Case Study · Salesforce Blog · 6/2/2026

How a SaaS Team Used AI Agents to Follow Up 1,000 Leads in 24 Hours

一支 SaaS 销售团队用 AI Agent 在 24 小时内跟进了 1000 条线索

# sales⚡ automationLangChain🟢 No-code
Why it matters
Response rate going from 4% to 23% is the real proof point. The agent reads intent signals before deciding whether to follow up at all — that's why quality improved alongside volume.

The Problem

A 12-person SaaS sales team was drowning in unqualified leads. With a pipeline of 1,000+ new contacts per day, each rep was manually deciding who to follow up with — a process that was slow, inconsistent, and based on gut feel rather than data.

The Agent Solution

They built a LangChain agent that connects to their HubSpot CRM via API. Each morning, the agent pulls the day's new leads, reads their activity history (email opens, page visits, demo requests), scores their purchase intent on a 1-10 scale, and drafts a personalized follow-up email for scores above 6. The rep only reviews and sends — no writing required.

Results

How It Works

The agent uses a two-step pipeline. First, a "scoring agent" reads CRM data and outputs a JSON object with intent score and reasoning. Second, a "writing agent" takes that score and drafts an email tailored to the lead's specific behavior — if they visited the pricing page three times, the email mentions pricing; if they opened the onboarding guide, the email mentions ease of setup.

Can You Do This Without Code?

Yes. You can replicate 80% of this with Make.com or Zapier + a ChatGPT/Claude action node. The main limitation is custom intent scoring — no-code tools can trigger on simple conditions (email opened = follow up) but can't do nuanced multi-signal scoring. For most small sales teams, that's still a massive improvement over manual work.

Related Cases