On this page

Cross-border identity verification fails in a predictable place. The ID check is rarely the problem. The trouble starts when a business customer has to prove who it is, who owns it, and who speaks for it in a country whose registry, documents, and ownership rules differ from the last one. Teams then stitch together document tools, screening vendors, registry lookups, and email threads, and analysts absorb the gap.

This case study looks at what happened when two European companies moved that work onto Duna, an AI-native business identity platform. SeQura, a buy-now-pay-later provider, ran merchant KYB on it. bol, the Dutch-Belgian marketplace, ran customer due diligence for its partners. Both programs are documented in first-party Duna case studies, and every figure below comes from those pages.

The problem: cross-border onboarding built from separate tools

Software that automates identity verification for cross-border operations has to solve more than document capture. Commonly cited challenges include regulatory fragmentation, document diversity, and data fragmentation (Microblink's cross-border identity verification guide lists these alongside evolving fraud and the speed/security tradeoff). For business onboarding, the list is longer. A KYB case needs entity verification, ultimate beneficial owner (UBO) resolution, legal representative checks, sanctions and PEP screening, and a documented decision.

Before Duna, SeQura's merchant review looked like many regulated onboarding programs. Evidence arrived through different channels, analysts validated documents by hand, and screening sat in a separate step. Review took an average of 243 minutes of analyst time per case. First-time right, meaning the share of onboarding cases a partner completes correctly without being sent back for fixes, sat at 50%. Half of all merchants had to return to the journey at least once.

Merchants were also doing a lot of work. SeQura's baseline active onboarding time was 27.5 minutes per merchant, a cost that falls on the customer's side of the conversion funnel.

bol faced a different version of the same problem. A marketplace has to know its sellers, and it can't ask for so much that good partners abandon the process. Its case study describes customer due diligence becoming a strategic priority, with the goal of building a trusted marketplace without making onboarding a burden.

The goals both teams set

The targets were consistent across both programs:

  • Cut analyst time per case without lowering review quality

  • Raise first-time right so fewer cases bounce between partner and analyst

  • Keep partner effort low during onboarding and re-verification

  • Make periodic reviews feasible at scale rather than a headcount problem

  • Keep a complete audit trail of evidence, decisions, and policy changes

The solution: a policy engine, AI agents, and one case view

Duna replaces the stack of point tools with a compliance infrastructure layer. Three components do the main work.

The policy engine turns KYB, KYC, and AML requirements into code. The rules decide what evidence a given business needs based on its country, entity type, ownership structure, and risk score. The engine runs these checks upstream, during the partner journey, instead of leaving them for an analyst to discover later. Duna's explainer on automated decisioning from first principles covers how this layer works.

AI agents gather and check the evidence. An onboarding agent collects documents and data before review starts. A due diligence agent builds the risk picture. A screening agent investigates alerts. Duna's homepage summarizes the division of labor in one line: agents collect the evidence, analysts make the judgment. A deeper look at where AI agents fit across the compliance lifecycle shows how each agent hands off to the next.

Automated case management puts everything an analyst needs in one view: registry data, uploaded documents, screening results, ownership structure, and the policy outcome. Four-eye reviews trigger on risk rules. Partner communication runs through in-platform messaging, so questions and answers sit on the case record rather than in an inbox.

How multi-jurisdiction onboarding works in practice

The cross-border logic follows a fixed order.

  1. Registry prefill first. Duna connects to 210+ local registries, so company data such as legal name, registration number, address, and officers can be pulled in before the partner types anything.

  2. Policy-driven collection second. The partner is asked only for what the registry couldn't supply and what the policy requires for that country and risk tier. The interface adapts dynamically, and it's available in 7+ languages.

  3. UBO resolution inside the case. Ownership data lives in the case data model, with access to UBO registers where they exist. Ownership structure feeds the risk tier, which in turn changes the evidence required.

  4. Analyst judgment last. Cases that clear policy can be accepted automatically. The rest reach an analyst with evidence already assembled.

The effect is that country-specific complexity gets encoded once, as policy, rather than rebuilt as manual procedure for each market. Adding a jurisdiction becomes a rules question instead of a tooling project.

Integration and resilience

Duna's data platform connects KYC providers, registries, sanctions lists, and UBO sources through a single API. Smart routing selects providers by performance, cost, and coverage, and fallback procedures take over when a source fails or returns nothing. For a partner in the middle of onboarding, that means a provider outage doesn't turn into a stalled application.

One boundary is worth stating plainly. Duna is built around business identity: entities, owners, and representatives. Where a program also needs biometric or document checks on individuals, such as a director's ID, Duna orchestrates specialist KYC providers through the same policy and case layer rather than replacing them.

The results: measured before and after

SeQura: 16.3x faster review and 76% higher first-time right

Metric

Before

On Duna

Average analyst time per KYB case

243 minutes

14.9 minutes

First-time right

50%

88%

Approval rate

n/a

95.2%

Review time fell 16.3x. First-time right rose 38 percentage points, which is the 76% relative increase in the case study's headline. The 95.2% approval rate matters for a specific reason: it shows that the cases reaching analysts are mostly clean, so faster review isn't coming from rubber-stamping, and good merchants aren't being turned away.

The reboarding project shows what the time savings mean at scale. SeQura had a periodic review queue that, at 243 minutes per case, would have consumed nearly 7,000 analyst-hours. On Duna the same queue took roughly 420 analyst-hours, about 6% of the old workload. By our arithmetic, that is a queue of around 1,700 cases handled in the time the old process would have needed for a small fraction of them. Periodic review stops being a project that competes with new-merchant onboarding for the same analysts.

bol: 5.4-minute reviews and an 83/100 partner effort score

bol's customer due diligence program shows the same mechanics from the partner's side:

  • Analysts review cases in an average of 5.4 minutes

  • Partners spend an average of 12.2 minutes actively completing onboarding requirements

  • 81% of partner cases close on first review

  • The post-onboarding partner effort score is 83 out of 100

  • Periodic re-verification across around 40,000 partners averages 11 minutes of active partner time per case

At roughly 40,000 partners, re-verification is where manual processes break. Eleven minutes per partner, with the system pre-filling what it already knows, is what makes an annual or risk-triggered cycle workable without hiring against it.

Duna's own aggregate figures point the same direction: 10.6x faster onboarding, a 37% conversion increase, and 4.8x analyst efficiency across its customer base, as reported by Duna.

Why the gains came from the workflow, not from faster clicking

Two mechanisms account for most of the movement.

Upstream checks cut round-trips. First-time right is a function of how much gets validated while the partner is still in the journey. When the policy engine knows a Dutch BV with a corporate shareholder needs ownership documents and a sole proprietorship doesn't, it asks for the right things once. A partner who's told what's missing at submission doesn't wait days for an analyst to say so. Moving SeQura from 50% to 88% first-time right removed most of the remediation cycles that had inflated both partner effort and analyst time.

Consolidated evidence cuts review time. Analysts at both companies stopped hunting across systems. When registry extracts, documents, screening hits, and UBO structure sit on one case, a decision takes minutes. The 14.9-minute and 5.4-minute averages reflect that.

False positives follow a similar logic. Better matching and filtering reduce screening noise before it reaches a person, a principle Duna details in its piece on AI in KYC and AML. The case studies don't publish a standalone false-positive rate, so the evidence here is indirect: high first-time right and a 95.2% approval rate indicate that few cases are being escalated or reworked unnecessarily.

Cross-border edge cases the system had to handle

Document diversity. Incorporation certificates, extracts, and proof-of-address documents look different in every country. AI agents extract and normalize fields from these documents, and registry prefill reduces how often a document is needed at all. The policy decides which document types satisfy a requirement in each jurisdiction.

Ownership complexity. Layered holding structures are the classic source of manual escalation. Because UBO resolution sits in the case model, an unresolved owner above the threshold becomes a defined policy state rather than an analyst's open question. For readers comparing approaches, Duna's overview of UBO mapping at scale explains how AI agents, graph tools, and KYB automation differ.

Enhanced due diligence. Higher-risk entities trigger additional evidence and four-eye review through the same rules, so escalation is consistent and documented instead of depending on who picked up the case.

Periodic review and re-KYC. Policies schedule re-verification and prefill what hasn't changed. The SeQura and bol figures above come from exactly this use.

Missing or poor data. When a registry returns nothing or a document is unreadable, fallback routing tries another source, and the partner gets an automated request for only the missing item, with reminders. Every step is logged.

What the evidence says about stakeholder experience

Both case studies include perspectives from the customers' own teams, and the full quotations are on the bol and SeQura pages. The numbers carry the same message. For analysts, review shrank from hours to minutes. For partners, bol's 83/100 effort score and 12.2 active minutes show that compliance requirements didn't have to feel like friction.

Lessons for teams automating cross-border KYB

  1. Encode policy before you automate collection. Automation is only as good as the rules telling it what to ask for. Both programs started from the evidence requirements per country and risk tier.

  2. Measure first-time right, not just speed. Review time can improve while partners still loop through remediation. First-time right captures the full cost of a bad request.

  3. Prefill from registries wherever they exist. Every field a partner doesn't have to type is a field that can't be mistyped.

  4. Plan for reboarding on day one. SeQura's 420 versus roughly 7,000 analyst-hours shows that lifecycle review is where manual cost accumulates.

  5. Budget for provider failure. Redundancy and smart routing keep coverage steady when a single data source degrades.

  6. Keep humans on judgment. The 95.2% approval rate works because analysts handle the cases that need them, with evidence already assembled.

Teams comparing vendors can use Duna's guide to the best KYB software in 2026 as a starting framework. Criteria worth testing in any evaluation: registry coverage in your actual markets, UBO depth, how the policy layer is configured and audited, and whether results are published as before-and-after numbers.

Next steps

If your onboarding program spans several countries and your analysts still assemble evidence by hand, the most useful first exercise is to map your current per-jurisdiction requirements and measure two numbers: average analyst minutes per case and first-time right. Those are the baselines SeQura and bol started from.

Duna's team will walk through your markets, entity types, and review volumes and show how the policy engine, registry prefill, and AI agents would handle them. You can request a demo at duna.com.

Cross-border identity verification fails in a predictable place. The ID check is rarely the problem. The trouble starts when a business customer has to prove who it is, who owns it, and who speaks for it in a country whose registry, documents, and ownership rules differ from the last one. Teams then stitch together document tools, screening vendors, registry lookups, and email threads, and analysts absorb the gap.

This case study looks at what happened when two European companies moved that work onto Duna, an AI-native business identity platform. SeQura, a buy-now-pay-later provider, ran merchant KYB on it. bol, the Dutch-Belgian marketplace, ran customer due diligence for its partners. Both programs are documented in first-party Duna case studies, and every figure below comes from those pages.

The problem: cross-border onboarding built from separate tools

Software that automates identity verification for cross-border operations has to solve more than document capture. Commonly cited challenges include regulatory fragmentation, document diversity, and data fragmentation (Microblink's cross-border identity verification guide lists these alongside evolving fraud and the speed/security tradeoff). For business onboarding, the list is longer. A KYB case needs entity verification, ultimate beneficial owner (UBO) resolution, legal representative checks, sanctions and PEP screening, and a documented decision.

Before Duna, SeQura's merchant review looked like many regulated onboarding programs. Evidence arrived through different channels, analysts validated documents by hand, and screening sat in a separate step. Review took an average of 243 minutes of analyst time per case. First-time right, meaning the share of onboarding cases a partner completes correctly without being sent back for fixes, sat at 50%. Half of all merchants had to return to the journey at least once.

Merchants were also doing a lot of work. SeQura's baseline active onboarding time was 27.5 minutes per merchant, a cost that falls on the customer's side of the conversion funnel.

bol faced a different version of the same problem. A marketplace has to know its sellers, and it can't ask for so much that good partners abandon the process. Its case study describes customer due diligence becoming a strategic priority, with the goal of building a trusted marketplace without making onboarding a burden.

The goals both teams set

The targets were consistent across both programs:

  • Cut analyst time per case without lowering review quality

  • Raise first-time right so fewer cases bounce between partner and analyst

  • Keep partner effort low during onboarding and re-verification

  • Make periodic reviews feasible at scale rather than a headcount problem

  • Keep a complete audit trail of evidence, decisions, and policy changes

The solution: a policy engine, AI agents, and one case view

Duna replaces the stack of point tools with a compliance infrastructure layer. Three components do the main work.

The policy engine turns KYB, KYC, and AML requirements into code. The rules decide what evidence a given business needs based on its country, entity type, ownership structure, and risk score. The engine runs these checks upstream, during the partner journey, instead of leaving them for an analyst to discover later. Duna's explainer on automated decisioning from first principles covers how this layer works.

AI agents gather and check the evidence. An onboarding agent collects documents and data before review starts. A due diligence agent builds the risk picture. A screening agent investigates alerts. Duna's homepage summarizes the division of labor in one line: agents collect the evidence, analysts make the judgment. A deeper look at where AI agents fit across the compliance lifecycle shows how each agent hands off to the next.

Automated case management puts everything an analyst needs in one view: registry data, uploaded documents, screening results, ownership structure, and the policy outcome. Four-eye reviews trigger on risk rules. Partner communication runs through in-platform messaging, so questions and answers sit on the case record rather than in an inbox.

How multi-jurisdiction onboarding works in practice

The cross-border logic follows a fixed order.

  1. Registry prefill first. Duna connects to 210+ local registries, so company data such as legal name, registration number, address, and officers can be pulled in before the partner types anything.

  2. Policy-driven collection second. The partner is asked only for what the registry couldn't supply and what the policy requires for that country and risk tier. The interface adapts dynamically, and it's available in 7+ languages.

  3. UBO resolution inside the case. Ownership data lives in the case data model, with access to UBO registers where they exist. Ownership structure feeds the risk tier, which in turn changes the evidence required.

  4. Analyst judgment last. Cases that clear policy can be accepted automatically. The rest reach an analyst with evidence already assembled.

The effect is that country-specific complexity gets encoded once, as policy, rather than rebuilt as manual procedure for each market. Adding a jurisdiction becomes a rules question instead of a tooling project.

Integration and resilience

Duna's data platform connects KYC providers, registries, sanctions lists, and UBO sources through a single API. Smart routing selects providers by performance, cost, and coverage, and fallback procedures take over when a source fails or returns nothing. For a partner in the middle of onboarding, that means a provider outage doesn't turn into a stalled application.

One boundary is worth stating plainly. Duna is built around business identity: entities, owners, and representatives. Where a program also needs biometric or document checks on individuals, such as a director's ID, Duna orchestrates specialist KYC providers through the same policy and case layer rather than replacing them.

The results: measured before and after

SeQura: 16.3x faster review and 76% higher first-time right

Metric

Before

On Duna

Average analyst time per KYB case

243 minutes

14.9 minutes

First-time right

50%

88%

Approval rate

n/a

95.2%

Review time fell 16.3x. First-time right rose 38 percentage points, which is the 76% relative increase in the case study's headline. The 95.2% approval rate matters for a specific reason: it shows that the cases reaching analysts are mostly clean, so faster review isn't coming from rubber-stamping, and good merchants aren't being turned away.

The reboarding project shows what the time savings mean at scale. SeQura had a periodic review queue that, at 243 minutes per case, would have consumed nearly 7,000 analyst-hours. On Duna the same queue took roughly 420 analyst-hours, about 6% of the old workload. By our arithmetic, that is a queue of around 1,700 cases handled in the time the old process would have needed for a small fraction of them. Periodic review stops being a project that competes with new-merchant onboarding for the same analysts.

bol: 5.4-minute reviews and an 83/100 partner effort score

bol's customer due diligence program shows the same mechanics from the partner's side:

  • Analysts review cases in an average of 5.4 minutes

  • Partners spend an average of 12.2 minutes actively completing onboarding requirements

  • 81% of partner cases close on first review

  • The post-onboarding partner effort score is 83 out of 100

  • Periodic re-verification across around 40,000 partners averages 11 minutes of active partner time per case

At roughly 40,000 partners, re-verification is where manual processes break. Eleven minutes per partner, with the system pre-filling what it already knows, is what makes an annual or risk-triggered cycle workable without hiring against it.

Duna's own aggregate figures point the same direction: 10.6x faster onboarding, a 37% conversion increase, and 4.8x analyst efficiency across its customer base, as reported by Duna.

Why the gains came from the workflow, not from faster clicking

Two mechanisms account for most of the movement.

Upstream checks cut round-trips. First-time right is a function of how much gets validated while the partner is still in the journey. When the policy engine knows a Dutch BV with a corporate shareholder needs ownership documents and a sole proprietorship doesn't, it asks for the right things once. A partner who's told what's missing at submission doesn't wait days for an analyst to say so. Moving SeQura from 50% to 88% first-time right removed most of the remediation cycles that had inflated both partner effort and analyst time.

Consolidated evidence cuts review time. Analysts at both companies stopped hunting across systems. When registry extracts, documents, screening hits, and UBO structure sit on one case, a decision takes minutes. The 14.9-minute and 5.4-minute averages reflect that.

False positives follow a similar logic. Better matching and filtering reduce screening noise before it reaches a person, a principle Duna details in its piece on AI in KYC and AML. The case studies don't publish a standalone false-positive rate, so the evidence here is indirect: high first-time right and a 95.2% approval rate indicate that few cases are being escalated or reworked unnecessarily.

Cross-border edge cases the system had to handle

Document diversity. Incorporation certificates, extracts, and proof-of-address documents look different in every country. AI agents extract and normalize fields from these documents, and registry prefill reduces how often a document is needed at all. The policy decides which document types satisfy a requirement in each jurisdiction.

Ownership complexity. Layered holding structures are the classic source of manual escalation. Because UBO resolution sits in the case model, an unresolved owner above the threshold becomes a defined policy state rather than an analyst's open question. For readers comparing approaches, Duna's overview of UBO mapping at scale explains how AI agents, graph tools, and KYB automation differ.

Enhanced due diligence. Higher-risk entities trigger additional evidence and four-eye review through the same rules, so escalation is consistent and documented instead of depending on who picked up the case.

Periodic review and re-KYC. Policies schedule re-verification and prefill what hasn't changed. The SeQura and bol figures above come from exactly this use.

Missing or poor data. When a registry returns nothing or a document is unreadable, fallback routing tries another source, and the partner gets an automated request for only the missing item, with reminders. Every step is logged.

What the evidence says about stakeholder experience

Both case studies include perspectives from the customers' own teams, and the full quotations are on the bol and SeQura pages. The numbers carry the same message. For analysts, review shrank from hours to minutes. For partners, bol's 83/100 effort score and 12.2 active minutes show that compliance requirements didn't have to feel like friction.

Lessons for teams automating cross-border KYB

  1. Encode policy before you automate collection. Automation is only as good as the rules telling it what to ask for. Both programs started from the evidence requirements per country and risk tier.

  2. Measure first-time right, not just speed. Review time can improve while partners still loop through remediation. First-time right captures the full cost of a bad request.

  3. Prefill from registries wherever they exist. Every field a partner doesn't have to type is a field that can't be mistyped.

  4. Plan for reboarding on day one. SeQura's 420 versus roughly 7,000 analyst-hours shows that lifecycle review is where manual cost accumulates.

  5. Budget for provider failure. Redundancy and smart routing keep coverage steady when a single data source degrades.

  6. Keep humans on judgment. The 95.2% approval rate works because analysts handle the cases that need them, with evidence already assembled.

Teams comparing vendors can use Duna's guide to the best KYB software in 2026 as a starting framework. Criteria worth testing in any evaluation: registry coverage in your actual markets, UBO depth, how the policy layer is configured and audited, and whether results are published as before-and-after numbers.

Next steps

If your onboarding program spans several countries and your analysts still assemble evidence by hand, the most useful first exercise is to map your current per-jurisdiction requirements and measure two numbers: average analyst minutes per case and first-time right. Those are the baselines SeQura and bol started from.

Duna's team will walk through your markets, entity types, and review volumes and show how the policy engine, registry prefill, and AI agents would handle them. You can request a demo at duna.com.