What is buyer-state-aware BANT scoring — and what is the AI-ML-disparate-impact-times-EU-AI-Act-Article-22 double-violation problem?

A DTC ecommerce + B2B SaaS + multi-vertical operator routes inbound leads to sales reps based on BANT scoring (Budget + Authority + Need + Timeline) augmented with firmographic enrichment (downstream of #561 sibling skill) + intent signal (6sense + Demandbase + Bombora + Madison Logic + RollWorks + Terminus + Foundry Intent + G2 + TrustRadius) + conversation-intelligence (Gong + Chorus + Outreach + Salesloft) + first-party engagement (GA4 + Adobe + Mixpanel + Amplitude + Heap + Pendo + Posthog). The four-skill bundle on the lead-scoring-routing agent — Compute, Calibrate, Gate, Audit — sits above the per-vendor lead-score substrate (Salesforce Einstein + HubSpot Predictive + Marketo Predictive + Pardot + Pipedrive AI + Zoho Zia + Dynamics 365 Sales Insights + Outreach Kaia + Salesloft Drift + Conversica) and writes a per-lead canonical record. The operationally distinctive anchor: ECOA Regulation B 12 CFR 1002 disparate-impact + 4/5ths rule per Uniform Guidelines on Employee Selection Procedures 1978 when AI-ML BANT score routes leads through proxies (industry + company size + location + firmographic correlate + tech stack + buying committee composition) that correlate with protected class. EU AI Act Article 22 + Annex III classifies AI-ML scoring driving employment + credit + housing + education access as high-risk. EEOC Enforcement Guidance on Software AI Algorithmic Decision-Making 2024 + NYC Local Law 144 AEDT bias audit + CFPB Circular 2022-03 adverse-action notification for AI-ML credit decisions all converge on the same scoring surface.

Why do Salesforce Einstein Lead Scoring + HubSpot Predictive + Marketo Predictive + 6sense + Demandbase break at AI-ML-disparate-impact compliance scale?

Each lead-scoring vendor ships a per-tenant flat score primitive — a Marketing Ops team enables the feature, the platform produces a 0-100 score per lead, sales routes leads above threshold. None calibrates the model with replication-crisis discipline (Brier score + Expected Calibration Error + reliability diagram + Rosenbaum bounds + E-value + sensitivity analysis + multiple-comparisons correction + pre-registration of model thresholds + 4/5ths rule audit). None gates the score against ECOA Regulation B disparate-impact + Title VII Bostock + ADEA + ADA + FCRA Section 604 prescreening + EEOC AI Guidance 2024 + NYC Local Law 144 AEDT bias audit + EU AI Act Article 22. None coordinates adverse-action notification when score drives material consequence (FCRA Section 615 + ECOA 12 CFR 1002.9 + CFPB Circular 2022-03). None writes a per-lead audit trail with regulatory-defense retention. None enforces FDD Item 12 territorial-protection when franchise BANT score crosses ADI/protected-territory. None enforces Sarbanes-Oxley 302/404/906 + FASB ASC 606 revenue-recognition when BANT score informs revenue accrual. The four-skill bundle Compute + Calibrate + Gate + Audit sits above the lead-scoring-vendor surface — it does not replace it.

How does Compute + Calibrate work across BANT scoring with replication-crisis discipline?

Compute runs per-portfolio per-vertical per-lead buyer-state-aware BANT scoring across 5 signal layers: (1) firmographic enrichment from #561 sibling skill (Apollo + ZoomInfo + Clearbit + Cognism + Lusha + LeadIQ + DiscoverOrg + People Data Labs + Coresignal + UpLead); (2) intent signal (6sense + Demandbase + Bombora Company Surge + Madison Logic + RollWorks + Terminus + Foundry Intent + Cyance + Aberdeen + TechTarget Priority Engine + G2 + TrustRadius + Capterra + Gartner Digital Markets + SoftwareReviews + IT Central Station); (3) conversation-intelligence (Gong + Chorus + Outreach + Salesloft + Mindtickle + Avoma); (4) second-party engagement (LinkedIn Sales Navigator + LinkedIn Insight Tag + Lead Gen Forms + Smart Links); (5) first-party engagement (GA4 + Adobe Analytics + Mixpanel + Amplitude + Heap + Pendo + Posthog + Mutiny + Optimizely + VWO + Statsig + GrowthBook). ML modeling: scikit-learn + XGBoost + LightGBM + CatBoost + TensorFlow + PyTorch + AutoML + DataRobot + H2O.ai + Alteryx + KNIME + Dataiku. Vector-similarity for intent classification: Algolia + Elasticsearch + Pinecone + Weaviate + Chroma + Qdrant + Milvus + Vespa + Marqo. Per-vendor LLM tie-breakers (OpenAI + Anthropic + Mistral + Llama) under zero-retention. Calibrate runs replication-crisis discipline: Brier score + Expected Calibration Error (ECE) + reliability diagram + isotonic regression / Platt scaling recalibration + Rosenbaum bounds for unmeasured-confounder sensitivity + E-value + multiple-comparisons correction (Bonferroni + Benjamini-Hochberg false-discovery-rate) + pre-registration of model thresholds + holdout-set replication + temporal cross-validation against drift + 4/5ths rule disparate-impact audit per protected-class proxy.

What does Gate + Audit do?

Gate runs 5 anchors per-lead before any score-driven routing decision commits. (1) ECOA Regulation B 12 CFR 1002 disparate-impact + 4/5ths rule per Uniform Guidelines on Employee Selection Procedures 1978 + Title VII Bostock + ADEA + ADA + Fair Housing Act + GINA + Equal Pay Act + EEOC Enforcement Guidance on Software AI Algorithmic Decision-Making 2024 + NYC Local Law 144 AEDT bias audit + Illinois HB 3773 AI Video Interview Act + CA SB-942 + CO SB 21-169 algorithmic discrimination + NY AI bill. (2) FCRA 15 USC 1681 Section 604 prescreening permissible-purpose + Section 615(d) firm-offer-of-credit + Section 615(e) opt-out + Section 1681i adverse-action notice + ECOA adverse-action notice + CFPB Circular 2022-03 adverse-action notification for AI-ML credit decisions + ECOA 12 CFR 1002.9 + GLBA Safeguards Rule. (3) EU AI Act Article 22 + Article 26 + Article 50 + Article 13 + 14 + 15 + Annex III high-risk + Article 6 + 27 Fundamental Rights Impact Assessment + Digital Services Act + DMA. (4) Sarbanes-Oxley 302/404/906 when public-company + FASB ASC 606 revenue-recognition + FASB ASC 842 + FASB ASC 326 CECL + SEC Reg S-K Item 303 MD&A + SEC Form 10-K segment. (5) FDD Item 12 territorial-protection + 15-state franchise registration + state UDTPA + Robinson-Patman + FDD Item 17 + 19 + FTC Act Section 5 + FTC Endorsement Guides when third-party-review intent signal + FTC Fake Review Rule + CFPB UDAAP + Lanham Act. Audit writes a per-lead WORM canonical record: signal layer snapshot + ML model snapshot + calibration metrics (Brier + ECE + reliability diagram + Rosenbaum bounds + E-value) + 4/5ths rule audit snapshot + per-anchor gate-pass + adverse-action-notice content + Fundamental Rights Impact Assessment artifact + AEDT bias-audit artifact + routing decision + revenue-accrual impact + per-lead replay capability. Storage: AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi WORM. Retention stacks: 7-year FTC + 7-year IRS + 6-year SEC + 3-year FINRA + 7-year FDD + 5-year FCRA Section 615(d) firm-offer + 4-year ECOA Reg B record-keeping + 3-year EEOC + 7-year SOX + GDPR Article 30 + EU AI Act Article 12 + SOC 2 CC7/CC8.

What does this skill connect to on the lead-scoring-routing agent and across the swarm?

On the lead-scoring-routing agent: lead-scoring-routing (parent commercial pillar) + firmographic enrichment + lead routing (#561 sibling build-pillar — UPSTREAM SOURCE of enrichment + identity-resolution feeding BANT score) + multi-source attribution-preserving lead ingestion (#557 sibling build-pillar — upstream lead source) + multi-location BANT lead-scoring (sibling commercial). Across the swarm: governance-decision-router five-destination routing for borderline cases + per-jurisdiction compliance multi-state franchise (sibling build-pillar) + master-record + per-location dynamic content + cs-agent-assist + chat-deflection compliance. Build-pillar siblings: tiered pre-filter deterministic gates for AI content compliance + marketing AI autonomy profile configuration + per-platform compliance gating for social posts (#564 same EU AI Act Article 50 substrate) + per-vertical compliance overlay + per-vertical pre-built compliance overlay templates. Commercial-pillar parent: /lead-scoring-routing.

What does the 6-workstream pre-engagement-baseline reporting cycle look like for this skill?

Every two weeks during the Tier 3 Fractional CMO with AI Swarm engagement, six workstreams report against the pre-engagement baseline. Workstream 1: signal-layer coverage — per-portfolio per-vertical per-layer (firmographic + intent + conversation + second-party + first-party) volume + source attestation + DPA chain. Workstream 2: Compute model distribution — per-vertical model selection + AutoML provenance + feature-importance per protected-class proxy + vector-similarity intent classification distribution. Workstream 3: Calibration metric distribution — per-vertical Brier + ECE + reliability diagram + Rosenbaum bounds + E-value + multiple-comparisons correction applied + pre-registered threshold compliance + 4/5ths rule disparate-impact audit results. Workstream 4: Gate-pass/gate-fail distribution — per-anchor gate-fail rate + per-vertical breakdown + adverse-action-notice issued count + Fundamental Rights Impact Assessment coverage + AEDT bias-audit coverage. Workstream 5: Regulatory-defense audit coverage — ECOA Reg B + 4/5ths rule + Title VII + ADEA + ADA + FCRA 604/615 + EEOC AI Guidance 2024 + NYC Local Law 144 + EU AI Act Article 22 + Annex III + CFPB Circular 2022-03 + SOX 302/404/906 + FASB ASC 606 + FDD Item 12. Workstream 6: FBC feedback-loop pattern-learning — per-vertical realized-vs-predicted reconciliation + multi-arm-bandit regret + recalibration + per-protected-class disparate-impact retrospective + per-jurisdiction enforcement-update impact.

Build pillar · lead-scoring-routing agent

How to build buyer-state-aware BANT scoring

Salesforce Einstein Lead Scoring + HubSpot Predictive + Marketo Predictive + Pardot Einstein + Pipedrive AI + Zoho Zia + Microsoft Dynamics 365 Sales Insights + 6sense + Demandbase + Bombora ship per-tenant flat lead-score primitives. The Compute + Calibrate + Gate + Audit skill bundle on the lead-scoring-routing agent sits above the per-vendor BANT-score substrate (downstream consumer of #561 firmographic enrichment) and writes a per-lead canonical record with named regulatory anchors covering ECOA Regulation B + 4/5ths rule + Title VII Bostock + ADEA + ADA + FCRA Section 604/615 + EEOC AI Guidance 2024 + NYC Local Law 144 AEDT + EU AI Act Article 22 + Annex III + CFPB Circular 2022-03.

Published October 7, 2026 · 3,200 words

The 4-skill bundle on the lead-scoring-routing agent

One agent. Four coordinated skills. The Compute + Calibrate + Gate + Audit bundle runs above the per-vendor BANT-score substrate (Salesforce Einstein + HubSpot Predictive + Marketo Predictive + Pardot + Pipedrive AI + Zoho Zia + Dynamics 365) + intent signal (6sense + Demandbase + Bombora + Madison Logic) + conversation- intelligence (Gong + Chorus + Outreach + Salesloft) and writes one canonical per-lead record.

Compute

5 signal layers: firmographic enrichment (from #561 sibling skill) + intent (6sense + Demandbase + Bombora + Madison + G2 + TrustRadius) + conversation-intelligence (Gong + Chorus + Outreach + Salesloft) + second-party engagement (LinkedIn Sales Navigator) + first-party engagement (GA4 + Adobe + Mixpanel + Amplitude). ML modeling: scikit-learn + XGBoost + LightGBM + CatBoost + AutoML + DataRobot. Vector-similarity intent classification.

Calibrate

Replication-crisis discipline: Brier score + Expected Calibration Error (ECE) + reliability diagram + isotonic / Platt recalibration + Rosenbaum bounds for unmeasured- confounder sensitivity + E-value + multiple-comparisons correction (Bonferroni + Benjamini-Hochberg FDR) + pre- registration of model thresholds + holdout-set replication + temporal cross-validation + 4/5ths rule disparate-impact audit per protected-class proxy.

Gate

5 anchors per-lead before score-driven routing commits. ECOA Reg B + 4/5ths rule + Title VII Bostock + ADEA + ADA + Fair Housing + GINA + EEOC AI Guidance 2024 + NYC Local Law 144 AEDT + IL HB 3773 + CA SB-942 + CO SB 21-169 + FCRA 604/615 + ECOA 1002.9 + CFPB Circular 2022-03 + EU AI Act Article 22 + Annex III + FRIA + SOX + FASB ASC 606 + FDD Item 12.

Audit

Per-lead WORM canonical record: signal-layer snapshot + ML model snapshot + calibration metrics + 4/5ths rule audit + per-anchor gate-pass + adverse-action-notice content + FRIA artifact + AEDT bias-audit + routing decision + revenue- accrual impact + replay capability. Retention: 7-year FTC + 7-year IRS + 6-year SEC + 3-year FINRA + 7-year FDD + 5-year FCRA + 4-year ECOA Reg B + 3-year EEOC + 7-year SOX + GDPR Article 30 + EU AI Act Article 12 + SOC 2 CC7/CC8.

The real ecosystem this sits above

Compute + Calibrate + Gate + Audit does not replace the lead- scoring vendors or the intent platforms. It sits above them, coordinates them, and writes one canonical per-lead record with named regulatory anchors.

Lead-scoring substrate

Salesforce Einstein + HubSpot Predictive + Marketo
Pardot Einstein + Pipedrive AI + Zoho Zia + Dynamics 365
Outreach Kaia + Salesloft + Drift + Conversica
ActiveCampaign Predictive Sending
scikit-learn + XGBoost + LightGBM + CatBoost + AutoML

Intent + signal source

6sense + Demandbase + Bombora + Madison Logic
RollWorks + Terminus + Foundry Intent + Cyance
Aberdeen + TechTarget Priority Engine + G2
TrustRadius + Capterra + Gartner + SoftwareReviews
LinkedIn Sales Navigator + Insight Tag + Lead Gen Forms

Conversation + first-party

Gong + Chorus + Outreach + Salesloft + Mindtickle
GA4 + Adobe Analytics + Mixpanel + Amplitude + Heap
Pendo + Posthog + Mutiny + Optimizely + VWO + Convert
Statsig + GrowthBook + Kameleoon + AB Tasty
Pinecone + Weaviate + Chroma + Qdrant + Milvus + Vespa

Compliance overlay

Five anchors run per-lead before any score-driven routing decision commits. The first anchor is operationally distinctive to BANT scoring: AI-ML disparate-impact under ECOA Regulation B + 4/5ths rule + Title VII Bostock intersects EU AI Act Article 22 high-risk classification + EEOC AI Guidance 2024 + NYC Local Law 144 AEDT.

Anchor 1: ECOA + 4/5ths rule + Title VII + ADEA + ADA + EEOC AI Guidance + NYC Local Law 144 AEDT (operationally distinctive)

ECOA Regulation B 12 CFR 1002 disparate-impact + 4/5ths rule per Uniform Guidelines on Employee Selection Procedures 1978. Title VII Civil Rights Act 1964 + Griggs v Duke Power Co + Ricci v DeStefano + Texas Dept of Housing v Inclusive Communities Project. Title VII Sex (Bostock v Clayton County 2020 + sex-orientation + gender-identity). ADEA Age Discrimination in Employment Act when employment-decisioning BANT routes leads with age proxy. ADA Title I when accommodation needed for BANT scoring. Pregnancy Discrimination Act. Genetic Information Nondiscrimination Act (GINA). Equal Pay Act. EEOC Enforcement Guidance on Software AI Algorithmic Decision-Making 2024. EEOC Uniform Guidelines 1978. NYC Local Law 144 automated employment decision tools (AEDT) bias audit. Illinois HB 3773 Artificial Intelligence Video Interview Act. CA SB-942 + CO SB 21-169 algorithmic discrimination + NY AI bill. Fair Housing Act 42 USC 3601 when housing-decisioning.

Anchor 2: FCRA + ECOA + CFPB adverse-action notification

FCRA 15 USC 1681 Section 604 prescreening permissible- purpose + Section 615(d) firm-offer-of-credit + Section 615(e) opt-out + Section 1681i adverse-action notice. ECOA adverse-action notice + ECOA 12 CFR 1002.9. CFPB Circular 2022-03 adverse-action notification for AI-ML credit decisions + CFPB ECOA examination procedures + CFPB UDAAP. GLBA Safeguards Rule 16 CFR Part 314 (amended 2023).

Anchor 3: EU AI Act Article 22 + Annex III high-risk + FRIA

EU AI Act Article 22 transparency for automated decisions + Article 26 deployer obligations + Article 50 + Article 13 + 14 + 15 + Annex III high-risk classification (employment + credit + housing + education) when AI-ML BANT score routes leads with material consequence + Article 6 + 27 Fundamental Rights Impact Assessment. Digital Services Act + DMA. GDPR Article 22 right not to be subject to automated decision + Article 6 + 7 + 17 + Article 28 + 30. CCPA + CPRA right to opt out of automated decision-making + right to know + right to correct + 18-state comprehensive privacy. LGPD + DPDP + PIPEDA + Quebec Law 25 + COPPA.

Anchor 4: SOX + FASB + SEC + franchise

Sarbanes-Oxley 302/404/906 when public-company AI-ML BANT score drives Forecast disclosure. FASB ASC 606 revenue recognition when BANT score informs revenue accrual. FASB ASC 842 lease. FASB ASC 326 CECL when BANT informs credit- loss estimate. SEC Reg S-K Item 303 MD&A. SEC Form 10-K segment. FDD Item 12 territorial-protection per FTC Franchise Rule 16 CFR 436 when franchise BANT score crosses ADI/protected-territory. 15-state franchise registration + state UDTPA + Robinson-Patman + FDD Item 17 + 19.

Anchor 5: Replication-crisis discipline + AI governance

Brier score + Expected Calibration Error + reliability diagram + Rosenbaum bounds + E-value + sensitivity analysis + multiple-comparisons correction (Bonferroni + Benjamini- Hochberg FDR) + pre-registration of model thresholds + 4/5ths rule audit. FTC Act Section 5 + FTC Endorsement Guides when third-party-review intent signal + FTC Fake Review Rule + Lanham. NIST AI Risk Management Framework. ISO 42001. ISO 27001. SOC 2 Type II. Per-vendor LLM zero-retention verified per call. Per-source DPA. Per-API rate-limit. Policy-as-code via OPA Rego + AWS Cedar + Casbin + Cerbos + Oso + Styra DAS + Permit.io. Storage: AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi WORM.

6-workstream reporting cycle

Every two weeks during a Tier 3 Fractional CMO engagement, six workstreams report against the pre-engagement baseline. No forecast accuracy claims. Process commitments only.

1. Signal-layer coverage. Per-portfolio per-vertical per-layer (firmographic + intent + conversation + second-party + first-party) volume + source attestation + DPA chain.
2. Compute model distribution. Per-vertical model selection + AutoML provenance + feature- importance per protected-class proxy + vector-similarity intent classification distribution.
3. Calibration metric distribution. Per-vertical Brier + ECE + reliability diagram + Rosenbaum bounds + E-value + multiple-comparisons correction applied + pre-registered threshold compliance + 4/5ths rule audit results.
4. Gate-pass/gate-fail distribution. Per-anchor gate-fail rate + per-vertical breakdown + adverse- action-notice issued count + FRIA coverage + AEDT bias-audit coverage.
5. Regulatory-defense audit coverage. ECOA Reg B + 4/5ths rule + Title VII + ADEA + ADA + FCRA 604/615 + EEOC AI Guidance 2024 + NYC Local Law 144 + EU AI Act Article 22 + Annex III + CFPB Circular 2022-03 + SOX + FASB ASC 606 + FDD Item 12.
6. FBC feedback-loop pattern-learning. Per-vertical realized-vs-predicted reconciliation + multi-arm- bandit regret + recalibration + per-protected-class disparate- impact retrospective + per-jurisdiction enforcement-update impact.

FAQ

What is buyer-state-aware BANT scoring — and what is the AI-ML-disparate-impact-times-EU-AI-Act-Article-22 double-violation problem?: A DTC ecommerce + B2B SaaS + multi-vertical operator routes inbound leads to sales reps based on BANT scoring (Budget + Authority + Need + Timeline) augmented with firmographic enrichment (downstream of #561 sibling skill) + intent signal (6sense + Demandbase + Bombora + Madison Logic + RollWorks + Terminus + Foundry Intent + G2 + TrustRadius) + conversation-intelligence (Gong + Chorus + Outreach + Salesloft) + first-party engagement (GA4 + Adobe + Mixpanel + Amplitude + Heap + Pendo + Posthog). The four-skill bundle on the lead-scoring-routing agent — Compute, Calibrate, Gate, Audit — sits above the per-vendor lead-score substrate (Salesforce Einstein + HubSpot Predictive + Marketo Predictive + Pardot + Pipedrive AI + Zoho Zia + Dynamics 365 Sales Insights + Outreach Kaia + Salesloft Drift + Conversica) and writes a per-lead canonical record. The operationally distinctive anchor: ECOA Regulation B 12 CFR 1002 disparate-impact + 4/5ths rule per Uniform Guidelines on Employee Selection Procedures 1978 when AI-ML BANT score routes leads through proxies (industry + company size + location + firmographic correlate + tech stack + buying committee composition) that correlate with protected class. EU AI Act Article 22 + Annex III classifies AI-ML scoring driving employment + credit + housing + education access as high-risk. EEOC Enforcement Guidance on Software AI Algorithmic Decision-Making 2024 + NYC Local Law 144 AEDT bias audit + CFPB Circular 2022-03 adverse-action notification for AI-ML credit decisions all converge on the same scoring surface.
Why do Salesforce Einstein Lead Scoring + HubSpot Predictive + Marketo Predictive + 6sense + Demandbase break at AI-ML-disparate-impact compliance scale?: Each lead-scoring vendor ships a per-tenant flat score primitive — a Marketing Ops team enables the feature, the platform produces a 0-100 score per lead, sales routes leads above threshold. None calibrates the model with replication-crisis discipline (Brier score + Expected Calibration Error + reliability diagram + Rosenbaum bounds + E-value + sensitivity analysis + multiple-comparisons correction + pre-registration of model thresholds + 4/5ths rule audit). None gates the score against ECOA Regulation B disparate-impact + Title VII Bostock + ADEA + ADA + FCRA Section 604 prescreening + EEOC AI Guidance 2024 + NYC Local Law 144 AEDT bias audit + EU AI Act Article 22. None coordinates adverse-action notification when score drives material consequence (FCRA Section 615 + ECOA 12 CFR 1002.9 + CFPB Circular 2022-03). None writes a per-lead audit trail with regulatory-defense retention. None enforces FDD Item 12 territorial-protection when franchise BANT score crosses ADI/protected-territory. None enforces Sarbanes-Oxley 302/404/906 + FASB ASC 606 revenue-recognition when BANT score informs revenue accrual. The four-skill bundle Compute + Calibrate + Gate + Audit sits above the lead-scoring-vendor surface — it does not replace it.
How does Compute + Calibrate work across BANT scoring with replication-crisis discipline?: Compute runs per-portfolio per-vertical per-lead buyer-state-aware BANT scoring across 5 signal layers: (1) firmographic enrichment from #561 sibling skill (Apollo + ZoomInfo + Clearbit + Cognism + Lusha + LeadIQ + DiscoverOrg + People Data Labs + Coresignal + UpLead); (2) intent signal (6sense + Demandbase + Bombora Company Surge + Madison Logic + RollWorks + Terminus + Foundry Intent + Cyance + Aberdeen + TechTarget Priority Engine + G2 + TrustRadius + Capterra + Gartner Digital Markets + SoftwareReviews + IT Central Station); (3) conversation-intelligence (Gong + Chorus + Outreach + Salesloft + Mindtickle + Avoma); (4) second-party engagement (LinkedIn Sales Navigator + LinkedIn Insight Tag + Lead Gen Forms + Smart Links); (5) first-party engagement (GA4 + Adobe Analytics + Mixpanel + Amplitude + Heap + Pendo + Posthog + Mutiny + Optimizely + VWO + Statsig + GrowthBook). ML modeling: scikit-learn + XGBoost + LightGBM + CatBoost + TensorFlow + PyTorch + AutoML + DataRobot + H2O.ai + Alteryx + KNIME + Dataiku. Vector-similarity for intent classification: Algolia + Elasticsearch + Pinecone + Weaviate + Chroma + Qdrant + Milvus + Vespa + Marqo. Per-vendor LLM tie-breakers (OpenAI + Anthropic + Mistral + Llama) under zero-retention. Calibrate runs replication-crisis discipline: Brier score + Expected Calibration Error (ECE) + reliability diagram + isotonic regression / Platt scaling recalibration + Rosenbaum bounds for unmeasured-confounder sensitivity + E-value + multiple-comparisons correction (Bonferroni + Benjamini-Hochberg false-discovery-rate) + pre-registration of model thresholds + holdout-set replication + temporal cross-validation against drift + 4/5ths rule disparate-impact audit per protected-class proxy.
What does Gate + Audit do?: Gate runs 5 anchors per-lead before any score-driven routing decision commits. (1) ECOA Regulation B 12 CFR 1002 disparate-impact + 4/5ths rule per Uniform Guidelines on Employee Selection Procedures 1978 + Title VII Bostock + ADEA + ADA + Fair Housing Act + GINA + Equal Pay Act + EEOC Enforcement Guidance on Software AI Algorithmic Decision-Making 2024 + NYC Local Law 144 AEDT bias audit + Illinois HB 3773 AI Video Interview Act + CA SB-942 + CO SB 21-169 algorithmic discrimination + NY AI bill. (2) FCRA 15 USC 1681 Section 604 prescreening permissible-purpose + Section 615(d) firm-offer-of-credit + Section 615(e) opt-out + Section 1681i adverse-action notice + ECOA adverse-action notice + CFPB Circular 2022-03 adverse-action notification for AI-ML credit decisions + ECOA 12 CFR 1002.9 + GLBA Safeguards Rule. (3) EU AI Act Article 22 + Article 26 + Article 50 + Article 13 + 14 + 15 + Annex III high-risk + Article 6 + 27 Fundamental Rights Impact Assessment + Digital Services Act + DMA. (4) Sarbanes-Oxley 302/404/906 when public-company + FASB ASC 606 revenue-recognition + FASB ASC 842 + FASB ASC 326 CECL + SEC Reg S-K Item 303 MD&A + SEC Form 10-K segment. (5) FDD Item 12 territorial-protection + 15-state franchise registration + state UDTPA + Robinson-Patman + FDD Item 17 + 19 + FTC Act Section 5 + FTC Endorsement Guides when third-party-review intent signal + FTC Fake Review Rule + CFPB UDAAP + Lanham Act. Audit writes a per-lead WORM canonical record: signal layer snapshot + ML model snapshot + calibration metrics (Brier + ECE + reliability diagram + Rosenbaum bounds + E-value) + 4/5ths rule audit snapshot + per-anchor gate-pass + adverse-action-notice content + Fundamental Rights Impact Assessment artifact + AEDT bias-audit artifact + routing decision + revenue-accrual impact + per-lead replay capability. Storage: AWS S3 Object Lock + Azure Blob immutable + Google Cloud Storage Bucket Lock + Wasabi WORM. Retention stacks: 7-year FTC + 7-year IRS + 6-year SEC + 3-year FINRA + 7-year FDD + 5-year FCRA Section 615(d) firm-offer + 4-year ECOA Reg B record-keeping + 3-year EEOC + 7-year SOX + GDPR Article 30 + EU AI Act Article 12 + SOC 2 CC7/CC8.
What does this skill connect to on the lead-scoring-routing agent and across the swarm?: On the lead-scoring-routing agent: lead-scoring-routing (parent commercial pillar) + firmographic enrichment + lead routing (#561 sibling build-pillar — UPSTREAM SOURCE of enrichment + identity-resolution feeding BANT score) + multi-source attribution-preserving lead ingestion (#557 sibling build-pillar — upstream lead source) + multi-location BANT lead-scoring (sibling commercial). Across the swarm: governance-decision-router five-destination routing for borderline cases + per-jurisdiction compliance multi-state franchise (sibling build-pillar) + master-record + per-location dynamic content + cs-agent-assist + chat-deflection compliance. Build-pillar siblings: tiered pre-filter deterministic gates for AI content compliance + marketing AI autonomy profile configuration + per-platform compliance gating for social posts (#564 same EU AI Act Article 50 substrate) + per-vertical compliance overlay + per-vertical pre-built compliance overlay templates. Commercial-pillar parent: /lead-scoring-routing.
What does the 6-workstream pre-engagement-baseline reporting cycle look like for this skill?: Every two weeks during the Tier 3 Fractional CMO with AI Swarm engagement, six workstreams report against the pre-engagement baseline. Workstream 1: signal-layer coverage — per-portfolio per-vertical per-layer (firmographic + intent + conversation + second-party + first-party) volume + source attestation + DPA chain. Workstream 2: Compute model distribution — per-vertical model selection + AutoML provenance + feature-importance per protected-class proxy + vector-similarity intent classification distribution. Workstream 3: Calibration metric distribution — per-vertical Brier + ECE + reliability diagram + Rosenbaum bounds + E-value + multiple-comparisons correction applied + pre-registered threshold compliance + 4/5ths rule disparate-impact audit results. Workstream 4: Gate-pass/gate-fail distribution — per-anchor gate-fail rate + per-vertical breakdown + adverse-action-notice issued count + Fundamental Rights Impact Assessment coverage + AEDT bias-audit coverage. Workstream 5: Regulatory-defense audit coverage — ECOA Reg B + 4/5ths rule + Title VII + ADEA + ADA + FCRA 604/615 + EEOC AI Guidance 2024 + NYC Local Law 144 + EU AI Act Article 22 + Annex III + CFPB Circular 2022-03 + SOX 302/404/906 + FASB ASC 606 + FDD Item 12. Workstream 6: FBC feedback-loop pattern-learning — per-vertical realized-vs-predicted reconciliation + multi-arm-bandit regret + recalibration + per-protected-class disparate-impact retrospective + per-jurisdiction enforcement-update impact.

Engage Completions

Two ways to engage. The Tier 1 AI Readiness Assessment maps the lead-scoring substrate + intent platforms + per-vertical disparate-impact surface against the Compute + Calibrate + Gate + Audit bundle. The Tier 3 Fractional CMO with AI Swarm embeds 1-2 days per week for 6+ months and runs the bundle end-to-end against the lead-scoring-routing agent across the swarm.

Tier 1 AI Readiness Assessment Tier 3 Fractional CMO with AI Swarm Not sure which tier? Take the 3-question quiz