC trust boundary (with a $300M 2026 token-spend commitment under the expanded Anthropic partnership), OpenAI is integrated via the Agentforce 360 surface inside ChatGPT, Google Gemini powers parts of the Atlas Reasoning Engine, and customers can BYOLLM via Amazon Bedrock, Azure OpenAI Service and Google Vertex AI per Salesforce’s developer documentation. The strategic positioning differs from frontier-model labs and from productivity-suite-anchored assistants: Agentforce is sold to enterprise sales, service, marketing and commerce buyers as a CRM-workflow-embedded agent layer inside the Salesforce subscription stack, distributed through Salesforce’s existing direct, partner and reseller account channels. Agentforce is a product family inside Salesforce Inc. — not a separate company or a standalone P&L — and the relevant financial frame is Salesforce subscription revenue with Agentforce ARR trajectory noted on earnings calls.
Customers and Distribution
Salesforce reported FY26 revenue of $41.5B (+10% YoY) with FY27 guidance of $45.8–46.2B, non-GAAP operating margin of 34.1%, and Agentforce-specific deal-mix disclosures of 29,000 Agentforce deals closed in Q4 FY26 (+50% QoQ), 12,000 customers on Agentforce 360 at the October 2025 launch growing to 18,500+ by Q4 FY26, and Agentforce ARR of $800M (+169% YoY) 18 months from the September 2024 launch. Salesforce’s top-10 Q4 wins all included Agentforce 360 plus Data 360 bundled with five Agentforce products. The Salesforce CRM install base is approximately 150,000+ enterprise customers globally, weighted to large-enterprise sales, service, marketing and commerce. Distribution sits across three motions inside Salesforce’s mature enterprise GTM: direct enterprise sales into the Salesforce account base (the principal channel, owned by Miguel Milano as Chief Revenue Officer and Brian Millham as President and COO), partnerships with the global systems-integrator ecosystem (Accenture, Deloitte, IBM, Capgemini and the broader Big-4 SI bench as the implementation muscle for Agentforce rollouts), and the AppExchange reseller and ISV partner network around Salesforce platform extensibility. The P3c GTM-maturity sub-rubric was held at 9 in the v1.6 evidence pass on the Q4 FY26 deal-mix evidence.
Model Strategy
Agentforce is a Verticals-first generative-AI play: Salesforce’s strategic bet is that workflow-embedded AI inside the Salesforce CRM and Data Cloud — paired with the 150,000+ customer install base no rival CRM platform can match — beats generalist assistant access inside regulated sales, service, marketing and commerce procurement. The foundation-model stack is deliberately multi-supplier: Anthropic Claude is preferred and runs inside the Salesforce VPC trust boundary under Trusted Context AI Actions (with a $300M 2026 token-spend commitment per the expanded Anthropic partnership), OpenAI is the Agentforce 360 surface inside ChatGPT, Google Gemini powers parts of the Atlas Reasoning Engine, and BYOLLM is supported via Amazon Bedrock, Azure OpenAI Service and Google Vertex AI per Salesforce’s developer documentation. The D4a supplier-diversity sub-rubric was held at 9 in the v1.6 evidence pass on that basis — the highest defensibility credit Agentforce receives. Above the foundation-model layer, the Atlas Reasoning Engine is the orchestration surface; the hybrid consumption-credit-plus-seat pricing and the Agentforce 360 bundle are the monetisation surface; the workflow-schema integration across Sales Cloud, Service Cloud, Marketing Cloud, Commerce Cloud and Data 360, plus the Einstein Trust Layer, are the structural differentiators. Output and agent-definition portability are constrained by the CRM and Data 360 schema (the D1c portability sub-rubric was held at 5 on that evidence) — high CRM upstream lock-in, moderate Agentforce-segment substitutability at the model layer.
Leadership Team
Agentforce is a product family inside Salesforce Inc. rather than a separate company — the IM Framework v1.5-A8 segment rule applies and Salesforce-segment leadership is in scope here. Bench depth is via Salesforce’s wider engineering, sales and customer-success organisation (~76,000 employees globally per the most recent 10-K disclosure); the D4e sub-rubric was held at 5 reflecting Marc Benioff’s continued personal load-bearing role on AI narrative against a credible second-tier in Adam Evans (EVP and GM, Salesforce AI) and Clara Shih (previously CEO of Salesforce AI; transitioned to lead a new Business AI initiative inside Meta in early 2025). Other publicly active senior figures include Brian Millham (President & COO, owns the field GTM motion) and Miguel Milano (President & Chief Revenue Officer, owns the global sales organisation that distributes Agentforce into the Salesforce account base).
Funding History
| Date | Round | Raised | Post-money | Lead investor(s) |
|---|---|---|---|---|
| n/a | Internal funding | n/a | n/a | Salesforce Inc. (NYSE: CRM) |
Salesforce Agentforce is not externally funded. It is a product family within Salesforce Inc., funded from parent cash flow. Salesforce reported FY26 revenue of $41.5B (+10% YoY) with FY27 guidance of $45.8–46.2B, non-GAAP operating margin of 34.1% and FCF positive. Agentforce itself reached $800M ARR at Q4 FY26 (+169% YoY), 18 months from the Sep 2024 launch, with 29,000 deals closed in Q4 FY26 alone (+50% QoQ) and 18,500+ customers on Agentforce 360 from 12,000 at the Oct 2025 launch. Salesforce committed approximately $300M to Anthropic in 2026 against the $300M token-spend partnership that anchors Claude inside the Salesforce VPC trust boundary; the investment is a capital-allocation transaction inside Salesforce Inc. rather than an external round. No external rounds exist for Agentforce as a standalone entity.
Competitive Landscape
Agentforce’s competitive set sits in three concentric rings: the symmetric workflow-platform plays that mirror Salesforce’s positioning on adjacent surfaces (ServiceNow Now Assist on ITSM, SAP Joule on ERP), the productivity-suite-anchored generative AI with cross-workflow extensibility (Microsoft 365 Copilot with Copilot Studio and Dynamics 365 Sales Copilot), and the pure-play CX-agent startups (Sierra, Decagon) that flank Service Cloud workflows specifically. Salesforce Agentforce is unusual in the set because the competitive frame is CRM-workflow depth at install-base scale rather than horizontal model capability — the strategic bet is that workflow-embedded AI inside the Salesforce CRM and Data Cloud beats generalist assistant access in regulated sales, service, marketing and commerce procurement.
| Competitor | Positioning | Distribution edge | Threat profile |
|---|---|---|---|
| Now Assist (ServiceNow (NYSE: NOW)) |
ServiceNow’s generative-AI assistant family across ITSM, CSM, HR Service Delivery, Field Service Management and Security Operations — the closest direct mirror of Agentforce’s workflow-platform-embedded positioning, but anchored in IT-operations rather than CRM. | ServiceNow enterprise account base (8,400+ customers, 630 above $5M ACV); direct enterprise sales; global systems-integrator partnerships (Accenture, Deloitte, EY, KPMG); Pro Plus and Enterprise Plus SKUs as the monetisation surface. | High — structurally symmetric play on a workflow platform of comparable enterprise scale; the principal head-to-head in enterprise procurement where customer-service and field-service workflows overlap. |
| Microsoft 365 Copilot (Microsoft (NASDAQ: MSFT)) |
Productivity-suite-embedded generative AI (Word, Excel, Outlook, Teams, PowerPoint) with Copilot Studio extensibility into line-of-business workflows including sales and service-adjacent surfaces. Dynamics 365 Sales Copilot is the more direct CRM-anchored counterpart inside the Microsoft estate. | Microsoft 365 enterprise install base (the world’s largest productivity-suite distribution channel); Azure account base; Copilot Studio low-code extensibility; Dynamics 365 CRM install base as the CRM-anchored channel. | High — Microsoft’s productivity-suite distribution reach materially exceeds Salesforce’s, and Dynamics 365 Sales Copilot is the structural CRM-anchored mirror; the open question is whether Copilot Studio extensions credibly substitute for CRM-workflow depth in regulated sales and service procurement. |
| Sierra | Conversational AI agent platform for customer experience — voice and chat agents that handle support, sales and operations across web, phone, SMS and app channels. Founder-led by Bret Taylor (ex-Salesforce co-CEO) and Clay Bavor (ex-Google Labs VP). | Direct enterprise sales driven by Bret Taylor and Clay Bavor’s senior-operator networks; growing partner ecosystem with system integrators; 40%+ of the Fortune 50 as paying customers. | Medium-High — pure-play CX-agent startup with Fortune 50 anchor-customer credibility and the Bret Taylor / ex-Salesforce co-CEO heritage as a credibility overlay; the principal head-to-head among CX-agent specialists where Service Cloud workflows overlap. |
| Decagon | AI customer-service agents for enterprise — vertical-CX pure-play with a similar valuation trajectory and Fortune 500 customer mix to Sierra, the second pure-play CX-agent flanking risk. | Direct enterprise sales; growing partner ecosystem; comparable Fortune 500 customer overlap including financial-services accounts. | Medium — structurally symmetric pure-play; flanks Service Cloud on customer-service-agent procurement but narrower than Agentforce on the full CRM-workflow surface. |
| Joule (SAP (NYSE: SAP)) |
SAP’s generative-AI assistant across S/4HANA, SuccessFactors, Ariba and the SAP Business Technology Platform — the ERP-anchored counterpart to Agentforce’s CRM anchor, with the SuccessFactors HR overlap as the principal CRM-adjacent surface. | SAP enterprise install base; BTP partner ecosystem; SuccessFactors HR distribution as the HR-workflow overlap. | Medium — ERP-anchored rather than CRM-anchored, but a credible workflow-platform competitor where finance, procurement and HR cases overlap with sales and service. |
Pricing benchmark: Agentforce is sold inside Salesforce’s CRM subscription stack on a hybrid model — consumption-based pricing on agent conversations (widely cited at $2 per conversation in launch coverage) layered onto seat-priced underlying CRM editions, plus the Agentforce 360 bundle introduced at Dreamforce 2025. Now Assist sits inside ServiceNow’s Pro Plus and Enterprise Plus SKUs on a per-user premium; Microsoft 365 Copilot and Dynamics Copilot are layered on parent-platform subscriptions on a comparable seat-plus-consumption structure; Sierra and Decagon are outcome-based per-resolution. The competitive frame is therefore enterprise procurement and workflow depth rather than headline per-conversation price; Salesforce’s structural differentiator is CRM-workflow schema lock-in and the 150,000+ customer install base.
Potential Risks
The case for Salesforce Agentforce at IM Framework 7.16 rests on the Salesforce CRM install base (150,000+ customers globally), the multi-model supplier-diversity position (Anthropic preferred with the $300M 2026 commitment, OpenAI, Google Gemini and BYOLLM across the three major hyperscalers), Salesforce’s self-funded balance sheet at FY26 revenue of $41.5B (+10% YoY) and FY27 guidance of $45.8–46.2B, the 29,000 Q4 FY26 Agentforce deal count (+50% QoQ), and the Agentforce ARR of $800M (+169% YoY) 18 months from launch. The case against splits into five risks of differing magnitude — with the pricing-model uncertainty the most active short-term, the rollout-pace risk against the post-launch trajectory the most structural, and the segment-of-conglomerate framing the most methodologically deliberate constraint.
Rollout pace against the $800M ARR trajectory and the Oct 2025 reset
Agentforce launched in Sep 2024 with the original “AI agents that act, not just assist” framing; the initial seat-plus-consumption pricing model drew mixed reception from the channel and the customer base, and Salesforce reset the GTM at Dreamforce 2025 with the Agentforce 360 launch and a re-platforming onto the Atlas Reasoning Engine plus multi-model routing. Agentforce ARR of $800M at Q4 FY26 (+169% YoY) and 29,000 Q4 deals (+50% QoQ) suggest the reset is working — but the trajectory is still 18 months in, against a parent revenue base of $41.5B, and the FY27 guide of $45.8–46.2B implies Agentforce ARR needs to compound aggressively for the segment to break out of “Emerging Player” framing. The P1d time-to-revenue stage-appropriateness sub-rubric was held at 7 on that basis — stage-appropriate for an incremental segment on a $41.5B parent, but not break-the-curve given the owned-channel-into-buyer distribution advantage.
Pricing-model uncertainty — consumption credits, seats and outcome-based hybrids
The Agentforce monetisation model has evolved through three phases inside 18 months: per-conversation consumption credits at launch (widely cited at $2 per conversation in the Sep 2024 press cycle), the Agentforce 360 bundle at Dreamforce 2025 with Foundations and Pro Plus framing, and ongoing commentary on outcome-based pricing experiments where the agent only bills against resolved cases. The bull case is that hybrid pricing lets Salesforce match each customer’s procurement preference and accelerate adoption; the bear case is that pricing-model instability slows the ARR trajectory, complicates the Salesforce reseller motion, and gives outcome-based competitors (Sierra, Decagon) a clearer GTM story in Fortune 500 procurement. The pricing-model evolution is a material watch-item for the next FY27 earnings cycle.
Portability and substitutability — CRM and Data 360 schema lock-in
Agentforce is bundled on top of Salesforce CRM and Data 360 — cannot churn from Agentforce without churning the underlying CRM and Data Cloud, which is the same workflow-platform lock-in that has anchored Salesforce’s subscription gross retention since IPO. The model layer is portable in principle (BYOLLM via Amazon Bedrock, Azure OpenAI Service, OpenAI and Google Vertex AI per Salesforce’s developer documentation), but the agent definitions, workflow triggers, Atlas Reasoning Engine grounding and Einstein Trust Layer integrations are not portable to a rival CRM. The D1c portability sub-rubric was held at 5 in the v1.6 evidence pass on that basis — high CRM upstream lock-in, but moderate Agentforce-segment substitutability at the model layer. The bull case is that this is the same lock-in that has defined CRM subscription gross retention for two decades — a feature, not a bug. The bear case is that Microsoft 365 Copilot with Dynamics 365 and Copilot Studio extensions, plus pure-play CX agents on customer-managed data warehouses, give some customers a credible exit from the CRM lock-in if Agentforce capability slips.
Foundation-model supplier interlock — Anthropic, OpenAI and Google all in scope
Agentforce runs Anthropic Claude as the preferred model inside the Salesforce VPC trust boundary (with the $300M 2026 token-spend commitment per Salesforce’s expanded Anthropic partnership), OpenAI via the Agentforce 360 surface inside ChatGPT, Google Gemini via the Atlas Reasoning Engine and BYOLLM across Amazon Bedrock, Azure OpenAI Service and Google Vertex AI. The D4a supplier-diversity sub-rubric was held at 9 in the v1.6 evidence pass on that basis — the highest defensibility credit Agentforce gets — but the structural exposure is that each supplier is also a potential competitor for the same enterprise inference workloads, and Anthropic, OpenAI and Google all ship direct enterprise APIs. None of this is fatal to Agentforce, but it caps how durable the multi-model story is if any one of those labs prices the partnership economics down or competes more aggressively for the same enterprise budget.
Regulatory exposure — EU AI Act, GDPR, HIPAA and the multi-vertical compliance load
Agentforce sits inside a regulatory perimeter that includes the EU AI Act general-purpose-AI obligations (multi-model Agentforce stack in scope), GDPR data-residency requirements served through Hyperforce local zones in EU, Australia, Japan, India, UAE and Brazil, HIPAA with Business Associate Agreement for healthcare customer workflows, and SOC 2 Type II plus the EU Cloud Code of Conduct certification for Agentforce on Hyperforce. Salesforce inherited the Salesforce Trust certifications and the Einstein Trust Layer architecture for Agentforce. The D4c regulatory-exposure sub-rubric was upgraded from 6 to 7 in the v1.6 evidence pass on the second-level EU AI Act compliance evidence. The bull case is that Salesforce’s regulated-industry compliance discipline forces a procurement advantage that smaller pure-play competitors cannot match; the bear case is that regulated-industry rollouts in financial services, healthcare and life sciences are slower-cycle than commercial-CRM deployments, capping the near-term ARR trajectory inside the regulated cohort.
Recent IM Coverage
Show recent press coverage of Salesforce Agentforce
- Feb 2026 — Salesforce delivers record fourth quarter fiscal 2026 results — revenue $9.99B (+8% YoY); FY26 revenue $41.5B (+10% YoY); FY27 guidance $45.8–46.2B; Agentforce ARR $800M (+169% YoY).
- Feb 2026 — Salesforce Q4 FY26 press release: Agentforce deal count, ARR trajectory and FY27 outlook.
- 2026 — Marc Benioff FY26 summary — Agentforce trajectory, 29,000 Q4 deals and the agentic-enterprise framing.
- 2026 — Salesforce avoids Q3 danger zone with explosive Agentforce momentum — 12,000 customers on Agentforce 360 launch growing to 18,500+; top-10 Q4 wins all included Agentforce 360 plus Data 360.
- 2026 — Salesforce Q3 FY26 analysis: AI agents and Data 360 lift bookings and FY26 outlook.
- 2026 — Anthropic and Salesforce expanded partnership — Claude inside Salesforce VPC trust boundary, Trusted Context AI Actions on Claude, $300M 2026 token-spend commitment.
- 2026 — Salesforce News — Trusted Context AI Actions on Claude for Agentforce.
- 2026 — Salesforce News — Agentforce achieves EU Cloud Code of Conduct GDPR-compliance certification on Hyperforce.
Curated feed of named-source coverage — Salesforce Investor Relations, the Salesforce Newsroom, Anthropic’s own partnership disclosures, and named-author specialist coverage (Salesforce Ben, Futurum Group, Parameter). Excludes paywalled article bodies of The Information, WSJ, FT and Bloomberg (headline plus free-snippet only), wire-aggregator reposts and unsourced AI round-up pieces.
Show the source register for the figures on this page
IM operates a primary-source-where-possible discipline. The figures above come from:
- Revenue (basis-disclosure note): Salesforce does not separately disclose Agentforce revenue on a clean GAAP basis. Closest published references: Salesforce FY26 Q4 results — FY26 revenue $41.5B (+10% YoY), FY27 guidance $45.8–46.2B, non-GAAP operating margin 34.1%; Agentforce ARR $800M (+169% YoY) at Q4 FY26 per Marc Benioff’s LinkedIn FY26 summary and the Salesforce Q4 FY26 press release. We label the headline as “Salesforce subscription revenue with Agentforce ARR trajectory noted” rather than a clean Agentforce-only GAAP figure and decline-to-publish a precise stand-alone Agentforce revenue number pending a primary segment disclosure.
- Customers and deal-mix: Salesforce disclosed 29,000 Agentforce deals closed in Q4 FY26 (+50% QoQ), 12,000 customers on Agentforce 360 at the Oct 2025 launch growing to 18,500+ by Q4 FY26, and top-10 Q4 wins all including Agentforce 360 plus Data 360 bundled with five Agentforce products per Salesforce Ben’s Q3 coverage and Futurum’s Q3 FY26 analysis. The 150,000+ customer install base is the standing reference for the Salesforce CRM account base into which Agentforce is sold per Salesforce’s investor materials.
- Model supplier mix and compliance: Multi-model routing across Anthropic Claude (preferred, inside Salesforce VPC trust boundary, with a $300M 2026 token-spend commitment per the Anthropic and Salesforce expanded partnership and Salesforce News on Trusted Context AI Actions on Claude), OpenAI (Agentforce 360 inside ChatGPT), Google Gemini (Atlas Reasoning Engine) and BYOLLM via Amazon Bedrock, Azure OpenAI and Google Vertex AI per Salesforce Developer documentation on supported models. SOC 2 Type II, HIPAA with BAA, GDPR with data residency and EU AI Act phased-compliance per Salesforce News on EU Cloud Code of Conduct.
- Funding to date: Not applicable. Salesforce Agentforce is funded internally from Salesforce Inc.’s balance sheet. Reference: Salesforce FY26 revenue $41.5B (+10% YoY) with FY27 guidance of $45.8–46.2B and non-GAAP operating margin of 34.1%. Salesforce Inc. is listed on NYSE as CRM and funds product development including Agentforce from operating cash flow rather than external rounds.
Methodology & Disclaimer
For metric definitions, source-tier hierarchy, and decline-to-publish rules, see the tracker methodology. Confidence dots (• green / • amber / • red) follow the same convention as the AI Tracker.
Spotted a figure you believe is wrong? Send corrections to info@informationmatters.net.
Information Matters Framework scores are the considered opinion of the IM team — human and AI — applied to publicly-available evidence under a disclosed methodology. They are not statements of fact about the companies scored and they are not investment advice.
