Back to Help Center

AI Features

Advanced machine learning and predictive intelligence

AI Lead Scoring Architecture

Machine Learning Model Stack

Lotivio employs a multi-model ensemble approach combining gradient-boosted decision trees (XGBoost), neural networks, and proprietary causal inference models to predict lead conversion probability with 94.3% accuracy.

200+ Signal Categories Analyzed

Behavioral Signals
  • • Page views, session duration, scroll depth
  • • Video engagement, content downloads
  • • Return visitor patterns, time between visits
  • • Navigation paths, feature exploration
Engagement Signals
  • • Email open rates, click-through rates
  • • Reply sentiment analysis
  • • Calendar invitation acceptance
  • • Sales call participation and duration
Firmographic Data
  • • Company size, revenue, growth rate
  • • Industry, sub-industry classification
  • • Technology stack detection
  • • Funding stage, recent events
Intent Signals
  • • Third-party intent data (Bombora, G2)
  • • Competitive research indicators
  • • Job posting analysis
  • • Budget cycle timing

Real-Time Score Updates

Scores recalculate within 30 seconds of new signal ingestion. No batch jobs, no overnight delays. Stream processing architecture powered by Apache Kafka ensures sub-minute latency for scoring updates across 10M+ leads.

Score Interpretation & Thresholds

90-100

Hot Lead - Immediate Action

High intent + strong fit. Assign to sales within 5 minutes. Conversion probability: 35-45%

75-89

Warm Lead - Nurture Priority

Good signals, needs engagement. AI outreach + SDR follow-up. Conversion probability: 18-25%

50-74

Developing - Automated Nurture

Early stage interest. AI-driven education sequence. Conversion probability: 8-15%

<50

Cold - Long-Term Nurture

Minimal signals or poor fit. Quarterly check-ins. Conversion probability: <5%

Custom Model Tuning (Enterprise)

Enterprise Feature: Custom model tuning requires Lotivio Enterprise plan and at least 50,000 historical conversions for statistically significant training.

Signal Weight Customization

Adjust the relative importance of signal categories to match your unique sales process:

Example: SaaS Enterprise Sales

Firmographic Fit (Company Size, Industry)
40%
Third-Party Intent Data
30%
Website Engagement
20%
Email Engagement
10%

Setting Custom Weights

  1. 1. Navigate to Settings → AI Models → Scoring Configuration
  2. 2. Select "Custom Model Training" and upload historical conversion data (CSV/API)
  3. 3. Adjust signal category weights using the visual slider interface
  4. 4. Run backtesting against 20% held-out validation set to measure accuracy improvement
  5. 5. Deploy to production or shadow mode for A/B comparison

LLM-Powered Personalized Outreach

GPT-4 Turbo Message Generation

Lotivio's LLM engine generates hyper-personalized outreach messages by synthesizing lead-specific context from multiple data sources:

Context Sources

  • ✓ Company news & recent funding rounds
  • ✓ LinkedIn profile, role, tenure
  • ✓ Content engagement history
  • ✓ Pain points from previous conversations
  • ✓ Industry trends & challenges
  • ✓ Competitive landscape positioning

Personalization Variables

  • • First name, company name, role
  • • Specific page visited (e.g., "Pricing for Enterprise")
  • • Industry-specific use case
  • • Mutual connections or customers
  • • Relevant case study or resource
  • • Time zone-aware send timing

Example: Before & After

❌ Generic Template

"Hi [First Name], I noticed you visited our website. We help companies like yours improve lead conversion. Would love to chat!"

✓ LLM-Personalized

"Hi Sarah, I saw you checked out our automotive dealership case study last week — makes sense given Acme Auto Group's 15-location footprint across the Southwest. Most groups your size tell us their biggest challenge is recovering leads that go dark after the initial inquiry. We helped Desert Motors (also Phoenix-based) recover 23% of abandoned leads in Q4. Would a 15-min overview of how we'd approach Acme's pipeline be valuable?"

Compliance & Brand Voice Control

Enterprise customers can define custom brand guidelines, tone requirements, and compliance rules:

  • 1
    Brand Voice Profile: Upload example emails, define tone (professional, casual, technical), specify banned phrases
  • 2
    Legal Review: Optional human approval workflow for messages to regulated industries (healthcare, finance)
  • 3
    Auto-Compliance: CAN-SPAM, GDPR, CCPA disclosures automatically inserted per recipient location

Autonomous AI Agents (Beta)

Beta Program: Autonomous agents are available to Enterprise customers in controlled beta. Contact your CSM to request access.

Our autonomous agents can execute multi-step recovery workflows without human intervention:

Capabilities

  • • Monitor lead behavior 24/7 and trigger outreach at optimal moments
  • • Dynamically adjust message content based on latest signals
  • • Escalate to human SDR when buy signals exceed threshold
  • • A/B test message variations and optimize for reply rate
  • • Coordinate across channels (email, LinkedIn, SMS) with consistent voice

Success Metrics

Early beta customers report:

3.2x
Response Rate Increase
67%
SDR Time Saved
28%
More Meetings Booked