Agentforce

Learning Objectives

• Understand how Agentforce deploys autonomous AI agents within the Salesforce CRM ecosystem
• Identify the key agent types and their use cases in sales, service, and marketing
• Evaluate when Agentforce is the right choice vs. general-purpose agent frameworks

What Is Agentforce?

Agentforce is Salesforce's AI agent platform — autonomous agents that handle customer service tickets, qualify inbound leads, schedule meetings, and manage cases end-to-end for routine requests. By early 2026, approximately 12,000 customers were live on Agentforce, with annualized revenue exceeding $500 million.

Unlike most AI agent products that require developers to configure behavior in code, Agentforce offers Agent Studio: build agents through natural language and point-and-click configuration, without code. Agents have native access to CRM data — customer history, support tickets, contracts, product usage — making their responses more contextually accurate than general-purpose AI tools.

Pricing & Access

Core Agent Types

SDR Agent (Sales Development)

Autonomous AI that qualifies inbound leads, sends personalized outreach, handles objections from a trained knowledge base, and schedules meetings — before a human sales rep is involved. This changes the capacity calculation for sales development entirely: teams can handle significantly more inbound volume without proportional headcount increases.

Service Agent

Handles customer service tickets end-to-end for routine requests — order status, returns, account changes, basic troubleshooting. Escalates to human agents only when the issue requires judgment or falls outside its training. The agent uses CRM history to personalize responses.

Sales Coach

Provides real-time coaching during sales conversations — surfacing relevant product information, competitive intelligence, and next-best-action recommendations based on the deal context in the CRM.

The Informatica Advantage

Salesforce's $8 billion acquisition of Informatica (completed November 2025) strengthens the data foundation that Agentforce agents rely on. Informatica brings enterprise data catalog, integration, governance, and master data management — solving the biggest bottleneck in enterprise AI: getting clean, connected data to the models.

The Fin Acquisition

In 2026, Salesforce agreed to acquire Fin — the autonomous customer-service agent from the company formerly known as Intercom — for approximately $3.6 billion, with the deal expected to close in early 2027. Fin resolves customer questions end-to-end across chat, email, WhatsApp, SMS, phone, and Slack, and Salesforce plans to fold its technology and team into Agentforce. It is one of the largest acquisitions yet of a dedicated customer-service AI agent, and it directly deepens Agentforce's Service Agent capabilities.

Claudeforce and the Anthropic Partnership

On August 26, 2026 Salesforce and Anthropic announced Claudeforce, an expanded partnership that runs in both directions and materially changes where Agentforce work happens.

Inside Salesforce, Claude serves as a reasoning model for the Atlas Reasoning Engine, powers Agentforce Vibes and Agentforce Coworker by default, and is selectable in Agent Builder. Through Amazon Bedrock it runs inside the Salesforce Trust Boundary, so regulated-industry customers can deploy it without data leaving that environment.

The more striking half points outward. Salesforce in Claude is a plugin carrying 37 prebuilt sales skills — meeting prep, deal health review, pipeline review — that lets a seller work the revenue cycle from inside Claude rather than from the Salesforce interface. Actions still route through Salesforce so business rules are enforced, and an administrator connects it once for the whole team. Salesforce also made Claude the default model across Slack, including Slackbot.

Marc Benioff summarised the design as "the UI is the AI". Read plainly, that is a company whose product has always been its interface deciding that the interface is now somebody else's chat window — a response to investor pressure on whether assistants will displace business software rather than merely assist it.

Koa — Salesforce's First In-House Reasoning Model

On September 15, 2026 Salesforce announced Koa, its first CRM reasoning model, built by post-training NVIDIA's Nemotron 3 Super. Until Koa, multi-step reasoning inside Agentforce was routed out to a frontier model — Claude or ChatGPT — through the platform's AI gateway. Koa gives Salesforce an in-house option for that work.

The training corpus contains no customer data. Salesforce and NVIDIA generated synthetic scenarios modelled on enterprise workflows across more than 14 industries, pairing a persona with tasks and mapping the sequence of tool calls an agent must make to complete them. Salesforce holds the weights and runs post-training and inference inside its own trust boundary, so no customer data crosses that boundary at inference time either. On Salesforce's own CRM benchmark the company reports Koa matching or exceeding leading model performance on CRM actions with three times fewer errors — a vendor-run benchmark on a vendor-defined task set, so read it as a statement of design intent rather than an independent result.

Two things this does not mean. Koa does not replace Claudeforce: Salesforce announced the Anthropic partnership three weeks earlier and positions Koa as an alternative within Agentforce rather than a substitute for frontier models. And it is not generally available — Koa reached select pilot customers at announcement, among them 1-800Accountant, Baxter Credit Union, Engine, Formula 1, UChicago Medicine and Xero, with general availability expected in winter 2026 in US regions.

Strengths

• Native CRM data access: Agents operate within the world's most widely used CRM — no data integration required
• Frontier reasoning without leaving the trust boundary: Claude runs via Amazon Bedrock inside Salesforce's own environment, which is what makes it usable in regulated industries
• No-code Agent Studio: Business users build agents without developers
• Scale proven: 12,000+ customers, $500 million+ ARR demonstrates enterprise-grade reliability
• Salesforce ecosystem: Agents integrate with the full Salesforce suite (Sales Cloud, Service Cloud, Marketing Cloud, Data Cloud)

Limitations & Considerations

• Salesforce-only: Agents are deeply tied to the Salesforce ecosystem — not portable to other CRMs
• Vendor lock-in: Embedding Agentforce deeply into operations increases switching costs
• Pricing complexity: Per-conversation pricing can be unpredictable for high-volume use cases
• CRM scope: Best for sales, service, and marketing — not a general-purpose agent framework

Key Takeaways

• Agentforce is Salesforce's autonomous AI agent platform with 12,000+ customers and $500 million+ annualized revenue
• Agent Studio enables no-code agent building through natural language configuration
• Native CRM data access is the core differentiator — agents know customer history without integration work
• Best for organizations already on Salesforce who want AI agents for sales, service, and marketing workflows
• Salesforce's ~$3.6 billion acquisition of Fin (formerly Intercom) adds a proven customer-service agent to the Agentforce lineup