RPA in banking: automation for banks, credit unions and lenders
RPA in banking takes the copying, checking and chasing out of the back office: software robots read the core banking system, the CRM, the document archive and the spreadsheets, move data between them, and prepare complete cases for people to decide. Done well, it frees hours in operations and compliance without touching the core system. Done badly, it creates unmonitored accounts with access to customer data. We build robotic process automation for banks, credit unions and lenders with the controls a regulator expects, hosted in the EU or in your own environment, and the bank owns every bot.
- You see the result before it goes live
- Fixed price for the agreed scope, in writing before work starts
- Source code, domain and accounts in your company's name
Where RPA pays back in a bank
| Process | What the bot does | What the person does |
|---|---|---|
| Periodic customer reviews | Collects customer data, documents, registry extracts and transaction summaries into one review file | Assesses risk and decides on the relationship |
| Account opening and closing | Creates or closes the customer in core, CRM, cards and archive with the same data | Approves exceptions |
| Loan files | Checks completeness, reads key values from documents, compares with the application | Makes the credit decision |
| Reconciliations | Matches ledgers, statements and payment files, lists breaks | Investigates and resolves breaks |
| Payment exceptions | Gathers the payment, customer and history for each rejected or held payment | Releases, returns or escalates |
| Alert preparation | Assembles customer and transaction context for screening and monitoring alerts | Investigates and decides |
| Regulatory reporting data | Collects figures from several systems into the reporting template, with checks | Reviews and signs off |
The common pattern: the bot does the gathering and the typing, the person does the judgement. That keeps accountability where the regulator expects it and puts the bank's experts on the part of the work only they can do.
Customer reviews and the new AML rules
Customer due diligence reviews are one of the largest manual workloads in retail and business banking, and the EU's Anti-Money Laundering Regulation sets firm intervals. Obliged entities shall conduct ongoing monitoring of business relationships, ensure that customer documents, data and information are kept up to date, and the period between updates shall not exceed 1 year for higher risk customers and 5 years for all other customers; information must also be reviewed when relevant circumstances change. The Regulation applies from 10 July 2027 (Regulation (EU) 2024/1624, Article 26 and Article 90, read on 11 November 2026).
For a bank with a large customer base, that is a steady stream of review files. A bot can open each due review, pull the customer record, identity documents, beneficial ownership and registry data, recent transaction patterns and earlier review notes into one file, flag what is missing or has changed, and send the customer a request for updated information. The analyst starts from a complete file instead of six screens. The review decision, and any decision to end a relationship, stays with the analyst. For onboarding new customers through identity, ownership and screening checks, see KYC automation.
Bots under DORA
A software robot with access to customer data is an ICT asset, and the Digital Operational Resilience Act, which applies from 17 January 2025, treats it as one. Financial entities shall identify, classify and adequately document all ICT supported business functions, the information and ICT assets supporting them and their dependencies (Article 8(1)); implement policies that limit physical or logical access to information and ICT assets to what is required for legitimate and approved functions, and policies and protocols for strong authentication mechanisms (Article 9(4)(c) and (d)); and manage ICT third-party risk, including a register of information on all contractual arrangements on the use of ICT services provided by ICT third-party service providers (Article 28(1) and (3)) (Regulation (EU) 2022/2554, read on 11 November 2026).
In practice, every bot we build comes with:
- Its own identity. A named service account per bot, never a person's login, with only the rights its task needs.
- Credentials in the vault. Passwords and keys held in the bank's secrets vault and rotated, never stored in the bot or a script.
- A record in your inventory. The bot, its owner, the business function it supports, the systems it touches and its criticality, kept current.
- Logs for every run. What it read, what it changed and what it handed to a person, kept for your audit period.
- Segregation of duties. A bot that prepares a payment or a change cannot also approve it.
- A manual fallback. A written procedure so the work continues if the bot stops.
The RPA platform you use, and any outside party that builds or runs bots, belongs in the register of information. We supply the details for that entry.
RPA or an interface?
RPA works through screens, and screens change. Where a system offers an interface, an integration is faster, more reliable and easier to monitor, and many core banking platforms and payment hubs now offer one. We use RPA where nothing else reaches: a legacy application without an interface, a supplier portal, a system that is due to be replaced within two years, or a process that needs to run next month while the proper integration is built. For interface work, see API integration; for document-heavy flows such as loan files, see document workflow automation.
Where people must stay in charge
Some steps should not be handed to a bot or a model at all. Under the EU AI Act, AI systems intended to be used to evaluate the creditworthiness of natural persons or establish their credit score are high-risk, with the exception of AI systems used for detecting financial fraud (Regulation (EU) 2024/1689, Annex III, point 5(b)); under the amended Act the Annex III high-risk requirements apply from 2 December 2027 (AI Act Service Desk, Article 113). (Both read on 11 November 2026.) AML and sanctions decisions require human judgement and a named decision-maker. We design bots and any AI document reading so that they prepare, check and explain, and a person decides, with the reasoning recorded.
Making bots last
- Fail safely. A bot that meets an unexpected screen or value stops and hands the case to a person, rather than guessing.
- Monitor every run. Volumes, failures and exceptions on a dashboard the process owner checks daily.
- Know about changes first. Bots are listed as dependants of the systems they use, so release notes reach the bot owner before the change.
- Measure real results. Hours saved are confirmed with the people who did the work, not estimated once and repeated. Our guide to the benefits of RPA shows why that matters.
- Retire bots. When a system gains an interface or the process changes, the bot is replaced or switched off, not left running.
Common tools
Banks use established RPA platforms such as UiPath, Microsoft Power Automate, Automation Anywhere and Blue Prism, often already licensed somewhere in the group. We work with the platform you have where it fits, and with scripted automation or direct integration where that is the more robust choice. The bot design, the controls and the documentation are the same whichever tool runs it.
How we deliver
- Pick one process. With operations and compliance, we choose a process with volume, clear rules and a measurable baseline, and record how long it takes and how often it goes wrong.
- Design with risk and security. Access rights, credentials, logging, segregation of duties, the fallback and the inventory entry are agreed with your risk and security teams before building.
- Scope and fixed price in writing. The process, systems, controls, tests and what the first phase leaves out.
- Build and test in a test environment. On test data and test systems, then a parallel run next to the manual process.
- Go live and measure. The process owner watches the first weeks daily; hours and errors are compared with the baseline, then the next process is picked.
For automation across the business, see business process automation; for systems that run finance teams, see financial software development; for customer-facing banking apps, see fintech app development; for insurers, see insurance software development.
The bank owns it
As a standard term, bot definitions, code and configuration sit in the bank's own repositories, bots run in your environment or your cloud account in an EU region, and credentials never leave your vault. Documentation is written for your auditors and your own team. You can keep us for maintenance, run the bots yourselves or hand them to another team. Where an interface can replace a bot, see banking software development. See prices.
Questions we get
What is RPA in banking?
Robotic process automation in banking is software that operates a bank's existing systems the way a person does: it logs in, reads screens and documents, copies data between the core banking system, CRM, document archive and spreadsheets, and prepares cases for a person to decide. It is used where systems do not offer an interface, or where an integration would take far longer to build.
Which banking processes are good candidates for RPA?
High-volume, rule-based steps with structured inputs: collecting data for periodic customer due diligence reviews, account opening and closing across several systems, loan file completeness checks, reconciliations, payment exception handling, preparing alerts for investigators and gathering data for regulatory reports. The decision in each case stays with a person.
Is RPA allowed under DORA?
Yes, but bots are ICT assets and the platform and any outside provider are ICT third parties. Under DORA, a bank must identify and document the business functions and ICT assets that support them, limit access to what is required, use strong authentication, and manage ICT third-party risk, including a register of information on contracts for ICT services. Bots are designed to fit those rules from the start.
Should a bot make credit or AML decisions?
No. Bots collect, check and prepare; people decide. Creditworthiness assessment of natural persons by AI is a high-risk use under the EU AI Act, and AML decisions require human judgement and accountability. A bot can assemble the file and flag what is missing, but the decision and its reasoning belong to an employee.
What happens when a banking system changes its screens?
The bot that depends on those screens may stop. That is why we use interfaces wherever they exist, build bots to fail safely and visibly rather than continue with wrong data, monitor every run and keep a person able to do the task manually. Changes to core systems are part of the bank's change process, so bot owners are told before a release.
Who owns the bots and the data?
The bank. Bot definitions, code and configuration sit in your repositories, the bots run in your environment or your cloud account in an EU region, and credentials stay in your vault. You can keep us for maintenance, run them yourselves or hand them to another team.
Get a written quote
Tell us in a few lines what you need. We reply within one working day with a written quote: scope, timeline and a fixed price or an hour estimate. Asking is free and commits you to nothing.