AI’s next act: how Salesforce is turning efficiency gains into revenue

$100 million in annualized cost savings. Over 3,200 opportunities influenced. Those are the kinds of outcomes leaders everywhere are hoping AI will deliver. But in conversations with CEOs this year, one thing was clear: very few companies can point to results at this level yet.

Recommended Video

Many organizations made progress using AI to improve efficiency in 2025. A much smaller group, including companies like Salesforce, is now beginning to see AI contribute to revenue. Across industries, last year’s focus was operational: automate support, streamline internal workflows, reduce repetitive labor. The mandate was straightforward: use AI to improve efficiency and protect margins. Now the pressure has shifted. Boards and investors are asking a tougher question: Where does AI show up in growth?

As part of our research into this shift, who better to go to than a leader in enterprise AI and CRM, Salesforce, which is one of the relatively small number of companies that can show measurable results in both cost reduction and revenue impact from AI agents. Their experience offers an early look at what this transition can look like in practice, and what we can learn from how Salesforce uses its own products at scale as their own “Customer Zero.”

Phase One (2025): Removing the Constraints of Reactive Service

For decades, service organizations operated under a basic constraint: human capacity. Every decision about service levels, speed, and personalization ultimately came down to how many people were available to respond.

In 2025, Salesforce began breaking that constraint. The company deployed AI agents across its customer support ecosystem, starting with its self-service help portal, help.salesforce.com. The portal features a conversational, ChatGPT-style interface and the initial goal was simple:

- Answer common customer questions autonomously

- Maintain conversational context

- Escalate to humans when judgment, complexity, or extreme urgency was required

The Results: Better Service and Lower Cost

In just over one year, Agentforce handled 3 million support conversations, contributing to measurable operational and financial impact:

- Lower case volume despite customer growth — Year-over-year support caseload dropped by 8%, representing more than 170,000 fewer cases, with further reductions forecast.

- Customer experiences previously out of reach — Salesforce now delivers live, synchronous chat support in seven languages, with plans to expand to 14 or more languages by year end, something it had never achieved in its 27-year history.

- More proactive service — With AI agents handling routine questions, human teams can focus on more proactive services to help customers succeed, preventing problems, and building stronger relationships without increasing costs.

- $100M in annualized cost savings — Agentforce reduced support costs while maintaining customer satisfaction — a rare combination in service transformation efforts.

“We already know that AI agents can scale our cost structure infinitely, but the real unlock is that they can help us scale our capacity, too,” says Jim Roth, President, of Customer Success at Salesforce. “When our capacity is infinite, we can be proactive and build more incredible customer experiences. We can treat every customer like they’re our most important customer.”

AI agents helped Salesforce improve service while lowering the cost to deliver it. But at this stage, the primary impact was operational. Margins improved and service expanded, but revenue remained largely unchanged.

Phase Two: Turning AI Agents Toward Growth

By the time leaders reached 2026, the expectation had evolved. If AI could remove cost from service, could it also help create revenue?

Inside Salesforce, one experiment offered an early answer. It started with what employees informally called “sawdust.” Like most large B2B companies, Salesforce generates massive inbound interest through its digital channels: content downloads, webinar registrations, information requests. Each interaction is technically a lead.

But in practice, many of these leads never receive follow-up. They essentially collect like sawdust. Sales teams focus on the highest-scoring prospects. Marketing prioritizes defined segments. A long tail of lower-priority leads sits in the system, untouched. They weren’t worthless. They were simply uneconomical for humans to pursue. That’s where AI agents entered the picture.

The “Sawdust” Experiment

Salesforce deployed an AI agent to engage these dormant leads autonomously. The agent could:

- Send personalized outreach

- Ask qualifying questions

- Respond based on context

- Identify signals of genuine buying intent

- Route promising prospects to human teams

These were leads the company didn’t have the capacity to work anyway, which made them a low-risk but high-upside test case.

In a short period of time, the agent began working on hundreds of thousands of previously untouched leads. The