The short answer
"Intelligent automation" combines several tools:
- Workflow automation that routes work between people and systems.
- Robotic process automation (RPA) that operates existing applications the way a person would, useful where there is no API.
- Integration through APIs, which is more reliable than RPA whenever it is available.
- Document capture that reads statements, invoices and identity documents.
- Machine learning or AI that classifies, prioritises or flags items for review.
In payments and financial services, the best first candidates are processes that are high-volume, rule-based and already well documented. People stay responsible for judgement calls, exceptions and anything a regulator would expect a person to decide.
Good candidates
| Process | What automation does | What people keep |
|---|---|---|
| Settlement and bank reconciliation | Matches transactions across processor, bank and ledger files; raises breaks | Investigating and resolving breaks |
| Merchant or client onboarding | Collects documents, checks completeness, pre-fills records, runs screening | Risk decisions and approvals |
| Chargebacks and disputes | Gathers transaction evidence and assembles the response pack | Deciding whether and how to contest |
| Customer service case triage | Classifies and routes cases, drafts replies for common requests | Complex or sensitive cases |
| Regulatory and management reporting | Pulls and validates data, prepares schedules | Review, sign-off and submission |
| Periodic client reviews | Schedules reviews, requests updated information, tracks responses | Assessing changes in risk |
Our article on which workflows to automate first covers how to score and rank candidates.
Poor candidates
- Processes that change frequently or are not documented.
- Work that depends on judgement the organisation cannot explain as rules or examples.
- Very low-volume tasks, where the build and maintenance cost outweighs the time saved.
- Screen-scraping of systems that are about to be replaced; wait, or integrate with the replacement.
Controls that make automation trustworthy
Automation in a regulated business is part of the control environment. Design it that way from the start:
- A named business owner for each automation, accountable for its output.
- Dedicated bot identities (service accounts) with least-privilege access, never shared human logins, and credentials held in a secrets vault.
- Segregation of duties. A bot that prepares a payment file should not also approve and release it.
- Complete audit trails showing each item processed, the rule or model applied, the result and any human decision.
- Exception queues with clear ownership, so items the automation cannot handle are not silently dropped.
- Change management: automations are tested, approved and versioned like any other production change.
- Monitoring and alerts when volumes, error rates or processing times move outside normal ranges.
- Fallback procedures so staff can process critical work manually if an automation or a target system fails, including during power or network outages.
Where machine learning or AI is involved, add documentation of what the model does, the data it was trained or configured on, how its accuracy is checked and when a person must review its output.
The Nigerian regulatory context
Automation does not change who is accountable. Several frameworks shape what "controlled" means:
- Payment service providers and banks. The Central Bank of Nigeria (CBN) licenses and supervises banks and payment service providers, and issues cybersecurity and risk management requirements that cover how technology, including automated processes, is governed (CBN). Automated processes fall within that operational risk picture.
- Anti-money laundering. Financial institutions and designated non-financial businesses have anti-money laundering and counter-terrorist financing duties, with the Nigerian Financial Intelligence Unit (NFIU) receiving reports and SCUML supervising designated non-financial businesses (NFIU). Automation can gather information, screen and flag, but the compliance programme and its decisions remain the organisation's responsibility.
- Other sector regulators. Capital-market operators answer to SEC Nigeria, insurers to NAICOM and pension operators to PenCom, each of which may set its own expectations for technology and outsourcing.
- Data protection. Automations that process personal data need the same safeguards and purpose limits as manual processes under the Nigeria Data Protection Act 2023 (NDPA), enforced by the Nigeria Data Protection Commission (NDPC). Where an automation relies on a service hosted outside Nigeria, the NDPA's rules on transfers apply.
This is general information, not legal or compliance advice.
Instant payments and richer data
Nigerian customers already expect money to move in seconds: instant bank transfers (NIP), card payments and wallet payments are part of everyday business. That speed shortens the time available to catch errors or fraud, which raises the value of automated checks that run before money moves. It also produces a high volume of transactions from several channels, each with its own file format and references, which is exactly where automated matching and reconciliation earn their keep. Where transfer narrations are free text, matching rules need tolerance and a review queue rather than an assumption that references are clean.
A hypothetical payment facilitator reconciles daily settlement files from two processors against its bank statements and ledger using spreadsheets. An automation now:
- Collects the files each morning through the processors' and bank's secure channels.
- Matches transactions by reference, amount and date, applying documented tolerance rules.
- Posts matched items to the ledger and creates a break for each unmatched item, assigned to the operations team.
- Produces a summary for the finance manager, who reviews and signs off.
The bot has read-only access to the processor portals and can post only to a reconciliation clearing account. Any change to matching rules goes through the same approval process as a finance policy change.
Measuring value
Agree on measures before building: time spent per item, backlog age, error and rework rates, exceptions per thousand items and time to close a break. Measure the manual baseline first; otherwise any claimed improvement is a guess.
Readiness checklist
Process
- Process documented step by step, including exceptions
- Volumes and current handling time measured
- Business owner named
Controls
- Bot identities and least-privilege access designed
- Segregation of duties confirmed
- Audit trail content agreed with compliance or internal audit
- Exception queue ownership and service levels agreed
Technology
- API options checked before choosing RPA
- Monitoring, alerting and fallback procedure defined
- Change management and testing approach agreed
Compliance
- Applicable regulatory obligations reviewed
- Privacy impact of the automation assessed
Limitations
Automation amplifies whatever process it runs: a flawed process becomes a faster flawed process. It also needs maintenance whenever the systems it touches change. Start with one well-understood process, prove the controls, then expand. Our workflow and business process automation service follows that approach.
Sources and further reading
Product capabilities and guidance change. These are the primary sources this article relies on, checked on the review date above.
- Central Bank of Nigeria, Central Bank of Nigeria
- Nigerian Financial Intelligence Unit, NFIU
- Nigeria Data Protection Commission (NDPA and GAID 2025), Nigeria Data Protection Commission
- Cybersecurity Framework, NIST
This article is general information, not legal, accounting or security advice for your specific situation. Examples are hypothetical unless stated otherwise.