The problem: the client runs dozens of deals at once in their CRM. Management had no real-time view of which deals were stalling or at risk of being lost. Daily pipeline reporting was assembled by hand, took hours, and still missed context buried in notes and call logs.
What I built:
A multi-tier system in n8n with an AI agent.
Daily reporting: the agent analyzes the sales funnel and autonomously calls a tool to pull deeper detail on problematic deals. The finished report lands in Telegram.
Weekly deal health: a separate workflow uses Loop Over Items and a dedicated sub-workflow to aggregate notes, calls and tasks for each deal. An AI agent with a Structured Output Parser evaluates the aggregate and returns JSON: a health score from 1 to 10, risk factors, recommended actions.
History: results are written to Google Sheets with Append or Update logic mapped by Deal ID, so no duplicates build up.
Routing: a Switch node instantly pushes Critical (score 4 or below) and Medium (7 or below) alerts to management via Telegram.
Result: manual pipeline reporting eliminated, management gets alerted before a deal collapses, and the weekly forecast is based on real CRM activity instead of gut feel.
Stack: n8n (sub-workflows, loops), CRM API, OpenAI and Claude API, Google Sheets, Telegram API.
#n8n #automation #AI #CRM #sales
What I built:
A multi-tier system in n8n with an AI agent.
Daily reporting: the agent analyzes the sales funnel and autonomously calls a tool to pull deeper detail on problematic deals. The finished report lands in Telegram.
Weekly deal health: a separate workflow uses Loop Over Items and a dedicated sub-workflow to aggregate notes, calls and tasks for each deal. An AI agent with a Structured Output Parser evaluates the aggregate and returns JSON: a health score from 1 to 10, risk factors, recommended actions.
History: results are written to Google Sheets with Append or Update logic mapped by Deal ID, so no duplicates build up.
Routing: a Switch node instantly pushes Critical (score 4 or below) and Medium (7 or below) alerts to management via Telegram.
Result: manual pipeline reporting eliminated, management gets alerted before a deal collapses, and the weekly forecast is based on real CRM activity instead of gut feel.
Stack: n8n (sub-workflows, loops), CRM API, OpenAI and Claude API, Google Sheets, Telegram API.
#n8n #automation #AI #CRM #sales