| Phase | Strategy | Where to Find It | Impact |
|---|---|---|---|
| Week 1 | Curate focused knowledge base docs — small & targeted beats massive dumps | Documents tab | Faster RAG, fewer tokens per query |
| Week 1 | Tune LLM Instructions to shape AI tone, persona, and guardrails | LLM Instructions tab | Consistent, on-brand responses |
| Week 1 | Monitor Discussion Activity to see what visitors actually ask | Analytics → Discussion Activity | Reveal content gaps & high-intent topics |
| Week 2 | Set your DQL Threshold — score visitors by conversation quality, not clicks | AI Scoring tab | Qualify pipeline with real buying signals |
| Week 2 | Run AI Segmentation to discover natural visitor groupings you didn't anticipate | Segments → Run AI Now | Precision targeting for campaigns |
| Week 2 | Build Smart Engagement Campaigns with compound rules (visits + conversations + time) | Campaigns → Targeting | Hyper-targeted messaging |
| Week 3 | Build Agents: 'When score > 75, email sales + update CRM + Slack notify' | Agents tab | Full pipeline automation, zero latency |
| Week 3 | Enable bidirectional CRM Sync with field-level mapping | CRM Sync tab | Real-time revenue data alignment |
| Week 3 | Use drip campaigns with scheduled sends for nurture sequences | Campaigns tab | Automated multi-touch engagement |
| Week 4 | Enable Visitor Memory for cross-session personalization | App Config → Visitor Memory | Returning visitors feel known |
| Week 4 | Activate Response Caching to slash token costs 40–70% | App Config → Cache Management | Major cost reduction at scale |
| Week 4 | Enable bulk enrichment via Apollo/PDL for full visitor database firmographics | App Config → Enrichment Providers | Richer scoring & segmentation signals |
| Trigger Event | Action Chain | Business Outcome |
|---|---|---|
| Visitor score exceeds 75 | Email sales team + create CRM task + Slack webhook | Instant DQL handoff to revenue team |
| Form submitted | Add to segment + trigger drip campaign + update CRM record | Automated nurture entry |
| New DQL created | Send personalized re-engagement email + notify account owner | No lead falls through the cracks |
| Campaign triggered | Push custom field update to HubSpot/Salesforce + send confirmation email | Full lifecycle tracking in CRM |
| Score threshold crossed | Assign to specific segment + update visitor profile + fire external API | Dynamic audience management |
| Technique | How to Enable | Expected Savings |
|---|---|---|
| Response Caching | App Config → Cache Management | 40–70% token reduction on repeat queries |
| Use Claude Haiku for high-volume simple Q&A | App Config → LLM Providers | Significant cost reduction vs Sonnet/Opus |
| Curate smaller, targeted knowledge base collections | Collections tab → Organize by topic | Fewer tokens per RAG query |
| Use the Web Crawler selectively | App Config → KB Tuning & Crawler → Crawl Content | Avoid over-indexing irrelevant pages |
| Set API token limits per conversation | App Config → API Limits | Cap runaway token usage per session |
| Condition A | Condition B | Condition C | Use Case |
|---|---|---|---|
| Visited 3+ times | Had 2+ conversations | Spent 60+ seconds on site | Warm engaged prospect campaign |
| DQL status = true | Score > 80 | Missing: Job Title field | Progressive profiling for hot leads |
| Segment = Enterprise | Visit count = 1 | Anonymous visitor | First-touch high-value prospect capture |
| Known visitor | Customer status = false | Visited pricing page | Bottom-of-funnel conversion push |
| Champion lifecycle | 3+ return sessions | Has email | Advocacy and expansion campaign trigger |