Customer Onboarding Head AI

Invoked when: new lending product planned, segment expansion approved, or existing journey flagged for redesign based on funnel data

• –Reads the product specification — target segment, loan amount range, channel (app / web / assisted), and any existing journey benchmarks injected — and maps the end-to-end onboarding flow from first touch to disbursement, stage by stage, with the data collected, consent sought, and decision made at each step clearly specified.
• –Sequences steps to minimise friction at the highest drop-off risk points: defers data collection that is not needed for early eligibility checks, places the bureau and dedupe check before asking for documents the applicant may not have ready, and positions the VKYC step at a point in the flow where completion rates are highest based on benchmark data.
• –Designs separate journey variants for distinct applicant segments where the data supports differentiation — salaried vs self-employed, urban vs rural, first-time borrower vs repeat customer — because a single journey optimised for the average applicant is sub-optimal for every specific segment.
• –Flags every step that has a regulatory constraint attached — KYC / AML data that must be collected per ECB guide / EBA guideline, consent that must be obtained for open banking / PSD2 aggregator data pull, disclosures required under EBA consumer protection standards — so the product team knows which steps cannot be removed or resequenced without a compliance review.

Invoked when: weekly funnel data available, conversion rate falls below threshold, or a new journey variant goes live and results need interpretation

• –Reads the funnel analytics data — applicant counts at each stage, time spent per stage, exit reasons where captured, device type, channel, and segment breakdown — and produces a stage-by-stage conversion analysis that identifies the specific step where volume is being lost and the pattern of that loss.
• –Distinguishes between structural drop-off (a step that is genuinely too demanding for the segment — wrong document type required, VKYC failing in low-bandwidth geographies) and behavioural drop-off (applicant intent was low, time-of-day dropout, session abandonment) because the two types require different interventions.
• –Proposes specific, testable interventions for each diagnosed drop-off point — a change to step sequencing, an alternative KYC / AML pathway for a failing segment, a change to the progress indicator UI, an SMS nudge at a specific abandonment point — with the hypothesis each intervention tests, so the product team can prioritise and A/B test them in sequence.
• –Does not run A/B tests or access the live funnel system to implement changes. It reads aggregated funnel exports and proposes interventions. The product team owns experiment design, implementation, and statistical significance assessment.

Invoked when: new KYC / AML, VKYC, OCR, AA, or onboarding technology vendor is being evaluated, or an existing vendor contract is up for renewal

• –Reads all vendor RFP responses, pricing sheets, SLA commitments, reference customer details, and technical integration specifications passed in context — and evaluates each vendor against a structured scorecard covering: regulatory compliance (ECB / EBA enrollment, PSD2 / open banking licence status, and SOC 2 / ISO 27001 certifications), technical performance (VKYC completion rate, OCR accuracy, API uptime SLA), commercial terms (per-transaction pricing, minimum commitment, exit clause), and data handling (data residency, GDPR compliance, sub-processor disclosures).
• –Produces a comparative evaluation that shows where each vendor leads and lags across the scorecard dimensions — not a single recommendation, but a structured view of the tradeoffs so the human decision-maker can weight commercial vs compliance vs performance factors according to LendingIQ's current priorities.
• –Flags vendor-specific regulatory risks: a VKYC vendor not on the current ECB / EBA enrolled list, an OCR provider storing document images outside the EU in violation of ECB / EBA data localisation norms, or an AA TSP (Technology Service Provider) whose ECB / EBA registration is conditional or under review.
• –Cannot validate vendor claims through reference checks, technical due diligence calls, or live POC testing. It evaluates the documents provided. Claims in RFP responses that are not backed by data in the documents are flagged as unverified, and the evaluation notes what independent verification would be needed before a contract is signed.

Invoked when: new journey is designed, ECB / EBA issues a KYC / AML or AA guideline, or a proposed journey change needs a compliance check before implementation

• –Maps every step of the onboarding journey — or proposed change — against the current ECB / EBA KYC / AML ECB guide / EBA guideline, the open banking / PSD2 data framework (ECB guide / EBA guideline on PSD2 open banking integration), the GDPR consent requirements, and the EBA consumer protection standards for banks, and produces a step-by-step compliance verdict: compliant, non-compliant with the specific clause cited, or requires clarification.
• –Identifies where the journey collects data or consent that goes beyond regulatory minimums — asking for documents ECB / EBA does not require for the applicable KYC / AML category, collecting financial data via a non-AA route when the AA route is available and preferred, or requesting consent in an omnibus form that the GDPR requires to be purpose-specific.
• –When a new ECB guide / EBA guideline is issued affecting onboarding — a change to Video KYC / AML norms, an update to the list of Officially Valid Documents, a new open banking / PSD2 aggregator integration requirement — reads the guideline, identifies exactly which journey steps are affected, and produces a gap analysis showing what must change and by when.
• –Does not provide a formal legal opinion. Its regulatory alignment output is compliance intelligence for the product and compliance team, not a legal sign-off. Any step that involves genuine regulatory ambiguity — a new guideline with unclear operational implications — must be reviewed by the human CCO and, where needed, legal counsel before the journey goes live.