{"slug":"conversation-ai","title":"GHL Conversation AI Agent — Bot Goals + Pipeline Filtering","tags":["conversation-ai","automation","workflows","pipeline","bot"],"agent_summary":"How to configure GHL's Conversation AI with goal-driven bots, pipeline-stage filtering to avoid wasting AI on hot leads, dead-lead revival blasts, and model selection guidance.","trigger_phrases":["conversation ai ghl","ghl bot","ai agent ghl","pipeline stage filter","dead lead revival","ghl automation bot","ghl ai followup","bot goals"],"runnable":true,"markdown":"\n# GHL Conversation AI Agent\n\n## Bot Goals Architecture\nConversation AI in GHL is goal-driven, not script-driven. Each bot has:\n- **Primary goal** — what success looks like (booked call, qualified lead, form fill)\n- **Guardrails** — off-topic handling, objection responses\n- **Handoff trigger** — \"book a call\" intent → transfer to human / booking flow\n\n## Pipeline-Stage Filtering (the trick)\nDon't run the bot on everyone. Filter by pipeline stage:\n\n```\nTrigger: Inbound SMS\n  → Find Opportunity (by contact)\n  → If/Else: Sales Pipeline stage = \"Engaged\"\n    ├─ True: Conversation AI action\n    └─ False: Do nothing (or route to sales)\n```\n\nKeeps the bot focused on the funnel stage where AI outperforms humans (tire-kicker nurture), skips hot leads who need a real rep.\n\n## Auto-Followup Config\n- Max touchpoints: **5**\n- Max delay per touch: **24 hours**\n- Escalating urgency tone across touches\n- Kill switch: any reply that includes \"stop\", \"not interested\", booking confirmation, or human-handoff keyword\n\n## Dead-Lead Revival\nRun the bot as a one-shot blast on contacts with stage = `Cold` and `last_contacted > 30d`. Conservative messaging, no pressure. Recovers ~5-8% to booked calls on Iron Automations' own data.\n\n## Model Choice\nGHL lets you pick GPT-4 / Claude / Gemini. Claude Sonnet wins on natural reply cadence; GPT-4o wins on aggressive closing. A/B by pipeline vertical.\n","html":"<h1>GHL Conversation AI Agent</h1>\n<h2>Bot Goals Architecture</h2>\n<p>Conversation AI in GHL is goal-driven, not script-driven. Each bot has:</p>\n<ul>\n<li><strong>Primary goal</strong> — what success looks like (booked call, qualified lead, form fill)</li>\n<li><strong>Guardrails</strong> — off-topic handling, objection responses</li>\n<li><strong>Handoff trigger</strong> — \"book a call\" intent → transfer to human / booking flow</li>\n</ul>\n<h2>Pipeline-Stage Filtering (the trick)</h2>\n<p>Don't run the bot on everyone. Filter by pipeline stage:</p>\n<pre><code>Trigger: Inbound SMS\n  → Find Opportunity (by contact)\n  → If/Else: Sales Pipeline stage = \"Engaged\"\n    ├─ True: Conversation AI action\n    └─ False: Do nothing (or route to sales)\n</code></pre>\n<p>Keeps the bot focused on the funnel stage where AI outperforms humans (tire-kicker nurture), skips hot leads who need a real rep.</p>\n<h2>Auto-Followup Config</h2>\n<ul>\n<li>Max touchpoints: <strong>5</strong></li>\n<li>Max delay per touch: <strong>24 hours</strong></li>\n<li>Escalating urgency tone across touches</li>\n<li>Kill switch: any reply that includes \"stop\", \"not interested\", booking confirmation, or human-handoff keyword</li>\n</ul>\n<h2>Dead-Lead Revival</h2>\n<p>Run the bot as a one-shot blast on contacts with stage = <code>Cold</code> and <code>last_contacted > 30d</code>. Conservative messaging, no pressure. Recovers ~5-8% to booked calls on Iron Automations' own data.</p>\n<h2>Model Choice</h2>\n<p>GHL lets you pick GPT-4 / Claude / Gemini. Claude Sonnet wins on natural reply cadence; GPT-4o wins on aggressive closing. A/B by pipeline vertical.</p>\n"}