ING Bank

ING Bank, one of Europe's largest retail banking groups, faced mounting operational and regulatory pressure around Know Your Customer (KYC) and customer due diligence (CDD) requirements. The existing process required clients to respond to roughly 100 discrete questions and submit extensive supporting documentation — a burden repeated not just at onboarding but during periodic reviews of existing clients. Processing times ranged from a single day in straightforward cases to several weeks when follow-up information was needed, creating friction across the customer journey and tying compliance staff to low-value data-gathering work rather than substantive risk assessment.

ING deployed a machine learning model that automatically answers 70–80 of the 100 questions in a standard customer due diligence review by drawing on publicly available data sources and behavioral signals — including how clients conduct their banking activity with ING and with other institutions. The model is applied uniformly across both new client onboarding and scheduled periodic reviews of existing clients, eliminating repeated outreach for information the bank can already infer or source externally. To maintain governance discipline, ING centralized all AI exploration under the COO with strict controls to prevent ungoverned proliferation across business units — a deliberate structural decision that shaped how the capability was scoped, tested, and scaled across the organization.

The impact on processing speed was transformational: customer due diligence reviews that previously took days or weeks now complete in seconds. Key outcomes include:

• 70–80% of KYC questions resolved automatically, without customer contact
• Processing time reduced from days or weeks → seconds
• 25% operations productivity gain — a figure ING's COO described as consistent across every process where AI has been introduced
• Compliance staff redeployed from data collection to genuine risk analysis
• ING projects that certain operations will require 25% fewer headcount, with those employees redirected toward growth-oriented and complex judgment tasks

• Compliance workflows often contain a high proportion of data-gathering steps that can be automated using existing internal and public data — ING's 70–80% automation rate illustrates the ceiling available before human judgment is actually required.
• Behavioral data already held by the institution is an underused asset; combining it with public sources can eliminate customer touchpoints without reducing regulatory rigor.
• Applying AI to both new onboarding and periodic reviews compounds efficiency gains — point solutions that address only one stage leave significant volume on the table.
• Centralizing AI governance under a single executive sponsor with explicit controls helps prevent fragmented, ungoverned deployments that create compliance and reputational risk.
• A consistent 25% productivity uplift across multiple operations processes suggests the gains are structural, not one-off — making the business case for each successive use case easier to justify.