Lindy HubSpot Sales Workflow: Inbox-Zero Agent Guide

Every sales ops team I’ve worked with has the same inbox problem. Inbound demo requests, pricing questions, and “just checking in” replies pile up overnight, reps triage manually in the morning, and the genuinely hot lead from 11 PM gets a response at 10 AM. Speed-to-lead matters more than almost any other variable in the conversion funnel, and a single missed hour compounds. This tutorial walks through exactly how I’d build a Lindy AI agent that monitors a sales inbox, classifies intent, creates or updates contacts in HubSpot, and pings the right rep, all without a Zapier chain holding it together.

What This Workflow Does and Why It Works

The core idea is dead simple. Give an AI agent the inbox as input, CRM write access as output, and let it handle the mechanical triage that currently steals 20 to 40 minutes from every rep’s morning. According to HubSpot Research’s State of Sales 2024, reps spend only 28% of their week actually selling. The rest disappears into data entry, internal communication, and exactly this kind of email admin.

Lindy is purpose-built for this pattern. Unlike Relevance AI, which excels at research-heavy, multi-tool agent chains, Lindy’s trigger model starts from communication events (email received, calendar invite, Slack message) and ships with native CRM write actions. That makes it genuinely faster to wire up for inbox-to-HubSpot workflows than either Relevance AI or a hand-rolled n8n flow. My opinion: if your primary use case is email triage into a CRM, don’t start with n8n. You’ll spend three days on HTTP request nodes that Lindy covers in three clicks.

The workflow I’ll build has five layers:

1. Email received in a Gmail or Outlook sales inbox
2. Lindy classifies intent (demo request, pricing, existing customer, spam)
3. Lindy looks up or creates a contact in HubSpot
4. If intent is “hot,” Lindy drafts a reply and notifies the rep in Slack
5. If intent is “nurture,” Lindy logs a note and enrolls the contact in a HubSpot sequence

Building the Agent Step by Step

Before you start, you need four things: a Lindy account on the Pro plan or higher (the free tier does not expose multi-action workflows), a HubSpot account with API access enabled, a Gmail or Outlook inbox dedicated to sales (shared alias or individual rep), and a Slack workspace where reps actually live.

The Gotcha: Prompt Drift and Malformed JSON

The single most common failure mode I’ve seen is the classification prompt returning free-text instead of valid JSON when the email body is unusually long or written in a language Lindy’s instruction wasn’t expecting. The entire downstream mapping breaks silently. No contact update. No Slack ping. The lead just dies.

// Expected Lindy output (enforce this format in prompt)
{
  "intent": "demo-request",
  "urgency": "hot",
  "summary": "Prospect at 200-person SaaS co asking for a live demo this week, mentions budget approved."
}

// Prompt closing line to add:
// "Respond ONLY with valid JSON matching
// the schema above. No markdown, no
// explanation, no code fences."

A second gotcha: HubSpot’s sequence enrollment API requires the contact to have a valid, opted-in email address and the enrolling user to have sequence permissions. I’ve had Lindy throw a silent 403 on enrollment for half a batch because the OAuth token was scoped to read-only. Double-check your HubSpot private app or OAuth scopes include crm.objects.contacts.write and automation.sequences.write before you assume the workflow is working. According to HubSpot’s Sequences API documentation, missing scope errors won’t always surface as obvious failures in the UI.

Which Teams Should Build This (and Which Shouldn’t)

Is This Agent Right for Your Team?

Choose Lindy + HubSpot if

• You have a dedicated inbound sales alias with 20 or more emails per day
• Your reps are currently spending 30 or more minutes per day on inbox triage
• You already use HubSpot as your CRM and have sequence infrastructure in place

From $49/mo (Lindy Pro) Try Lindy →

Choose n8n + HubSpot if

• You need more custom logic than Lindy's action library covers (e.g., enrichment waterfalls, multi-source lookups)
• You have an engineer or technical RevOps person who can maintain a self-hosted workflow
• Budget is tight and you want to avoid per-action pricing at scale

From $20/mo (cloud) or self-hosted Try n8n →

Choose Relevance AI + HubSpot if

• Your triage requires deep research on the sender before routing (company size, tech stack, funding)
• You want to chain multiple AI tools (web search, LinkedIn lookup, scoring model) in one agent
• You're building a more complex SDR agent, not just an email classifier

From $19/mo Try Relevance AI →

This workflow pairs well with signal-based outbound if you want to close the loop between inbound intent and outbound triggers. And if you’re enriching contacts as part of the triage step, the Clay enrichment waterfall approach is worth layering in before the HubSpot upsert.

Ship It, Then Tune It

The workflow above is a 60 to 90 minute build. The prompt is the only thing that needs ongoing attention. Review misclassified emails weekly for the first month, add concrete examples to the instruction block, and your accuracy will compound. According to Google Cloud’s research on AI agent reliability, structured output constraints are the single highest-leverage reliability lever for production agents. That’s exactly why the JSON enforcement step is non-negotiable and not optional polish.

Once the agent is stable, the math is straightforward. If you recover 30 minutes per rep per day and pay $49 per month for Lindy, you’ve paid for the tool before the first coffee break. For teams I’ve set this up with, the real win isn’t the time savings. It’s that a hot lead at 11 PM gets a draft reply in Slack by 11:05, and the rep sends it before the prospect opens their laptop the next morning.