Automated tasks
What Can Be Automated: A Practical Automated Task Reference
Most automation conversations stall at the abstract level. Teams know their processes are painful but can’t articulate what exactly should be automated. This framework solves that by mapping every major category of automatable task — with concrete examples from real operational environments. Use it as a discovery tool: read through each category and ask “do we do this manually today?”
How to Use This Framework
A Discovery Tool, Not a Shopping List
Every example in this framework is something an organisation is doing manually somewhere right now. The goal is not to automate everything — it is to make the manual work visible so the right things get prioritised. Read each category with your own operations in mind. Mark the ones your team does today. Then score them for volume, error rate, and business impact. That list is your automation backlog.
Category 01
Data Tasks
Data tasks are the most automatable category — and the most consistently underestimated. Any task that involves moving data from one place to another, transforming it, validating it, or comparing it against another dataset can be automated. If a human is doing it with copy-paste or re-keying, it can be automated.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| Data entry | Reads data from a source (form, email, file, system) and writes it into a target system without human intervention | A new client submits an onboarding form. The automation reads every field and creates the client record in the CRM, the compliance system, and the core banking platform simultaneously — with no manual re-keying. |
| Data transfer between systems | Moves data from System A to System B on a trigger or schedule — replacing the manual export-then-import process | Every evening, a treasury team exports a position report from the trading platform and manually imports it into the risk system. The automation connects the two systems directly and transfers the data at close of business without anyone touching it. |
| Data validation | Checks incoming data against defined rules — format, range, completeness, consistency — and flags or rejects records that don’t comply | A loan application arrives with a missing employer name and a date of birth that doesn’t match the ID document. The automation flags both fields before the application reaches the credit team, preventing a review session based on incomplete data. |
| Data reconciliation | Compares two datasets, identifies discrepancies, and either resolves them automatically or routes them for human review | A bank’s operations team spends three hours every morning reconciling the previous day’s payment confirmations against internal records. The automation runs at 07:00, matches every transaction, highlights the 12 discrepancies, and assigns them to the right analyst — reducing the process from 3 hours to 20 minutes. |
| Data enrichment | Takes an incoming record and augments it with data from external sources — credit bureaux, sanctions lists, company registries — before passing it forward | A new corporate client record is created. The automation queries the company registry, pulls the beneficial ownership structure, checks the entity against a sanctions database, and adds a risk classification — all before the relationship manager sees the record. |
| Data aggregation and reporting | Collects data from multiple sources, aggregates it to the required level, and produces a report — on a schedule or on demand | A compliance team manually compiles a daily regulatory position report from six different system extracts. The automation runs at 17:30, pulls every extract, aggregates to the required format, and emails the report to the compliance officer — with a flag if any source data is missing. |
| Duplicate detection | Scans incoming records against existing data to identify potential duplicates before they are created | A client is onboarded through a second channel without the team realising they already exist in the system. The automation checks name, date of birth, and ID number against the existing client database and alerts the team before a duplicate record is created. |
Category 02
Document Tasks
Banking operations run on documents. Most teams spend significant time either extracting data from documents or generating new documents from data — both of which can be automated. Intelligent Document Processing handles the extraction side; template-based generation handles the creation side.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| Field extraction from incoming documents | Reads an incoming document — PDF, image, scanned form — identifies the relevant fields, extracts the data, and passes it to the downstream process as structured data | A supplier invoice arrives by email as a PDF. The automation reads the invoice number, supplier name, line items, amounts, VAT, and bank details — regardless of the supplier’s template format — and creates the payable record in the finance system. No manual data entry, no re-keying errors. |
| Document classification | Receives a document and identifies what type it is — then routes it to the correct process or queue | A private bank receives a batch of client documents: some are passport scans, some are utility bills, some are bank statements, some are corporate documents. The automation classifies each one and routes it to the correct verification workflow — without a team member opening each document manually. |
| Document generation from a template | Takes a data record and populates a pre-defined template to produce a complete, formatted document — without manual drafting | An HR team sends 40 offer letters per month. Each one requires pulling the candidate’s name, role, salary, start date, and reporting line from the HRIS and inserting them into the standard letter template. The automation generates each letter in seconds from the approved data — branded, formatted, and ready to send. |
| Contract and agreement generation | Generates client-facing legal documents from approved clause libraries and live deal data — reducing drafting time and eliminating manual copy errors | A wealth management team generates a discretionary mandate for every new client. The automation pulls the client’s name, risk profile, fee structure, and investment mandate from the onboarding record and populates the standard agreement template — producing a document that requires only a legal review before signing. |
| Document expiry monitoring | Tracks documents with expiry dates — passports, licences, certifications, insurance certificates — and triggers renewal alerts or workflow steps automatically | A compliance team manually tracks 800 client passport expiry dates in a spreadsheet. The automation monitors every expiry date and sends a renewal request to the client 90 days, 60 days, and 30 days before expiry — with an escalation to the relationship manager if no response is received. |
| Document delivery and archiving | Sends a generated document to the right recipient by the right channel and stores it in the correct location in the document management system — automatically | After a loan is approved, the automation generates the approval letter, emails it to the client, stores a copy in the client’s digital file, and updates the loan management system with the dispatch timestamp — all without anyone from the operations team intervening. |
Category 03
Notification and Communication Tasks
A significant portion of manual communication in operations teams exists only because no system is tracking status and triggering the right message at the right moment. Every notification that is sent manually because “the system doesn’t do it” is automatable.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| Status update notifications | Sends a notification to the client or internal team every time a process reaches a defined milestone — without anyone manually drafting or sending the message | A client submits a loan application. They receive an automated acknowledgement immediately. When credit review begins, they receive an update. When the decision is made, they receive the outcome — all without the relationship manager sending a single email. |
| SLA breach alerts | Monitors how long each task or case has been open and sends an alert — to the assigned team or their manager — when a deadline is approaching or has been missed | A KYC review must be completed within 5 business days. The automation monitors every open case and sends an alert to the analyst at day 3 and an escalation to the team lead at day 4. By day 5, an unreviewed case is automatically escalated to the compliance manager with a full case history attached. |
| Approval request routing | When a task requires a decision, sends the request to the correct approver based on defined rules — amount, risk level, product type — rather than relying on an email chain | A payment instruction above €250,000 is submitted. The automation identifies the threshold, sends an approval request to the authorised signatory for that amount tier, and tracks the response. If not approved within 4 hours, it escalates to the backup approver. |
| Reminder and follow-up sequences | Sends a sequence of follow-up communications on a defined schedule when an action has not been completed — stopping automatically when the action is taken | A client is asked to provide an updated proof of address for their annual KYC review. They don’t respond. The automation sends a reminder after 7 days, a second reminder after 14 days, and notifies the relationship manager after 21 days — then stops as soon as the document is received. |
| Regulatory and compliance notifications | Generates and sends required regulatory communications — client disclosures, transaction confirmations, statutory notices — at the correct time and in the required format | Every trade confirmation must be sent to the client within 24 hours of execution. The automation generates the confirmation from the trade data, attaches the required regulatory disclosures, sends it to the client’s registered email, and logs the delivery confirmation — meeting the regulatory requirement without any manual step. |
Category 04
Approval and Workflow Tasks
Approval processes are where most operational delay lives. A step that takes 10 minutes to complete takes 2 days because nobody knows it’s waiting, the right person isn’t available, or the routing happened by email. Automated approval workflows eliminate the routing problem entirely.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| Multi-level approval routing | Routes a request to the correct approver — based on amount, risk, product, or entity — and moves to the next level automatically when each approval is given | A credit limit increase request is submitted. The automation routes it to the relationship manager for recommendation, then to the credit committee for decision, then to operations for implementation — with each step tracked, timed, and escalated if not actioned within the defined window. |
| Conditional routing | Assesses the characteristics of a request and routes it to different teams or processes depending on the outcome of defined conditions | A new account opening request arrives. The automation assesses the client type (individual vs. corporate), jurisdiction, and risk score — then routes it to the standard onboarding queue, the enhanced due diligence queue, or the PEP review queue based on the combination of those three factors. |
| Parallel approval steps | Sends a request to multiple approvers simultaneously — rather than sequentially — and waits for all required approvals before moving forward | A new product launch requires sign-off from Legal, Compliance, Risk, and Operations. Instead of routing sequentially — which takes weeks — the automation sends to all four simultaneously and moves to implementation as soon as all four have approved. Weeks become days. |
| Delegated authority management | Applies the current delegated authority matrix automatically — routing to the right level based on current parameters — and adjusts when the matrix is updated | When a senior approver is on leave, their requests don’t pile up in an inbox. The automation applies the active delegation rule, routes to the designated alternate, and returns to the primary approver when the delegation period ends — without any manual intervention. |
| Four-eyes control enforcement | Ensures that defined transactions or decisions are always reviewed by a second person — not as a manual policy, but as a technical control embedded in the workflow | Any payment above a defined threshold cannot be released by the person who entered it. The automation enforces this rule by requiring a second named authoriser to review and approve before the payment instruction is transmitted — making the four-eyes control structural rather than behavioural. |
Category 05
Monitoring and Exception Tasks
Most monitoring in manual organisations is reactive — something goes wrong and then someone notices. Automated monitoring is proactive — the system watches continuously and alerts before the problem becomes an incident. This category covers everything a human currently checks periodically that a system could watch continuously.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| Threshold and limit monitoring | Watches a value continuously and triggers an action — alert, escalation, or automatic intervention — when it crosses a defined threshold | A corporate client’s daily payment volume approaches their credit limit. The automation monitors the running total in real time and alerts the relationship manager when the client reaches 80% of their limit — before a payment is rejected, not after. |
| Compliance check automation | Runs defined compliance checks — sanctions screening, AML rules, KYC status verification — automatically at defined trigger points rather than on a sampling basis | Every new transaction counterparty is screened against the current sanctions list before the payment is processed. Not sampled — every one. The automation runs the check in seconds, passes clear counterparties through, and routes flagged ones to the compliance team with the match details attached. |
| Deadline and regulatory calendar monitoring | Tracks all regulatory submission deadlines, report due dates, and review obligations — and triggers the preparation workflow automatically in advance | A regulatory report is due to the central bank on the 15th of each month. The automation triggers the data collection process on the 8th, sends reminder notifications to data owners on the 10th and 12th, compiles the report on the 13th, and routes it for sign-off on the 14th — with the submission automated on the 15th if approval is received. |
| Exception queue management | Collects all transactions or cases that could not be processed automatically, assigns them to the appropriate analyst based on type and workload, and monitors resolution time | During a payment processing run, 47 transactions fail validation. Instead of landing in a shared mailbox, the automation categorises each failure type, assigns each to the analyst responsible for that category, and tracks resolution time — escalating unresolved cases after 4 hours. |
| Data quality monitoring | Continuously checks incoming data against quality rules and alerts when error rates or missing field rates begin trending upward — before they become a problem | The input data from a key upstream system starts showing a 3% increase in missing reference fields over a two-week period. The automation detects the trend, alerts the data owner, and creates a ticket before the missing data causes downstream failures. Without monitoring, this would have been discovered during a reconciliation failure three weeks later. |
Category 06
Scheduled and Batch Tasks
Any task that runs on a fixed schedule — daily, weekly, monthly — and follows the same steps every time is a strong automation candidate. These are often the tasks that extend working hours or create a single point of failure when the person who runs them is absent.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| End-of-day processing | Runs a defined sequence of closing tasks — position calculations, reconciliations, report generation, system updates — automatically at close of business | A treasury team’s end-of-day process involves pulling positions from four systems, calculating net exposure, reconciling against the previous day, and sending a summary to risk management. This takes 90 minutes manually. Automated, it runs at 17:30 and delivers the summary by 17:45 — without anyone staying late. |
| Periodic report generation | Generates standard reports — management information, regulatory submissions, client statements, performance reports — on a defined schedule, pulling live data at the time of generation | A fund administration team produces 120 client portfolio statements on the last business day of each month. Each requires pulling holdings, valuation, transactions, and performance data from three systems and populating a branded template. The automation generates all 120 overnight and delivers them by 08:00 on the first of the month. |
| System maintenance tasks | Runs routine system hygiene tasks — archiving completed records, purging temporary files, updating reference data, resetting counters — on a schedule without IT intervention | Client records flagged as dormant for more than 7 years are automatically archived to the long-term storage system, removed from the active client view, and logged in the compliance record — in line with data retention policy, without a quarterly manual purge process. |
| Fee and charge calculation | Calculates fees, interest, charges, or commissions based on defined rules applied to current data — and posts the results to the relevant accounts or billing system | A private bank calculates management fees for 350 client portfolios at the end of each quarter. Each calculation uses a different fee schedule based on the client agreement. The automation applies each client’s specific schedule to their current AUM and posts the fee to their account — in minutes, not the three days the manual process takes. |
Category 07
System Integration Tasks
Every time a person manually moves data from one system to another — by exporting a file, re-keying values, or copying information between screens — they are performing a task that exists only because two systems are not connected. System integration eliminates this class of work entirely.
How system integration works: APIs and connectors. An API (Application Programming Interface) is a channel that allows two software systems to exchange data directly — without a human intermediary. When System A has an API and System B has an API, an integration can be built that moves data between them automatically, in real time or on a schedule. A connector is a pre-built integration component that handles the technical translation between two specific systems — so instead of building the integration from scratch, a BPA platform can use a connector that already knows how to talk to Salesforce, SAP, or your core banking system. The result: data entered once flows everywhere it is needed, automatically.
| Task Type | What the Automation Does | Real Example |
|---|---|---|
| CRM to core banking synchronisation | Keeps client data consistent between the relationship management system and the core banking system — without manual updates in both | A client changes their address in the CRM. The automation detects the change and updates the core banking record, the compliance system, and the document management system simultaneously — preventing the situation where the same client has three different addresses across three systems. |
| ERP to payment system integration | Takes approved payment instructions from the ERP and transmits them to the payment system — eliminating the manual re-entry step that introduces errors and delay | After an invoice is approved in the finance system, the automation creates the payment instruction in the payment platform, applies the correct payment rail based on currency and destination, and transmits it for processing — without a payments team member re-keying the details. |
| Legacy system data extraction via RPA | Where a legacy system has no API, RPA bots log into the system, navigate to the required data, extract it, and pass it to the target system — mimicking what a human would do, but without the human | A regulatory reporting system built in 2003 has no API. A team member logs in daily, navigates to three different screens, copies position data, and pastes it into the reporting tool. An RPA bot performs exactly the same sequence — every day at 06:00 — without any human involvement, and with zero transcription errors. |
| Third-party data feed integration | Connects to external data providers — market data vendors, credit bureaux, sanctions list providers, company registries — and pulls current data automatically at the point it is needed | Every new corporate client triggers an automatic query to the company registry, pulling the current registered address, directors, and filing status. The data is added to the client record without the compliance analyst manually searching the registry website and copying the results. |
| Cross-entity data consolidation | Aggregates data from multiple entities or subsidiaries into a consolidated view — automatically, on a schedule — replacing the manual process of collecting and combining spreadsheets | A banking group with entities in five jurisdictions produces a daily consolidated liquidity report. Each entity’s treasury team previously sent a spreadsheet to the group treasury function, where someone spent two hours combining them. The automation pulls data directly from each entity’s system at 07:00 and delivers the consolidated report by 07:15. |
Where to Start
How to Turn This List Into an Automation Backlog
Step 1: Identify your candidates
- Go through each category above and mark every task your team does manually today
- For each one, estimate: how many times per month? How long does it take each time?
- Note the error rate — how often does the manual version produce a mistake that requires correction?
- Note the risk — what happens if this task is delayed or done incorrectly?
Step 2: Score and prioritise
- Volume × time = total monthly manual hours. High numbers first.
- Error rate × volume = monthly rework burden. Automate to eliminate.
- Regulatory or compliance risk = automatic priority regardless of volume
- Client-facing impact = prioritise processes where failures are visible to clients
- Standardisation = only prioritise tasks that already run the same way every time
One rule before you start
Do not add a task to the automation backlog if the process it belongs to is not yet standardised. Automating an inconsistent process produces consistent errors. The sequence is always: standardise first, then automate. A task that runs differently depending on who handles it is a process design problem — not an automation opportunity yet.