As 50,000 attendees descend on Salesforce's Dreamforce conference this week, the enterprise software program big is making its most aggressive guess but on synthetic intelligence brokers, positioning itself because the antidote to what it calls an industry-wide "pilot purgatory" the place 95% of enterprise AI projects never reach production.
The corporate on Monday launched Agentforce 360, a sweeping reimagination of its total product portfolio designed to rework companies into what it calls "agentic enterprises" — organizations the place AI brokers work alongside people to deal with as much as 40% of labor throughout gross sales, service, advertising, and operations.
"We’re really within the agentic AI period, and I believe it's in all probability the largest revolution, the largest transition in expertise I've ever skilled in my profession," mentioned Parker Harris, Salesforce's co-founder and chief expertise officer, throughout a latest press briefing. "Sooner or later, 40% of the work within the Fortune 1000 might be going to be carried out by AI, and it's going to be people and AI really working collectively."
The announcement comes at a pivotal second for Salesforce, which has deployed greater than 12,000 AI agent implementations over the previous yr whereas constructing what Harris referred to as a "$7 billion enterprise" round its AI platform. But the launch additionally arrives amid uncommon turbulence, as CEO Marc Benioff faces fierce backlash for latest feedback supporting President Trump and suggesting Nationwide Guard troops ought to patrol San Francisco streets.
Why 95% of enterprise AI initiatives by no means launch
The stakes are huge. Whereas corporations have rushed to experiment with AI following ChatGPT's emergence two years in the past, most enterprise deployments have stalled earlier than reaching manufacturing, in line with recent MIT research that Salesforce executives cited extensively.
"Prospects have invested loads in AI, however they're not getting the worth," mentioned Srini Tallapragada, Salesforce's president and chief engineering and buyer success officer. "95% of enterprise AI pilots fail earlier than manufacturing. It's not due to lack of intent. Individuals need to do that. All people understands the facility of the expertise. However why is it so exhausting?"
The reply, in line with Tallapragada, is that AI instruments stay disconnected from enterprise workflows, knowledge, and governance programs. "You're writing prompts, prompts, you're getting annoyed as a result of the context is just not there," he mentioned, describing what he referred to as a "immediate doom loop."
Salesforce's answer is a deeply built-in platform connecting what it calls 4 components: the Agentforce 360 agent platform, Information 360 for unified knowledge entry, Buyer 360 apps containing enterprise logic, and Slack because the "conversational interface" the place people and brokers collaborate.
Slack turns into the entrance door to Salesforce
Maybe probably the most important strategic shift is the elevation of Slack — acquired by Salesforce in 2019 for $27.7 billion — as the first interface for Salesforce itself. The corporate is successfully reimagining its conventional Lightning interface round Slack channels, the place gross sales offers, service circumstances, and knowledge insights will floor conversationally slightly than by kinds and dashboards.
"Think about that you simply possibly don't log into Salesforce, you don't see Salesforce, but it surely's there. It's coming to you in Slack, as a result of that's the place you're getting your work carried out," Harris defined.
The technique consists of embedding Salesforce's Agentforce brokers for gross sales, IT service, HR service, and analytics straight into Slack, alongside a totally rebuilt Slackbot that acts as a private AI companion. The corporate can be launching "Channel Expert," an always-on agent that gives instantaneous solutions from channel conversations.
To allow third-party AI instruments to entry Slack's conversational knowledge, Salesforce is releasing a Real-Time Search API and Model Context Protocol server. Companions together with OpenAI, Anthropic, Google, Perplexity, Author, Dropbox, Notion, and Cursor are constructing brokers that can stay natively in Slack.
"One of the best ways to see the facility of the platform is thru the AI apps and brokers already being constructed," Rob Seaman, a Salesforce govt, mentioned throughout a technical briefing, citing examples of startups "attaining tens of 1000’s of consumers which have it put in in 120 days or much less."
Voice and IT service take goal at new markets
Past Slack integration, Salesforce introduced main expansions into voice-based interactions and worker service. Agentforce Voice, now typically obtainable, transforms conventional IVR programs into pure conversations that may replace CRM information, set off workflows, and seamlessly hand off to human brokers.
The IT Service providing represents Salesforce's most direct problem to ServiceNow, the market chief. Mudhu Sudhakar, who joined Salesforce two months in the past as senior vp for IT and HR Service, positioned the product as a elementary reimagining of worker assist.
"Legacy IT service administration could be very portals, kinds, tickets centered, handbook course of," Sudhakar mentioned. "What we had just a few key tenets: dialog first and agent first, actually centered on having a conversational expertise for the individuals requesting the assist and for the individuals offering the assist."
The IT Service platform consists of what Salesforce describes as 25+ specialised brokers and 100+ pre-built workflows and connectors that may deal with every little thing from password resets to complicated incident administration.
Early clients report dramatic effectivity beneficial properties
Buyer outcomes counsel the strategy is gaining traction. Reddit decreased common assist decision time from 8.9 minutes to 1.4 minutes — an 84% enchancment — whereas deflecting 46% of circumstances solely to AI brokers. "This effectivity has allowed us to supply on-demand assist for complicated duties and enhance advertiser satisfaction scores by 20%," mentioned John Thompson, Reddit's VP of gross sales technique and operations, in a press release.
Engine, a journey administration firm, decreased common deal with time by 15%, saving over $2 million yearly. OpenTable resolved 70% of restaurant and diner inquiries autonomously. And 1-800Accountant achieved a 90% case deflection charge through the essential tax week interval.
Salesforce's personal inner deployments could also be most telling. Tallapragada's buyer success group now handles 1.8 million AI-powered conversations weekly, with metrics revealed at help.salesforce.com displaying what number of brokers reply versus escalating to people.
Much more considerably, Salesforce has deployed AI-powered gross sales growth representatives to comply with up on leads that might beforehand have gone uncontacted as a result of price constraints. "Now, Agentforce has an SDR which is doing 1000’s of leads following up," Tallapragada defined. The corporate additionally elevated proactive buyer outreach by 40% by shifting employees from reactive assist.
The belief layer downside enterprises can't ignore
Given enterprise issues about AI reliability, Salesforce has invested closely in what it calls the "belief layer" — audit trails, compliance checks, and observability instruments that allow organizations monitor agent conduct at scale.
"You must consider an agent as a human. Digital labor. It’s essential handle efficiency similar to a human. And also you want these audit trails," Tallapragada defined.
The corporate encountered this problem firsthand when its personal agent deployment scaled. "Once we began at Agentforce at Salesforce, we’d observe each message, which is nice till 1,000, 3,000," Tallapragada mentioned. "After getting 1,000,000 chats, there's no human, we can’t do it."
The platform now consists of "Agentforce Grid" for looking out throughout tens of millions of conversations to establish and repair problematic patterns. The corporate additionally launched Agent Script, a brand new scripting language that permits builders to outline exact guardrails and deterministic controls for agent conduct.
Information infrastructure will get a serious improve
Underlying the agent capabilities is important infrastructure funding. Salesforce's Data 360 consists of "Clever Context," which mechanically extracts structured data from unstructured content material like PDFs, diagrams, and flowcharts utilizing what the corporate describes as "AI-powered unstructured knowledge pipelines."
The corporate can be collaborating with Databricks, dbt Labs, and Snowflake on the "Common Semantic Interchange," an try and standardize how totally different platforms outline enterprise metrics. The pending $8 billion acquisition of Informatica, anticipated to shut quickly, will broaden metadata administration capabilities throughout the enterprise.
The aggressive panorama retains intensifying
Salesforce's aggressive AI agent push comes as just about each main enterprise software program vendor pursues comparable methods. Microsoft has embedded Copilot throughout its product line, Google presents agent capabilities by Vertex AI and Gemini, and ServiceNow has launched its personal agentic choices.
When requested how Salesforce's announcement in comparison with OpenAI's recent releases, Tallapragada emphasised that clients will use a number of AI instruments concurrently. "More often than not I'm seeing they're utilizing OpenAI, they're utilizing Gemini, they're utilizing Anthropic, similar to Salesforce, we use all three," he mentioned.
The actual differentiation, executives argued, lies not within the AI fashions however within the integration with enterprise processes and knowledge. Harris framed the competitors in phrases acquainted from Salesforce's founding: "26 years in the past, we simply mentioned, let's make Salesforce automation as straightforward as shopping for a e book on Amazon.com. We're doing that very same factor. We need to make agentic AI as straightforward as shopping for a e book on Amazon."
The corporate's buyer success tales are spectacular however stay a small fraction of its buyer base. With 150,000 Salesforce clients and a million Slack clients, the 12,000 Agentforce deployments symbolize roughly 8% penetration — sturdy for a one-year-old product line, however hardly ubiquitous.
The corporate's inventory, down roughly 28% year to date with a Relative Strength rating of just 15, suggests buyers stay skeptical. This week's Dreamforce demonstrations — and the months of buyer deployments that comply with — will start to supply solutions as to whether Salesforce can lastly transfer enterprise AI from pilots to manufacturing at scale, or whether or not the "$7 billion enterprise" stays extra aspiration than actuality.
