Salesforce’s Agentforce 360 Aims to Help Companies Escape ‘AI Pilot Purgatory’

Welcome to the era of the agentic enterprise.

At this year’s Dreamforce event, Salesforce is unveiling Agentforce 360, a major upgrade to its agentic development platform. The company says it brings together humans, AI agents, apps, and data to help organizations forge stronger connections with customers, employees, and suppliers.

“What we want to see is all companies becoming agentic enterprises…we’re all trying to have more employees and more labor in the market, and yet, we can’t hire enough people—you can’t find them, or maybe you don’t have enough money to hire them,” Parker Harris, Salesforce’s co-founder and Slack’s chief technology officer, remarked in a press conference last week. “But what we’re entering this world is that you can hire AI, that in the future, 40 percent of the work in the Fortune 1000 is probably going to be done by AI, and it’s going to be humans and AI actually working together. It’s going to be an incredible future, and we’re helping bring all of our customers forward into that future.”

Salesforce has moved expeditiously to innovate on Agentforce since its debut last year at Dreamforce. The AI platform, according to chief executive Marc Benioff, was “the biggest, most exciting piece of technology” the company has ever worked on. Agentforce’s next evolution, just a few months later, introduced autonomous agents to Slack, followed by another update that transformed its bots into “proactive” ones. Then, with Agentforce 3’s launch in June, it ushered in what Salesforce believed was the rise of the digital labor force.

Agentforce 360 is perhaps the culmination of what Benioff once proclaimed as the “third age of AI.”

Disclosure: I’m attending Dreamforce as a guest of Salesforce. The company has paid for most of my expenses. However, Salesforce in no way dictated the content of this post. These are my words.

Eliminating ‘Pilot Purgatory’

Despite Salesforce’s bold marketing and lofty promises, Agentforce adoption has lagged. Critics argue that the platform’s features are still immature and have yet to live up to the hype. In addition, the company faces pressure to respond to industry headwinds as organizations struggle to see returns from their enterprise AI pilot programs. Salesforce, like others, cites the much-debated MIT study, which claimed 95 percent of AI pilots fail, but says Agentforce 360 can reverse that trend.

“It’s not because of a lack of intent. People want to [use AI],” Srini Tallapragada, Salesforce’s president and chief engineering officer, points out. “Everybody understands the power of the technology, but why is it so hard? Because…when you are using ChatGPT, the first three days are cool, but are you using it every second? No, because that’s not in the line of your work.”

“It has to be embedded, not an add-on,” he contends. “You can’t go somewhere else. It has to be in the line of flow of your work. Similarly, if you are an enterprise customer in a regulated industry, you need a lot of governance and trust…Customers want to do it. They’re stuck in what I call a ‘pilot purgatory,’ and this is the problem we are trying to solve.”

Having the right AI model isn’t enough. Enterprise AI programs require deep context to provide accurate answers. This is why Salesforce identifies five pillars for creating an agentic enterprise: a development platform, a hub that stores all structured and unstructured data, an application that contains business logic and workflows, and a conversational interface. To support that vision, Salesforce is positioning five key products as the foundation: Agentforce 360, Data Cloud 360, Customer 360, and Slack.

What’s Included In Agentforce 360?

Agentforce 360 is more than the unification of humans, agents, apps, and data. The platform comes with its own set of features, which Salesforce boasts are designed to “deliver controlled, contextual, and consistent experiences” across all interactions. It promises to move AI from prototype to production-ready, scalable bots at a much faster pace than ever before.

Admittedly, it could be tricky to understand all of Salesforce’s AI moves as of late. The company is moving so quickly that it’s announcing new Agentforce use cases left and right. Along with the agentic platform, the technology was embedded into Tableau and launched versions for the U.S. government, ITSM, and human resources. How does Agentforce 360 fit into everything then? Tallapragada suggests thinking of it less like a product and more like a rethinking of how we do business tasks:

“Agentforce 360 is the evolution. It’s a functional redefinition of how we get jobs done…Because we have this embedded knowledge of deep jobs-to-be-done workflows, we encoded that into Agentforce 360 so that when companies want to become an agentic 360 enterprise, they are not starting at ground zero. They’re starting, if it’s a 100-meter race, at 90 meters. That’s what the power is, and we’ll continue to focus on simplifying that experience, working very closely with customers, and ensuring that they get fast time to value.”

The emphasis on Slack is notable, though Salesforce has plans to extend Agentforce 360 to competing team messaging apps. “We are doing integrations with [Microsoft] Teams,” Harris admits. “We find most of our customers use it, certainly for video conferencing. They’ll use it for some direct messaging. If you’re in Slack, maybe you want to join that Teams meeting—we’re going to make that super easy. So we are doing some integrations, but people aren’t really doing their work in a product like Teams. And when you think about Slack, that’s where we want you to get your work done.”

As the name suggests, Agentforce 360 consolidates all information and context across the Salesforce landscape, including third-party apps tied to the CRM. It’s the only enterprise AI platform that provides organizations with a complete view. Not only does it unite all the various Agentforce offerings, but it also incorporates Slack, Heroku, Salesforce’s Trailhead, MuleSoft, and Informatica.

Here are some of the key features of the new platform:

Agent Scripting and Improved Atlas Engine

This capability gives organizations control over an AI agent’s behavior. Developers provide a “script” in a readable, structured JSON format that instructs the bot on how to act, what steps it should follow, and under what conditions. The agent will then follow those instructions while also balancing its own reasoning and flexibility, powered by Salesforce’s Atlas Reasoning Engine. The tech firm notes that models are limited to interpreting human language on their own, but most tasks have more sequential steps that require strict adherence. Developers can use the new Agent Script feature to create “if/then” conditionals, hand-offs, and tool calls, turning reasoning patterns into clear, governed code.

And speaking of Atlas, Salesforce is giving it an upgrade. Not only can the reasoning engine be configured to better balance an LLM’s creativity with structured business logic, but it now supports Google Gemini. This is the third AI model that the Atlas Reasoning Engine uses to reason and plan, joining OpenAI and Anthropic.

Agent Script is expected to launch in beta in November 2025.

Agentforce Voice

This feature enables developers to create agents that utilize the power of speech. Interestingly, it puts Salesforce more in competition not only with Sierra, the AI startup founded by Benioff’s former co-CEO Bret Taylor, but also with Zendesk, Dialpad, Twilio, Google Cloud, Talkdesk, Genesys, and Five9.

“Traditional phone systems weren’t built for the age of AI and unified data,” Abby Todd, Salesforce’s director of product marketing for Service Cloud, writes in a post. “They live outside your CRM. They can’t access knowledge articles or case history. And when a customer moves from chat to phone‌—or vice versa—the conversation starts over from scratch.”

A source of frustration for customers is the interactive voice response (IVR) system they’re forced to deal with when calling for support. It’s estimated that 35 percent of all these calls are actually resolved, but when an AI agent is involved, that number rises to 50 percent. Robin Gareiss, the chief executive of research firm Metrigy, states, “By 2030, most [IVRs] will be replaced by AI triage agents…36.7 percent of CX leaders say they’re replacing IVRs with AI agents, and 62.5 percent of our research success group—those generating the highest business metric improvements from AI—are doing the same.”

According to Salesforce, Agentforce Voice powers natural, human-like conversations that match the organization’s tone and personality. It was developed by Salesforce’s AI Research team using the new eVerse framework and has been thoroughly tested against most scenarios, including noisy backgrounds and distracted callers. Agents powered by this tool can deliver real-time audio streaming and low-latency transcription. It also takes into account interruptions, phonetic dictionaries, and exercises patience to avoid rushing customers.

Agentforce Voice can also be integrated into an organization’s existing contact center infrastructure, including NiCE, Genesys, Amazon Connect, Five9, and Vonage.

Salesforce reveals that Agentforce Voice will be generally available on October 21. However, it’s also working on two additional voice offerings, one involving Session Initiation Protocol (SIP) and another for WebRTC and digital channels. Pilot programs for both are set to begin this December.

Agentforce Builder

Salesforce doesn’t want anything to slow workers down from building agents, which is why it’s launching Agentforce Builder, a low-code conversational development environment made for everyone. With the aid of an AI-powered “vibe building” tool, users can describe the agent they want built in plain language, and Agentforce will handle the logic, topics, instructions, and more using Agent Script. In addition, the workspace canvas supports inline editing and suggestions, and also enables quick simulation generation with real-time debugging.

Every agent created will have verifiable metadata, meaning users can see who authored, tested, and deployed the bot. All agents can also be exported, shared, or extended through APIs and JSON representations.

Along with Agentforce Builder, Salesforce is also formally introducing Agentforce Vibes, its vibe coding platform. News about the company’s GitHub Copilot, Vercel, Lovable, and Claude Code rival first emerged earlier this month. While Agentforce Vibes may be similar to Agentforce Builder, programmatic developers may rely more heavily on it to help automate complex development work.

Developers can try the new Agentforce Builder in beta in November 2025.

Context Indexing

The final piece of Agentforce 360 is context indexing, which ensures AI agents are not just grounded in structured data, but also in an organization’s unstructured data (e.g., PDFs, tables, images, and flowcharts). Salesforce’s new Data 360 platform powers this feature, which automatically extracts and formats the information to help agents deliver accurate and relevant answers in that business-level context.

Context Indexing will be generally available later this month.

Virtual accounting firm 1-800Accountant reports that Agentforce powered a 90 percent case deflection rate, with AI agents resolving 70 percent of customer chats during the busy tax week. “By becoming an agentic enterprise and expanding our use of Agentforce, we’ll soon be giving customers real-time updates on their tax returns, automatically flagging potential savings for them, and giving our pros pre-built agendas. This intelligent automation makes sure our team is focused on the complex needs of our customers, which is how we truly keep our clients successful,” Chief Technology Officer Ryan Teeples says in a statement.

However, Agentforce 360 isn’t the only announcement Salesforce is making at Dreamforce this year. The company is also unveiling the Slackification of its platform, bringing some of its CRM capabilities natively into its team communication app for the first time. It’s part of a growing trend in which Salesforce views Slack as becoming what it believes is the agentic operating system for the enterprise.

Featured Image: Salesforce co-founders Marc Benioff and Parker Harris stand side-by-side at Dreamforce 2024. Credit: Salesforce