Salesforce just saved $100 million with AI – and then found revenue no one was even trying to close.
$100 million in cost savings. 3,200+ new sales opportunities. And it all started with a pile of leads nobody wanted.
That’s what Salesforce reported in April 2026 after deploying Agentforce – its own AI agents – across customer support and sales. The results are worth studying carefully, because the playbook is transferable to businesses of almost any size.
What They Did
Starting in 2025, Salesforce deployed AI agents on help.salesforce.com – its main customer portal. The setup was straightforward: the AI handled routine support questions conversationally, maintained context across the conversation, escalated to human agents only when necessary, and operated in seven languages around the clock.
In just over one year, those agents handled 3 million support conversations. The impact: a year-over-year drop in support caseload of 8% (that’s 170,000 fewer tickets), $100 million in annualized operating savings, and improved customer satisfaction scores. Not a tradeoff – a simultaneous improvement in quality and cost.
Then they turned the same technology toward revenue.
Like most large B2B companies, Salesforce generates a massive inbound stream of leads from webinars, content downloads, and information requests. Most of these leads receive no follow-up. They’re not bad leads – they’re just uneconomical for a human sales rep to pursue. Salesforce called this pile “sawdust.”
They pointed AI agents at all of it. The agents sent personalized outreach, asked qualifying questions, responded based on context, identified signals of real buying intent, and routed only the promising prospects to human reps. From leads that would otherwise have been ignored, they influenced 3,200+ sales opportunities and generated actual closed revenue.
Why This Actually Worked
First, AI has no capacity ceiling. A human sales team can only contact so many leads. An AI agent can contact all of them, simultaneously, at 2 a.m. if needed. The economic calculus that made “sawdust” leads unprofitable for humans simply doesn’t apply to software.
Second, they started with cost reduction, not revenue. This gave the team time to build trust in the system, refine how agents escalate, and prove reliability before betting the revenue pipeline on it. By the time they turned AI toward growth, the foundation was solid.
Third, they used AI on their own products internally first – being “Customer Zero” – which created real feedback loops and made the case studies credible.
I’m Mike Partners. I founded AiExpert.org because I believe the strategies behind billion-dollar AI deployments should be accessible to every business owner. Here’s how to put this one into practice.
How to Apply This to Your Business
Start by finding your own “sawdust” pile – pull a list of leads from the last 18-24 months that never converted, people who downloaded something, attended a webinar, or filled out a form, because most businesses have hundreds or thousands of these sitting untouched. Then deploy an AI outreach tool like Clay, Reply.io, Instantly, or Apollo, which now have AI-personalized outreach features that can send individualized messages at scale for $50-$300 per month. And before you hire another customer service person, add AI to your support first by testing whether an AI agent can handle your 10 most common inbound questions, since most SMBs find that 60-70% of support volume is the same questions on repeat.
The companies that figure this out first won’t just be more efficient. They’ll compound. Every saved hour becomes a growth hour. Every “sawdust” lead becomes a potential customer. That’s the real unlock.
Frequently Asked Questions
How did Salesforce save $100 million with AI agents?
Salesforce deployed Agentforce AI agents on their customer support portal, handling 3 million conversations in just over a year. The agents resolved routine questions conversationally in seven languages, reducing support caseload by 8% (170,000 fewer tickets) while simultaneously improving customer satisfaction scores.
What is the “sawdust” lead strategy Salesforce used?
Salesforce used the term “sawdust” to describe the massive pile of inbound leads that marketing captured but sales never had time to follow up on. By pointing AI agents at these neglected leads with personalized outreach, they generated over 3,200 new sales opportunities from prospects that would have otherwise been ignored entirely.
Can small businesses use AI for lead follow-up like Salesforce?
Absolutely. Mike Partners and the team at AiExpert.org recommend tools like Clay, Reply.io, Instantly, or Apollo, which offer AI-personalized outreach features for $50-$300 per month. These tools can send individualized messages at scale, qualifying leads and routing promising prospects to your sales team automatically.
Should I use AI for customer support or sales first?
Salesforce’s approach suggests starting with customer support. It’s lower risk, easier to measure, and builds organizational trust in AI before you bet your revenue pipeline on it. Once your support AI is running reliably, you can extend the same technology toward revenue generation with confidence.
How quickly can a small business implement AI customer support?
Most small businesses can deploy a basic AI support agent within a day using tools like Tidio, Intercom AI, or Gorgias AI. Start by identifying your 10 most common customer questions and training the AI on those answers. According to AiExpert.org, most SMBs find that 60-70% of their support volume consists of the same repetitive questions.
