SDRs chasing the right 20%, not the hot-guessed 80%.
AI scoring on top of your CRM. Enriched firmographic signals + LLM inference + explanations. Routed to right-fit SDR, quickly.
Who this is for
RevOps or marketing lead with noisy inbound where SDRs are chasing unqualified leads and CRM lead scores are based on stale heuristics from 2021.
The pain today
- Inbound leads exceeding SDR capacity to qualify manually
- HubSpot or Salesforce lead score based on rules nobody tunes
- SDRs ignoring the score because it doesn't match their gut
- Enterprise and SMB leads routed to the same queue
- No explanation of WHY a lead is scored high or low
The outcome you get
- AI scoring trained on your historical conversion data
- Enrichment layer (Clearbit, Apollo, custom) feeding the score
- Score explanations SDRs can read and trust
- Automatic routing — enterprise to AEs, SMB to SDRs, dead to nurture
- Weekly score recalibration as new conversion data accumulates
Lead data you already have vs need
Most B2B companies have more scoring signal than they think. Already captured: firmographic (company size, industry, revenue from CRM), behavioral (pages viewed, time on site, form fields, content downloaded), engagement (email opens, clicks, event attendance). Missing but accessible: tech stack (BuiltWith, Wappalyzer), funding signals (Crunchbase, PitchBook), intent data (G2, TrustRadius, search-based intent providers). Enrichment cost is usually $0.10–1 per lead through providers like Clearbit, Apollo, or ZoomInfo. I audit what you already capture, recommend targeted enrichment for gaps, and tune the score against actual historical wins/losses, not generic ICP theory.
LLM + classic ML hybrid scoring
Best-performing scoring combines two approaches. Classic ML (logistic regression, gradient boosting): great on numeric features, interpretable, cheap to run, handles the 'of similar leads in the past, X% converted' logic. LLM: great on unstructured signal — website copy, LinkedIn bio, email content — inferring fit from context. Pipeline: classic ML gives a baseline score from firmographics and behavior; LLM layer adjusts based on qualitative signals (e.g., 'website mentions cybersecurity compliance, matches our ICP signal'). Each contribution is explainable. This hybrid outperforms pure-LLM or pure-ML on most B2B scoring tasks and costs less than pure-LLM per lead.
CRM integration: HubSpot and Salesforce
Scoring has to land in the tools SDRs live in. HubSpot: scoring written back to custom property on contact record, score explanations in a custom field, routing via HubSpot Workflows based on score + firmographic rules. Salesforce: scoring on lead object, routing via Salesforce assignment rules or Distribution for Salesforce. Both: daily rescoring for leads that accumulate new behavior, immediate scoring for new form submissions. Webhook-based so scoring happens in seconds, not hours. SDRs see the score plus the explanation plus confidence — they decide who to call based on real reasoning, not a black box number.
Explanations SDRs can trust
A lead score without an explanation is a number nobody trusts. Every score comes with a 2–3 sentence explanation: what signals pushed it up or down, what's missing that would increase confidence. Example: 'Score 85 (high). Company matches ICP (SaaS, 100–500 employees, US-based). Two decision-makers visited pricing page in last 7 days. Missing: budget signal — consider discovery to confirm.' SDRs see this at a glance, prioritize confidently, and can use the missing-signal insight to frame their call. Explanations are generated by the LLM layer and cached per lead until next rescoring.
Routing and workflow
Score alone doesn't convert — routing decides speed. Enterprise-profile leads (>500 employees, high industry fit) route to AE queue within 5 minutes of form submission. Mid-market to SDR round-robin. SMB to automated nurture with periodic rescoring. Cold/wrong-fit leads explicitly marked so your pipeline doesn't show false hope. Routing rules adjust over time as conversion data accumulates — what was SDR-worthy 6 months ago may now be AE-worthy or vice versa. I build the rules collaboratively with your RevOps team; the AI provides the score, humans provide the routing logic that fits your sales motion.
Pricing
AI lead scoring fits the AI Automation retainer at $3,000/mo. First-version timeline: 4–6 weeks to train model, integrate CRM, ship routing. Retainer continues through weekly recalibration and quarterly model retraining — SDR feedback on score accuracy is a continuous input. 14-day money-back, cancel anytime, Work Made for Hire. Enrichment provider costs (Clearbit, Apollo, ZoomInfo) and LLM API costs billed directly to you — typical total $200–2,000/mo depending on lead volume.
Frequently asked questions
The questions prospects ask before they book.
- How much historical data do I need for accurate scoring?
- Minimum 500 historical leads with known outcomes (closed-won, closed-lost, no-response) is enough for baseline scoring. 5,000+ leads produces more confident models. Less than 500 leads means starting with rules-based scoring and transitioning to ML as data accumulates — I'll tell you honestly what's possible with your current data.
- Will SDRs actually use the score?
- Depends on explanation quality and score-vs-outcome accuracy. I work with SDR leadership to train the team on interpreting explanations, and I share weekly accuracy reports (which high-scored leads converted, which low-scored leads didn't) so SDRs see the model calibrate. Skepticism is healthy and usually fades after 4–6 weeks of visible accuracy.
- Can it score existing database leads, not just new ones?
- Yes — rescoring sweeps the entire database on a schedule (typically weekly). Old leads that match ICP better than their current segment get bumped into the right queue. Dormant accounts showing new signals (new hire on LinkedIn, company raised funding) resurface. This is where most value appears — reactivating right-fit leads from the existing pipeline.
- What about GDPR and data privacy?
- Lead data processed with same privacy rules you already apply. Enrichment providers' DPAs cover their data sources. LLM processing routed through Anthropic or OpenAI Enterprise (both offer DPAs). EU-only data residency achievable with Azure OpenAI in EU regions or self-hosted models.
- Can you integrate with my MAP (Marketo, Pardot)?
- Yes. MAP integrations work through the same patterns as CRM — webhook-based score updates, custom field writes, workflow triggers based on score changes. Most common MAPs (Marketo, Pardot, Mailchimp, HubSpot Marketing Hub) supported. Integration phase typically adds 1 week to initial timeline.
Ready to start?
Tell me what you need in 60 seconds. Tailored proposal in your inbox within 6 hours.