Automate Stalled Deal Follow-Up in n8n

Deals don't usually die from a no. They die from silence. A prospect goes quiet after a demo, the rep means to follow up, a week passes, then a month, and the deal quietly rots in the pipeline. An n8n stalled deal follow-up workflow scans the CRM for exactly those gone-quiet deals and fires a warm, contextual re-engagement before they're cold.

The agency blog posts describe this as "scan the CRM daily and flag at-risk deals." They never show the three steps that make it work: the no-activity-since detection query, the dedupe that stops you nudging the same deal every day, and the AI step that writes copy referencing the actual conversation. This builds all three.

It's warm re-engagement, not cold prospecting. Different job, different copy, different node chain.

What you can automate around a stalled deal

A re-engagement workflow handles the follow-up a busy rep keeps forgetting:

• A daily scan for open deals with no activity past a cutoff
• Dedupe against deals already nudged inside a cooldown window
• AI re-engagement copy that references the deal stage and last touch
• A Slack alert when a high-value deal goes quiet, for a human reply
• Logging every nudge so the cadence stays sane
• Escalation after two unanswered nudges: hand it to the rep or mark it lost

The workflow does the chasing the rep intends to do and never gets to.

Why detection and dedupe are the hard parts

Writing a re-engagement email is the easy step now; an LLM does it in one node. The two steps that actually break are detection and dedupe.

Detection is a query, not a vibe. You need open deals whose last_activity_at is older than your cutoff. Most CRMs expose that field; the trick is filtering on it correctly and not catching deals that are legitimately mid-conversation. A deal touched yesterday isn't stalled even if it's old.

Dedupe is the one that quietly humiliates you. Run a daily scan without a last-nudged guard and you email the same stalled deal every morning. The prospect who went quiet now gets a "just checking in" every 24 hours. That's not re-engagement, that's harassment, and it's the single most common way this workflow goes wrong.

Here's the take: the cooldown matters more than the copy. A perfectly written nudge sent daily is worse than a mediocre one sent once a week. Store a last_nudged_at per deal, gate every send on a 5-to-7-day cooldown, and cap total nudges at two before escalating to a human. Get the cadence right and average copy converts. Get the cadence wrong and great copy still burns the lead.

The stalled-deal pipeline

Schedule Trigger (daily)
  → Read open deals (HubSpot / Pipedrive / Sheets)
  → Filter: last_activity older than cutoff
  → Filter: not nudged within cooldown AND nudges < 2
  → OpenAI: write re-engagement copy from deal context
  → Send email + log nudge
  → Slack alert if high-value

Two filters back to back: one finds the stalled deals, one removes the ones you've recently touched. That second filter is the dedupe, and it's the node the blog posts never mention.

1. Detect the stall

The Schedule Trigger fires daily and reads open deals from the CRM. A filter keeps only those whose last_activity_at is older than the cutoff. Ten days is a reasonable start; tune it to your sales cycle.

// Filter: no activity in 10+ days, still open
const d = items[0].json;
const days = (Date.now() - new Date(d.last_activity_at)) / 864e5;
return d.stage !== 'closed' && days >= 10;

2. Dedupe the already-nudged

The second filter is the dedupe. Drop any deal nudged inside the cooldown, or already nudged twice. Read last_nudged_at and nudge_count from your tracking sheet or a CRM custom field.

// Filter: respect cooldown and the 2-nudge cap
const d = items[0].json;
const since = (Date.now() - new Date(d.last_nudged_at || 0)) / 864e5;
return since >= 7 && Number(d.nudge_count || 0) < 2;

3. Write the re-engagement copy

The OpenAI node writes a short, warm email referencing the deal stage and the last real interaction, not a generic "circling back." Feed it the deal name, stage, and the last activity note. Force structured output (subject plus body) and parse it with a Code node. Skip the parse and the email node sends raw model text where it expects a subject field; the send fails or mangles. AI workflows break at the parse step more than anywhere else.

4. Send, log, escalate

Send the email (Gmail or your sending node), then immediately write back last_nudged_at and increment nudge_count. Order matters: if you send first and the write fails, tomorrow's run re-sends. For high-value deals, a Slack alert lets a human take the reply instead of the bot. After the second nudge with no response, escalate: assign to the rep or mark the deal dormant. Don't nudge a third time.

5. Close the loop

When a prospect replies, a separate inbound flow should clear the deal from the stalled set so it stops getting scanned. The reply is the signal that re-engagement worked; let it reset the state.

Implementation patterns

Pattern 1: two-filter detection. One filter finds stale deals, a second removes recently-nudged ones. Both are required; the second is the one everyone forgets.

Read deals → Filter(stale) → Filter(cooldown + cap) → act

Pattern 2: context-aware AI copy. Feed deal stage and last-touch into the prompt so the email reads like a real follow-up, not a template. The AI Lead Scoring and Email Routing template uses the same GPT-4o-mini structured-output pattern if you want the scoring-and-routing core pre-built.

deal context → OpenAI(structured: subject+body) → parse → send

Pattern 3: state-driven escalation. Track nudge_count; after two, route to a human instead of sending again. Keeps the automation from outstaying its welcome.

n8n nodes you'll use most

Getting started

1. Pick your CRM source and confirm it exposes a last-activity timestamp per deal.
2. Add a Schedule Trigger on a daily Cron.
3. Read open deals and add the stale-detection filter on last_activity_at.
4. Add the dedupe filter on last_nudged_at and a 2-nudge cap.
5. Add an OpenAI node with structured output and a parse step for the copy.
6. Send the email, then write last_nudged_at and increment nudge_count.
7. Add a Slack escalation for high-value deals and a reply-handler that clears the stall.

Re-engaging warm deals is one half of a healthy pipeline; scoring and routing the new ones is the other. For the inbound side, see lead scoring automation in n8n, and for the broader CRM picture there's automating your CRM with n8n. Detect with a real query, dedupe with a cooldown, and the silence stops costing you deals.