An input-to-output workflow for a back-book refresh programme, with explicit exception handling and a practical technology acceptance checklist.
On this page
Short answer
Automate the repeatable work: reconcile the population, identify gaps, retrieve current evidence, resolve ownership where sources permit, refresh screening and assemble the case. Route uncertainty and policy-required decisions to your own reviewers. Release only after testing data lineage, exceptions, quality controls and completion evidence.
This page covers the high-volume remediation workflow: how cases move from extract to evidenced completion. For choosing a vendor, see how to evaluate UK CDD software; for the capabilities Agora publishes, see KYC remediation software.
This is an implementation blueprint, not a claim that every file can be completed automatically. The existing remediation guide explains triggers and programme governance; this page concentrates on the data contract, processing states and acceptance tests used to buy and configure technology. The workflow recommendations are Agora practitioner interpretation, not additional statutory requirements or legal advice.
Define what goes in and what counts as complete
Start with a versioned target standard by customer type and risk category. Identify mandatory evidence, acceptable sources, review triggers, approval responsibilities and the firm's treatment of unresolved information. A field being populated is not the same as the underlying CDD requirement being satisfied. Do not set the completion rule to “the automation ran”.
Agree an input specification with the source-system owners before loading a population. Preserve the original extract and its scope approval. Use stable customer identifiers so later exports and reviewer decisions can be reconciled to the right source record. Where duplicates are suspected, preserve the relationship between records and resolve uncertainty rather than silently combining customers.
Example input specification
- Identity: source customer ID, customer type, legal name, available registry identifier and jurisdiction.
- Existing CDD: current risk rating, last review date, evidence references and existing ownership information.
- Scope: inclusion reason, source system, extraction date and population version.
- Known gaps: missing, stale, conflicting or unverified information recorded separately.
- Operational linkage: relationship owner and authorised outreach channel where applicable.
These are suggested buyer inputs, not a promised Agora upload schema. Agree actual field mappings, formats and integrations during technical scoping. Limit transfers to necessary data and apply the firm's privacy, access and retention controls. Use synthetic examples for early demonstrations.
A controlled processing workflow
1. Reconcile the scope before running checks
Validate required identifiers, flag conflicting records and separate rejected inputs from accepted cases. Produce a reconciliation of the source population, excluded records and cases opened. Hold exclusions with a reason and owner. A successful upload does not establish that the population is complete or that its customer identities are correct.
2. Classify gaps and prioritise work
Assess each record against the signed-off standard. Separate “obtainable from a registry”, “requires a configured data provider”, “requires customer response” and “requires judgement”. Combine these categories with the firm's risk-sensitive priority rules. A high-risk case with weak evidence should not sit behind a routine record merely because the routine record is easier to complete.
3. Enrich and resolve ownership with provenance
Retrieve entity information from supported registries and approved sources, then reconstruct ownership where sufficient evidence is available. Preserve source and retrieval date alongside the conclusion. Flag an incomplete chain, conflicting ownership or unsupported jurisdiction. A registry match can help identify the entity; it should not be treated as blanket proof that all due diligence is satisfied.
4. Refresh screening and customer risk
Run sanctions, PEP and adverse media screening through the agreed provider and matching configuration. Keep match details and the rule basis for routing or discounting. Apply the versioned customer risk methodology to the refreshed evidence and route additional due diligence where policy requires it. Do not treat a screening alert as a confirmed match, or an automatically assigned rating as final customer acceptance.
5. Route exceptions and outreach deliberately
Define distinct routes for identity uncertainty, unresolved ownership, screening decisions, missing purpose and intended nature, and EDD approval. Customer-only information belongs in a structured outreach process, with request, response and outstanding status linked to the case. Non-response is an unresolved state with escalation under the firm's policy, not a reason to invent a missing answer.
6. Recheck quality and assemble completion evidence
Apply the agreed checking model to the routine and exception paths. Capture defects, return cases for rework and record recheck results. Maintain the reviewer and approval trail where required. The completion pack should explain the input, checks, source evidence, methodology version, outstanding limitations, decisions and quality outcome. Required retention periods and access arrangements should come from the firm's applicable obligations, not a generic software default.
Worked example: three synthetic cases, three different outcomes
The cases below are invented solely to illustrate processing logic. They are not customers, case studies, production results or a representation of exact platform screen or export fields. No automation percentage or delivery time is implied.
Case A: a routine corporate refresh
Input: a fictional UK company has an existing customer ID, a valid registry reference, dated company information and complete relationship-purpose evidence. Ownership is direct and supported by available evidence. Processing: refresh registry information, validate the ownership basis, screen relevant parties and recalculate risk under the agreed method. Output: a sourced case pack, check outcomes and the applicable review or approval route. Routine processing can be automated where policy permits; final acceptance remains accountable to the firm.
Case B: ownership cannot be resolved
Input: a fictional company is owned through an overseas entity and the available sources do not establish the controlling natural persons. Processing: retain the resolved part of the chain and identify the precise evidence gap. Output: an ownership exception, an evidence request and an assigned reviewer, rather than an invented UBO or a completed status. The firm determines whether evidence is sufficient and whether escalation or further measures are needed.
Case C: screening requires judgement
Input: a fictional beneficial owner produces an ambiguous screening alert with insufficient identifiers to resolve it. Processing: retain the alert, identifiers, provider result and matching configuration. Output: a decision queue with relevant evidence and a place for a named reviewer to record the rationale. Any related risk change or EDD workflow follows the firm's approved policy. The automation must not clear the alert merely to improve throughput.
Programme outputs to request
Ask for population reconciliation, gaps by category, cases waiting for customer evidence, decision queues, case ageing, defects and accepted completions. Keep counts tied to defined states so a case waiting for outreach is not reported as complete. Agora's case reporting supports PDF, JSON, XML and CSV exports; agree which programme outputs and field mappings are required rather than assuming an example report is a standard deliverable.
Pilot the difficult paths before scaling
Choose a pilot that deliberately includes poor identifiers, layered ownership, stale evidence, alert decisions, outreach and rework. Document the selection bias: a convenient sample of clean files is not evidence for the whole back book. Set acceptance criteria before testing, then compare the actual result against those criteria. Investigate exceptions and incorrect outcomes, not just the cases the pipeline processed successfully.
Release in controlled cohorts with reconciliation and an agreed response to failed dependencies or incorrect rules. Track source availability, queue growth, incomplete evidence and quality findings. Preserve version history when changing a rule and identify cases requiring reassessment. At programme closure, reconcile exclusions and unresolved cases and agree how refreshed records enter periodic and perpetual KYC.
Buyer acceptance checklist
- Can the technology reconcile every input to a case or a documented exclusion?
- Can it distinguish missing, stale, conflicting and unverified evidence?
- Are supported data sources, jurisdictions and provider dependencies explicit?
- Does unresolved ownership remain visible rather than being silently accepted?
- Are screening decisions, risk rules and overrides traceable to versions and people?
- Do outreach, escalation, rework and recheck remain attached to the same case?
- Can management information separate progress from accepted completion?
- Can an auditor reconstruct a case from its export and linked evidence?
- Are live-data safeguards, operational ownership and the BAU destination agreed?
Agora supplies the remediation technology: population ingestion, gap identification, enrichment, ownership resolution where supported, screening, configurable workflow, outreach, quality control and case reporting. Your firm provides the reviewers and accountable approvals. Use the technology assessment to structure scope questions and the planning pack to record requirements. Neither is a staffing estimate or a price quotation.
Primary sources
- Money Laundering Regulations 2017, regulation 28: customer due diligence
- Money Laundering Regulations 2017, regulation 40: record-keeping
- FCA Financial Crime Guide, chapter 3
- JMLSG: financial-sector guidance
Frequently asked questions
What can be automated in KYC remediation?
Typically population ingestion and reconciliation, gap classification, registry and data enrichment, screening refresh, routing to queues, outreach tracking and evidence assembly. Judgements such as discounting an ambiguous alert or accepting incomplete ownership evidence should stay with named reviewers.
How should remediation progress be reported?
Use defined case states such as awaiting evidence, in review, in quality control and accepted, so that cases started or processed are not reported as complete.
Should a remediation programme start with a pilot?
Yes. Include difficult cases such as poor identifiers, layered ownership, stale evidence and screening decisions, set acceptance criteria before testing and document any sample selection bias.
What happens to records after remediation closes?
Refreshed records should move into the firm's ongoing review model, with a next review date, applicable event triggers and preserved evidence history.
Does automating remediation remove the need for reviewers?
No. Automation can reduce manual collection and routing, but your firm's reviewers still make and record risk decisions.
Related resources
Remediation and ongoing KYC
KYC Remediation: How to Modernise Customer File Remediation
CDD and onboarding
Customer Due Diligence Automation: A Practical Guide
CDD and onboarding
Best UK CDD Software: How to Compare Providers
Remediation and ongoing KYC
Fenergo Alternatives for KYC Remediation
Where the technology fits
Agora is a technology provider: the platform supports the control described above, and your own teams operate it and hold the accountable decisions.
Next step
Test the workflow against your remediation scope
Discuss your input data, target CDD standard, exception routes and required outputs in a demonstration. Review and acceptance remain with your own teams.