Bank KYB onboarding software with audit trails and case management

On this page
Banks run business customer onboarding on one of three things: a client lifecycle management (CLM) platform such as Fenergo or Pega, a commercial banking workflow suite such as nCino, Appian, or Appway, or a patchwork of CRM, document storage, and screening tools held together by spreadsheets and email. Audit trails and case management are available in all three. The difference is whether the evidence is produced by the process itself or reconstructed by people afterward.
This case study walks through a reference deployment of Duna for corporate KYB onboarding at a bank, from the audit problem to the controls to the metrics worth tracking. One note on method: this is an anonymized composite scenario built from Duna's product capabilities and Duna's reported aggregate outcomes. It doesn't name a bank and carries no invented customer quotes. Where bank-specific numbers belong, the article shows what to measure and how, so you can compare them against your own baseline.
What banks use for business onboarding, and where audit trails break
Vendor shortlists for commercial and corporate onboarding usually include Fenergo, Pega Client Lifecycle Management and KYC, Appian, nCino, Appway, Temenos, and Finastra, sometimes with Salesforce as the front end. Each can record who did what. Fenergo's public pages emphasize CLM, KYC, and compliance workflow, along with claims of faster onboarding and automation. Those pages say much less about integration architecture, deployment sequencing, or what an audit evidence package looks like in practice, which is the detail a compliance team needs for due diligence.
That gap matters because most audit failures aren't caused by missing logs. They come from evidence scattered across systems. A typical failure pattern in a patchwork setup:
The onboarding form lives in one tool and the documents in another.
Screening results are exported as PDFs and attached to a ticket.
The analyst's rationale sits in an email thread or a free-text notes field.
Policy changes are recorded in a wiki, not tied to the decisions they affected.
When an auditor asks why a specific company was approved in a specific month, someone has to rebuild the answer from five sources. For a deeper comparison of platform categories, see Duna's overview of KYB software for banks.
The reference scenario: a multi-country bank onboarding corporate clients
The scenario is a bank onboarding corporate and mid-market business customers across several European jurisdictions. Its scope is KYB onboarding: entity verification against local registries, ultimate beneficial owner (UBO) identification, legal representative verification, document collection, sanctions, PEP, and adverse media screening, and risk scoring. Periodic reviews follow after account opening.
The measurable challenge
Four problems define the starting point, and each can be baselined before any software is chosen:
Manual review load. Analysts open every case in full, even when registry data and documents are clean.
Slow time-to-decision. Missing documents trigger email loops, so a case waits on the customer more often than on the analyst.
Inconsistent evidence capture. Teams in different countries record rationale in different formats.
Audit friction. Preparing for an internal or regulatory review means assembling case files by hand.
The fourth problem is the one that justifies the project to a risk committee. Regulators don't ask whether a bank has a policy. They ask for the decision, the evidence behind it, the person who approved it, and the version of the policy in force at the time. If answering takes days of reconciliation, the process isn't audit-ready, whatever the logs technically contain.
Solution architecture: policy in code, evidence in one case file
The deployment centers on one principle: every onboarding step should write to the same record, and every decision should trace to a policy.
The modules in plain terms
Onboarding journeys. Customers complete a dynamic, no-code journey in Duna Onboard. Questions and document requests adapt to the entity type, country, and risk level, so only necessary data is collected. Registry prefills reduce re-keying.
Policy engine. The bank's KYB, KYC, and AML requirements are written as executable policies. The policy decides what data to collect, which checks to run, what risk score applies, and which cases can be accepted automatically versus routed to a person.
Automated case management. Every application becomes a case with a defined structure: entity data, UBO structure, documents, screening results, risk score, messages, and decision.
In-platform messaging. Requests for missing information go through the platform, not email, so customer communication sits inside the case file.
Lifecycle compliance. Daily screening for PEP, sanctions, and media, plus policy-driven re-KYC, run on the same record the onboarding created.
Audit trail. Interactions, decisions, and policy changes are logged in one place.
Why centralization beat a workflow-only patchwork
A workflow tool can route a task and stamp a time on it. What it can't do on its own is guarantee that the evidence attached to that task came from the right source, or that the decision followed the current policy. Duna's data platform orchestrates registry and screening providers, with smart routing and fallback when a provider is unavailable, and that provider output is stored with the case. Local depth (210+ registries, UBO registers, local VAT capabilities, 7+ languages) means analysts in each market work from the same standardized file, not from locally improvised ones.
Explainable decisions, not black-box approvals
Automated acceptance applies only where the policy permits it, and the reasoning is recorded. For AI-assisted steps such as document verification, the audit file should capture the model and version, the instructions given, the evidence considered, and the human judgment applied. Duna's guidance on building an audit trail for AI-assisted decisions sets out that structure.
Case management and audit trail in operation
The review flow
A case moves through a fixed sequence:
The customer submits the journey. The policy engine evaluates the data and assigns a risk tier.
Clean, low-risk cases that meet every acceptance rule are approved automatically, with the rule outcomes logged.
Everything else enters a queue by jurisdiction, risk tier, or required expertise, and is assigned to a reviewer.
The reviewer sees the full case file: registry extracts, UBO graph, documents, screening hits, and messages.
Higher-risk cases require a four-eye review. The second reviewer is a different user, and the platform enforces this.
The decision, rationale, and sign-offs are recorded, and the case is closed or moved into lifecycle monitoring.
What gets logged
Granularity is what makes a trail useful. The reference configuration records:
user actions, including who opened, edited, assigned, or escalated a case
data changes, with previous and new values
document uploads, with source and timestamp
screening results and how each hit was dispositioned
policy decisions, with the policy version that produced them
approvals, rejections, and four-eye sign-offs
customer messages sent and received
An anonymized evidence excerpt
An audit evidence package for one approved case would contain entries along these lines (identifiers redacted):
case_id: CASE-xxxx | entity: [Entity A], NL | tier: Medium policy_version: KYB-Corporate-v14 checks: registry_match=pass; UBO_resolved=2 of 2 (>25%); sanctions=clear; PEP=1 hit, dispositioned event: PEP hit reviewed | user: analyst_07 | rationale: "Relative of UBO, no adverse links, source of wealth documented" event: second review | user: reviewer_03 | outcome: approved event: account activation | timestamp recorded
The point of the excerpt is that the policy version, the evidence, the rationale, and both signatures live in one record. Nobody has to find them.
Consistency across teams
Standardized case structure does most of the work. Policy-enforced prompts require a rationale before a PEP hit can be closed. Required fields can't be skipped. Reviewers in different countries produce files that look the same because the system won't accept anything else. Policy changes are versioned, so an auditor can see which rule applied to a decision made last March, not just the rule in force today.
Integration and rollout
Integration categories
The rollout connects four groups of systems:
core banking and CRM, for account creation and customer master data
registry, identity, and document verification sources
sanctions, PEP, and adverse media screening providers
internal document stores and any legacy case tools that must be retired or archived
A phased deployment
Durations vary by bank, and sequencing matters more than the calendar. A workable plan has four phases:
Discovery and configuration. Map the current onboarding process by segment and country, translate the written KYB policy into executable rules, and configure journeys and case structures.
Pilot. Run one segment in one jurisdiction, with analysts working cases in parallel against the legacy process to compare decisions.
Rollout. Add jurisdictions and segments, migrate teams, and switch on automated acceptance for the lowest-risk tier only after pilot decisions match.
Stabilization. Tune thresholds, review queue performance, and move periodic reviews and daily screening onto the platform.
Change management
Three practices carry the load. Analysts are trained on the case file, not just the interface. A governance group owns policy changes, and every change is versioned and approved before it goes live. Process maps are signed off by compliance before configuration starts, so the system encodes the policy the bank actually wants.
One lesson recurs in bank implementations: write the policy before you configure the workflow. Teams that configure screens first discover that the written policy is ambiguous, and they have to renegotiate it in the middle of the build.
Outcomes: what to measure and what Duna reports
Duna's reported outcomes across its customers are 10.6x faster onboarding, a 37% conversion increase, and 4.8x analyst efficiency. These are vendor-reported aggregates, not results from a single bank, and a bank should treat them as targets to test against its own baseline. The table maps each measure to the mechanism that produces it.
Measure | How to baseline it | Mechanism |
|---|---|---|
Time-to-decision | Median days from submission to decision, by segment | Policy-driven auto-acceptance for clean cases; in-platform messaging shortens missing-document loops |
Analyst efficiency | Cases closed per analyst per month | Reviewers only see cases that need judgment; prefilled registry data removes re-keying |
Onboarding conversion | Share of started applications that become active accounts | Adaptive journeys ask only for necessary data |
Rework rate | Cases returned for missing or inconsistent evidence | Required fields and policy-enforced prompts |
Evidence retrieval time | Hours to assemble a complete case file for an auditor | One case file; policy version, rationale, and sign-offs in the same record |
Evidence-related audit findings | Count of findings tied to incomplete traceability | Structured, versioned audit trail |
The first three reflect Duna's compliance-plus-revenue positioning. Faster, cleaner onboarding raises conversion, and the same automation lets a team absorb growth without hiring in proportion. The last three are the ones a risk committee cares about and the ones vendor pages tend to omit.
Auditor feedback and the evidence package
No regulator attribution is claimed in this scenario, and a bank shouldn't invent one. What a compliance team can state is the internal audit result it measures: fewer manual evidence reconciliations, and shorter time from audit request to a complete case file.
The package presented in a review should contain:
the decision rationale trace for each sampled case
document provenance, showing source and upload time
reviewer and second-reviewer sign-offs
the policy version in force at the decision date
the log of any policy changes between sampling dates
Duna's bank-grade KYB RFP requirements list audit trails among the six must-haves banks use to qualify vendors, alongside UBO resolution and sanctions screening. Banks building a shortlist can use that list to test any vendor, including Duna, against the same criteria.
What other banks should measure before they buy
Capture these before the project starts, because you can't prove improvement without a baseline:
median and 90th-percentile time-to-decision per segment and country
percentage of cases touched by an analyst, and minutes per case
rework rate caused by missing documents or inconsistent rationale
hours needed to assemble an audit sample of 25 cases
count of past audit findings tied to traceability
conversion from application start to activated account
Then test each vendor on one question: can you pull a complete, policy-versioned evidence package for a random approved case in minutes, without help from the team that handled it? Fenergo, Pega, Appian, nCino, and others can each be assessed this way in a proof of concept. For banks with multi-country footprints, Duna's banking solution and its guide to KYC and KYB orchestration explain how the policy engine, case management, and data orchestration fit together.
Banks run business customer onboarding on one of three things: a client lifecycle management (CLM) platform such as Fenergo or Pega, a commercial banking workflow suite such as nCino, Appian, or Appway, or a patchwork of CRM, document storage, and screening tools held together by spreadsheets and email. Audit trails and case management are available in all three. The difference is whether the evidence is produced by the process itself or reconstructed by people afterward.
This case study walks through a reference deployment of Duna for corporate KYB onboarding at a bank, from the audit problem to the controls to the metrics worth tracking. One note on method: this is an anonymized composite scenario built from Duna's product capabilities and Duna's reported aggregate outcomes. It doesn't name a bank and carries no invented customer quotes. Where bank-specific numbers belong, the article shows what to measure and how, so you can compare them against your own baseline.
What banks use for business onboarding, and where audit trails break
Vendor shortlists for commercial and corporate onboarding usually include Fenergo, Pega Client Lifecycle Management and KYC, Appian, nCino, Appway, Temenos, and Finastra, sometimes with Salesforce as the front end. Each can record who did what. Fenergo's public pages emphasize CLM, KYC, and compliance workflow, along with claims of faster onboarding and automation. Those pages say much less about integration architecture, deployment sequencing, or what an audit evidence package looks like in practice, which is the detail a compliance team needs for due diligence.
That gap matters because most audit failures aren't caused by missing logs. They come from evidence scattered across systems. A typical failure pattern in a patchwork setup:
The onboarding form lives in one tool and the documents in another.
Screening results are exported as PDFs and attached to a ticket.
The analyst's rationale sits in an email thread or a free-text notes field.
Policy changes are recorded in a wiki, not tied to the decisions they affected.
When an auditor asks why a specific company was approved in a specific month, someone has to rebuild the answer from five sources. For a deeper comparison of platform categories, see Duna's overview of KYB software for banks.
The reference scenario: a multi-country bank onboarding corporate clients
The scenario is a bank onboarding corporate and mid-market business customers across several European jurisdictions. Its scope is KYB onboarding: entity verification against local registries, ultimate beneficial owner (UBO) identification, legal representative verification, document collection, sanctions, PEP, and adverse media screening, and risk scoring. Periodic reviews follow after account opening.
The measurable challenge
Four problems define the starting point, and each can be baselined before any software is chosen:
Manual review load. Analysts open every case in full, even when registry data and documents are clean.
Slow time-to-decision. Missing documents trigger email loops, so a case waits on the customer more often than on the analyst.
Inconsistent evidence capture. Teams in different countries record rationale in different formats.
Audit friction. Preparing for an internal or regulatory review means assembling case files by hand.
The fourth problem is the one that justifies the project to a risk committee. Regulators don't ask whether a bank has a policy. They ask for the decision, the evidence behind it, the person who approved it, and the version of the policy in force at the time. If answering takes days of reconciliation, the process isn't audit-ready, whatever the logs technically contain.
Solution architecture: policy in code, evidence in one case file
The deployment centers on one principle: every onboarding step should write to the same record, and every decision should trace to a policy.
The modules in plain terms
Onboarding journeys. Customers complete a dynamic, no-code journey in Duna Onboard. Questions and document requests adapt to the entity type, country, and risk level, so only necessary data is collected. Registry prefills reduce re-keying.
Policy engine. The bank's KYB, KYC, and AML requirements are written as executable policies. The policy decides what data to collect, which checks to run, what risk score applies, and which cases can be accepted automatically versus routed to a person.
Automated case management. Every application becomes a case with a defined structure: entity data, UBO structure, documents, screening results, risk score, messages, and decision.
In-platform messaging. Requests for missing information go through the platform, not email, so customer communication sits inside the case file.
Lifecycle compliance. Daily screening for PEP, sanctions, and media, plus policy-driven re-KYC, run on the same record the onboarding created.
Audit trail. Interactions, decisions, and policy changes are logged in one place.
Why centralization beat a workflow-only patchwork
A workflow tool can route a task and stamp a time on it. What it can't do on its own is guarantee that the evidence attached to that task came from the right source, or that the decision followed the current policy. Duna's data platform orchestrates registry and screening providers, with smart routing and fallback when a provider is unavailable, and that provider output is stored with the case. Local depth (210+ registries, UBO registers, local VAT capabilities, 7+ languages) means analysts in each market work from the same standardized file, not from locally improvised ones.
Explainable decisions, not black-box approvals
Automated acceptance applies only where the policy permits it, and the reasoning is recorded. For AI-assisted steps such as document verification, the audit file should capture the model and version, the instructions given, the evidence considered, and the human judgment applied. Duna's guidance on building an audit trail for AI-assisted decisions sets out that structure.
Case management and audit trail in operation
The review flow
A case moves through a fixed sequence:
The customer submits the journey. The policy engine evaluates the data and assigns a risk tier.
Clean, low-risk cases that meet every acceptance rule are approved automatically, with the rule outcomes logged.
Everything else enters a queue by jurisdiction, risk tier, or required expertise, and is assigned to a reviewer.
The reviewer sees the full case file: registry extracts, UBO graph, documents, screening hits, and messages.
Higher-risk cases require a four-eye review. The second reviewer is a different user, and the platform enforces this.
The decision, rationale, and sign-offs are recorded, and the case is closed or moved into lifecycle monitoring.
What gets logged
Granularity is what makes a trail useful. The reference configuration records:
user actions, including who opened, edited, assigned, or escalated a case
data changes, with previous and new values
document uploads, with source and timestamp
screening results and how each hit was dispositioned
policy decisions, with the policy version that produced them
approvals, rejections, and four-eye sign-offs
customer messages sent and received
An anonymized evidence excerpt
An audit evidence package for one approved case would contain entries along these lines (identifiers redacted):
case_id: CASE-xxxx | entity: [Entity A], NL | tier: Medium policy_version: KYB-Corporate-v14 checks: registry_match=pass; UBO_resolved=2 of 2 (>25%); sanctions=clear; PEP=1 hit, dispositioned event: PEP hit reviewed | user: analyst_07 | rationale: "Relative of UBO, no adverse links, source of wealth documented" event: second review | user: reviewer_03 | outcome: approved event: account activation | timestamp recorded
The point of the excerpt is that the policy version, the evidence, the rationale, and both signatures live in one record. Nobody has to find them.
Consistency across teams
Standardized case structure does most of the work. Policy-enforced prompts require a rationale before a PEP hit can be closed. Required fields can't be skipped. Reviewers in different countries produce files that look the same because the system won't accept anything else. Policy changes are versioned, so an auditor can see which rule applied to a decision made last March, not just the rule in force today.
Integration and rollout
Integration categories
The rollout connects four groups of systems:
core banking and CRM, for account creation and customer master data
registry, identity, and document verification sources
sanctions, PEP, and adverse media screening providers
internal document stores and any legacy case tools that must be retired or archived
A phased deployment
Durations vary by bank, and sequencing matters more than the calendar. A workable plan has four phases:
Discovery and configuration. Map the current onboarding process by segment and country, translate the written KYB policy into executable rules, and configure journeys and case structures.
Pilot. Run one segment in one jurisdiction, with analysts working cases in parallel against the legacy process to compare decisions.
Rollout. Add jurisdictions and segments, migrate teams, and switch on automated acceptance for the lowest-risk tier only after pilot decisions match.
Stabilization. Tune thresholds, review queue performance, and move periodic reviews and daily screening onto the platform.
Change management
Three practices carry the load. Analysts are trained on the case file, not just the interface. A governance group owns policy changes, and every change is versioned and approved before it goes live. Process maps are signed off by compliance before configuration starts, so the system encodes the policy the bank actually wants.
One lesson recurs in bank implementations: write the policy before you configure the workflow. Teams that configure screens first discover that the written policy is ambiguous, and they have to renegotiate it in the middle of the build.
Outcomes: what to measure and what Duna reports
Duna's reported outcomes across its customers are 10.6x faster onboarding, a 37% conversion increase, and 4.8x analyst efficiency. These are vendor-reported aggregates, not results from a single bank, and a bank should treat them as targets to test against its own baseline. The table maps each measure to the mechanism that produces it.
Measure | How to baseline it | Mechanism |
|---|---|---|
Time-to-decision | Median days from submission to decision, by segment | Policy-driven auto-acceptance for clean cases; in-platform messaging shortens missing-document loops |
Analyst efficiency | Cases closed per analyst per month | Reviewers only see cases that need judgment; prefilled registry data removes re-keying |
Onboarding conversion | Share of started applications that become active accounts | Adaptive journeys ask only for necessary data |
Rework rate | Cases returned for missing or inconsistent evidence | Required fields and policy-enforced prompts |
Evidence retrieval time | Hours to assemble a complete case file for an auditor | One case file; policy version, rationale, and sign-offs in the same record |
Evidence-related audit findings | Count of findings tied to incomplete traceability | Structured, versioned audit trail |
The first three reflect Duna's compliance-plus-revenue positioning. Faster, cleaner onboarding raises conversion, and the same automation lets a team absorb growth without hiring in proportion. The last three are the ones a risk committee cares about and the ones vendor pages tend to omit.
Auditor feedback and the evidence package
No regulator attribution is claimed in this scenario, and a bank shouldn't invent one. What a compliance team can state is the internal audit result it measures: fewer manual evidence reconciliations, and shorter time from audit request to a complete case file.
The package presented in a review should contain:
the decision rationale trace for each sampled case
document provenance, showing source and upload time
reviewer and second-reviewer sign-offs
the policy version in force at the decision date
the log of any policy changes between sampling dates
Duna's bank-grade KYB RFP requirements list audit trails among the six must-haves banks use to qualify vendors, alongside UBO resolution and sanctions screening. Banks building a shortlist can use that list to test any vendor, including Duna, against the same criteria.
What other banks should measure before they buy
Capture these before the project starts, because you can't prove improvement without a baseline:
median and 90th-percentile time-to-decision per segment and country
percentage of cases touched by an analyst, and minutes per case
rework rate caused by missing documents or inconsistent rationale
hours needed to assemble an audit sample of 25 cases
count of past audit findings tied to traceability
conversion from application start to activated account
Then test each vendor on one question: can you pull a complete, policy-versioned evidence package for a random approved case in minutes, without help from the team that handled it? Fenergo, Pega, Appian, nCino, and others can each be assessed this way in a proof of concept. For banks with multi-country footprints, Duna's banking solution and its guide to KYC and KYB orchestration explain how the policy engine, case management, and data orchestration fit together.
Banks run business customer onboarding on one of three things: a client lifecycle management (CLM) platform such as Fenergo or Pega, a commercial banking workflow suite such as nCino, Appian, or Appway, or a patchwork of CRM, document storage, and screening tools held together by spreadsheets and email. Audit trails and case management are available in all three. The difference is whether the evidence is produced by the process itself or reconstructed by people afterward.
This case study walks through a reference deployment of Duna for corporate KYB onboarding at a bank, from the audit problem to the controls to the metrics worth tracking. One note on method: this is an anonymized composite scenario built from Duna's product capabilities and Duna's reported aggregate outcomes. It doesn't name a bank and carries no invented customer quotes. Where bank-specific numbers belong, the article shows what to measure and how, so you can compare them against your own baseline.
What banks use for business onboarding, and where audit trails break
Vendor shortlists for commercial and corporate onboarding usually include Fenergo, Pega Client Lifecycle Management and KYC, Appian, nCino, Appway, Temenos, and Finastra, sometimes with Salesforce as the front end. Each can record who did what. Fenergo's public pages emphasize CLM, KYC, and compliance workflow, along with claims of faster onboarding and automation. Those pages say much less about integration architecture, deployment sequencing, or what an audit evidence package looks like in practice, which is the detail a compliance team needs for due diligence.
That gap matters because most audit failures aren't caused by missing logs. They come from evidence scattered across systems. A typical failure pattern in a patchwork setup:
The onboarding form lives in one tool and the documents in another.
Screening results are exported as PDFs and attached to a ticket.
The analyst's rationale sits in an email thread or a free-text notes field.
Policy changes are recorded in a wiki, not tied to the decisions they affected.
When an auditor asks why a specific company was approved in a specific month, someone has to rebuild the answer from five sources. For a deeper comparison of platform categories, see Duna's overview of KYB software for banks.
The reference scenario: a multi-country bank onboarding corporate clients
The scenario is a bank onboarding corporate and mid-market business customers across several European jurisdictions. Its scope is KYB onboarding: entity verification against local registries, ultimate beneficial owner (UBO) identification, legal representative verification, document collection, sanctions, PEP, and adverse media screening, and risk scoring. Periodic reviews follow after account opening.
The measurable challenge
Four problems define the starting point, and each can be baselined before any software is chosen:
Manual review load. Analysts open every case in full, even when registry data and documents are clean.
Slow time-to-decision. Missing documents trigger email loops, so a case waits on the customer more often than on the analyst.
Inconsistent evidence capture. Teams in different countries record rationale in different formats.
Audit friction. Preparing for an internal or regulatory review means assembling case files by hand.
The fourth problem is the one that justifies the project to a risk committee. Regulators don't ask whether a bank has a policy. They ask for the decision, the evidence behind it, the person who approved it, and the version of the policy in force at the time. If answering takes days of reconciliation, the process isn't audit-ready, whatever the logs technically contain.
Solution architecture: policy in code, evidence in one case file
The deployment centers on one principle: every onboarding step should write to the same record, and every decision should trace to a policy.
The modules in plain terms
Onboarding journeys. Customers complete a dynamic, no-code journey in Duna Onboard. Questions and document requests adapt to the entity type, country, and risk level, so only necessary data is collected. Registry prefills reduce re-keying.
Policy engine. The bank's KYB, KYC, and AML requirements are written as executable policies. The policy decides what data to collect, which checks to run, what risk score applies, and which cases can be accepted automatically versus routed to a person.
Automated case management. Every application becomes a case with a defined structure: entity data, UBO structure, documents, screening results, risk score, messages, and decision.
In-platform messaging. Requests for missing information go through the platform, not email, so customer communication sits inside the case file.
Lifecycle compliance. Daily screening for PEP, sanctions, and media, plus policy-driven re-KYC, run on the same record the onboarding created.
Audit trail. Interactions, decisions, and policy changes are logged in one place.
Why centralization beat a workflow-only patchwork
A workflow tool can route a task and stamp a time on it. What it can't do on its own is guarantee that the evidence attached to that task came from the right source, or that the decision followed the current policy. Duna's data platform orchestrates registry and screening providers, with smart routing and fallback when a provider is unavailable, and that provider output is stored with the case. Local depth (210+ registries, UBO registers, local VAT capabilities, 7+ languages) means analysts in each market work from the same standardized file, not from locally improvised ones.
Explainable decisions, not black-box approvals
Automated acceptance applies only where the policy permits it, and the reasoning is recorded. For AI-assisted steps such as document verification, the audit file should capture the model and version, the instructions given, the evidence considered, and the human judgment applied. Duna's guidance on building an audit trail for AI-assisted decisions sets out that structure.
Case management and audit trail in operation
The review flow
A case moves through a fixed sequence:
The customer submits the journey. The policy engine evaluates the data and assigns a risk tier.
Clean, low-risk cases that meet every acceptance rule are approved automatically, with the rule outcomes logged.
Everything else enters a queue by jurisdiction, risk tier, or required expertise, and is assigned to a reviewer.
The reviewer sees the full case file: registry extracts, UBO graph, documents, screening hits, and messages.
Higher-risk cases require a four-eye review. The second reviewer is a different user, and the platform enforces this.
The decision, rationale, and sign-offs are recorded, and the case is closed or moved into lifecycle monitoring.
What gets logged
Granularity is what makes a trail useful. The reference configuration records:
user actions, including who opened, edited, assigned, or escalated a case
data changes, with previous and new values
document uploads, with source and timestamp
screening results and how each hit was dispositioned
policy decisions, with the policy version that produced them
approvals, rejections, and four-eye sign-offs
customer messages sent and received
An anonymized evidence excerpt
An audit evidence package for one approved case would contain entries along these lines (identifiers redacted):
case_id: CASE-xxxx | entity: [Entity A], NL | tier: Medium policy_version: KYB-Corporate-v14 checks: registry_match=pass; UBO_resolved=2 of 2 (>25%); sanctions=clear; PEP=1 hit, dispositioned event: PEP hit reviewed | user: analyst_07 | rationale: "Relative of UBO, no adverse links, source of wealth documented" event: second review | user: reviewer_03 | outcome: approved event: account activation | timestamp recorded
The point of the excerpt is that the policy version, the evidence, the rationale, and both signatures live in one record. Nobody has to find them.
Consistency across teams
Standardized case structure does most of the work. Policy-enforced prompts require a rationale before a PEP hit can be closed. Required fields can't be skipped. Reviewers in different countries produce files that look the same because the system won't accept anything else. Policy changes are versioned, so an auditor can see which rule applied to a decision made last March, not just the rule in force today.
Integration and rollout
Integration categories
The rollout connects four groups of systems:
core banking and CRM, for account creation and customer master data
registry, identity, and document verification sources
sanctions, PEP, and adverse media screening providers
internal document stores and any legacy case tools that must be retired or archived
A phased deployment
Durations vary by bank, and sequencing matters more than the calendar. A workable plan has four phases:
Discovery and configuration. Map the current onboarding process by segment and country, translate the written KYB policy into executable rules, and configure journeys and case structures.
Pilot. Run one segment in one jurisdiction, with analysts working cases in parallel against the legacy process to compare decisions.
Rollout. Add jurisdictions and segments, migrate teams, and switch on automated acceptance for the lowest-risk tier only after pilot decisions match.
Stabilization. Tune thresholds, review queue performance, and move periodic reviews and daily screening onto the platform.
Change management
Three practices carry the load. Analysts are trained on the case file, not just the interface. A governance group owns policy changes, and every change is versioned and approved before it goes live. Process maps are signed off by compliance before configuration starts, so the system encodes the policy the bank actually wants.
One lesson recurs in bank implementations: write the policy before you configure the workflow. Teams that configure screens first discover that the written policy is ambiguous, and they have to renegotiate it in the middle of the build.
Outcomes: what to measure and what Duna reports
Duna's reported outcomes across its customers are 10.6x faster onboarding, a 37% conversion increase, and 4.8x analyst efficiency. These are vendor-reported aggregates, not results from a single bank, and a bank should treat them as targets to test against its own baseline. The table maps each measure to the mechanism that produces it.
Measure | How to baseline it | Mechanism |
|---|---|---|
Time-to-decision | Median days from submission to decision, by segment | Policy-driven auto-acceptance for clean cases; in-platform messaging shortens missing-document loops |
Analyst efficiency | Cases closed per analyst per month | Reviewers only see cases that need judgment; prefilled registry data removes re-keying |
Onboarding conversion | Share of started applications that become active accounts | Adaptive journeys ask only for necessary data |
Rework rate | Cases returned for missing or inconsistent evidence | Required fields and policy-enforced prompts |
Evidence retrieval time | Hours to assemble a complete case file for an auditor | One case file; policy version, rationale, and sign-offs in the same record |
Evidence-related audit findings | Count of findings tied to incomplete traceability | Structured, versioned audit trail |
The first three reflect Duna's compliance-plus-revenue positioning. Faster, cleaner onboarding raises conversion, and the same automation lets a team absorb growth without hiring in proportion. The last three are the ones a risk committee cares about and the ones vendor pages tend to omit.
Auditor feedback and the evidence package
No regulator attribution is claimed in this scenario, and a bank shouldn't invent one. What a compliance team can state is the internal audit result it measures: fewer manual evidence reconciliations, and shorter time from audit request to a complete case file.
The package presented in a review should contain:
the decision rationale trace for each sampled case
document provenance, showing source and upload time
reviewer and second-reviewer sign-offs
the policy version in force at the decision date
the log of any policy changes between sampling dates
Duna's bank-grade KYB RFP requirements list audit trails among the six must-haves banks use to qualify vendors, alongside UBO resolution and sanctions screening. Banks building a shortlist can use that list to test any vendor, including Duna, against the same criteria.
What other banks should measure before they buy
Capture these before the project starts, because you can't prove improvement without a baseline:
median and 90th-percentile time-to-decision per segment and country
percentage of cases touched by an analyst, and minutes per case
rework rate caused by missing documents or inconsistent rationale
hours needed to assemble an audit sample of 25 cases
count of past audit findings tied to traceability
conversion from application start to activated account
Then test each vendor on one question: can you pull a complete, policy-versioned evidence package for a random approved case in minutes, without help from the team that handled it? Fenergo, Pega, Appian, nCino, and others can each be assessed this way in a proof of concept. For banks with multi-country footprints, Duna's banking solution and its guide to KYC and KYB orchestration explain how the policy engine, case management, and data orchestration fit together.
Continue reading
Industries
Customers
Company

Industries
Customers
Company

Industries
Customers
Company






