Purpose
The course teaches bank staff to use generative AI tools safely for the reading, writing and checking work that fills most of their work: complaint responses, credit memo narratives, circular summaries, variance commentary, meeting notes, exception reports, and the data entry underneath all of it.
AI is treated as a drafting and analysis assistant working under a human owner. The course does not cover AI that decides anything: credit approvals, risk ratings, alert closures, or customer letters sent unread.
Staff are already using these tools on personal accounts, with no policy, no verification standard and no record. The question is whether that use becomes governed.
Who should attend
| Audience | Primary relevance |
|---|---|
| Branch staff: tellers, CSD, relationship officers | High-volume repetitive drafting and customer correspondence |
| Credit, operations, trade finance | Memo narratives, document extraction, exception handling |
| Compliance, AML, internal audit | Verification standards, and auditing the bank's own AI use |
| Finance, treasury, HR, planning, IT | Variance commentary, spreadsheet work, policy and circular drafting |
| Department heads | Attend with their teams |
Prerequisites
- No prior AI experience assumed. No participant needs to be technical.
- Cohort size: 20 to 25.
- A second facilitator is required above 20 participants.
- Not for the CEO and board. They can attend a separate briefing, and not this workshop.
Pre-workshop survey
A 10 mins Google Form is issued a week before the session.
| Section | What it collects |
|---|---|
| Role and workflows | Department, recurring tasks, approximate time spent on each, documents handled routinely |
| Pain points | The one task the participant would most like to speed up, and where the current process breaks |
| Current AI use | Which tools, how often, for what, on whose account. Optional; reported in aggregate only |
Owner: Bank HR in collaboration with the communication person.
Syllabus
Day 1
- 1
AI fundamentals and vocabulary
What these tools are and are not. The terms staff will meet, each defined once with a banking example. The three ground rules for the two days.
- 2
Capabilities and failure modes
What the tools do reliably against what they do badly. Live demonstration of a confident wrong answer. Nepal-specific traps: Bikram Sambat dates, lakh and crore, Devanagari output, Indian answers to Nepali questions.
- 3
Data handling: what goes where
A five-tier framework for what may be typed into a tool, applied to the bank's own document types. The habit of describing rather than pasting.
- 4
Prompting
One prompt pattern, drilled across three rounds of decreasing structure.
- 5
Context
What to feed a tool and what never to. Building a reusable context block for a recurring task.
- 6
AI slop
Recognising and removing hedging, filler and invented detail. Why this is a supervisory problem rather than a style problem.
- 7
The verification loop
Draft, critique, refine, verify. The four-question check, used at every gate for the rest of the course.
- 8
Documents and communication
Drafting from a facts skeleton, summarising, rewriting for a different reader, comparing versions, turning notes into minutes.
Day 2
- 9
Tools and data flow
Which tool suits which task. What projects and memory retain, and who can see them. Connectors a bank should refuse. Pilot cost.
- 10
Data entry, extraction and classification
Unstructured documents into clean tables. Classification and reformatting. Scanned and handwritten input. Printed fictional datasets throughout.
- 11
Spreadsheets and analysis
Formulas, data cleanup and variance commentary. Technique help against analysis of real data.
- 12
Ethics, governance and responsible use
Where the human gate sits and why accountability stays with the bank. Prompt injection, deepfakes and AI-enabled fraud. Labelling, audit trail and incident reporting. Individual escalation. Assistance against autonomous decision-making. NRB's draft AI guidelines as the governing position.
- 13
Function lab: your own deskOptional
Parallel departmental tracks on real workflows drawn from the pre-workshop survey, with planted errors. Every track ends at a human gate.
- 14
Workflow designOptional
Chatbot, workflow or agent, and when each is right. Four workflow patterns. The canvas, run on their own work.
Board and executive briefing
45-60 minsAudience: CEO, executive committee, board.
Covers what the tools do and do not do, the bank's current exposure from ungoverned use, the accountability position, and the four decisions the steering committee will be asked to make: approved tools, the data tier policy, the incident path, and ownership.
Workflows we cover
All scenarios use fictional data.Branch staff: tellers, customer service, relationship officers
| Workflow | Scenario | How AI helps |
|---|---|---|
| Replying to a customer complaint | A customer is upset that money sent from abroad hasn't arrived. | Writes a polite reply in Nepali and English for staff to check and send. |
| Explaining a new NRB circular | A new circular arrives from Nepal Rastra Bank. | Sums up in a few lines what counter staff need to do differently. |
| Checking account-opening forms | A handwritten KYC form comes in. | Reads the form and points out any missing or mismatched details. |
| Following up after a customer meeting | A relationship officer has rough notes from a client visit. | Turns the notes into a follow-up email and a to-do list. |
Credit, operations, trade finance
| Workflow | Scenario | How AI helps |
|---|---|---|
| Checking a loan file | A personal loan application with salary slips and supporting documents. | Pulls out the income details and lists any documents that are missing. |
| Summarising a borrower's credit history | A credit report (CIB) for a borrower. | Gives a one-page view of what they already owe and how well they have repaid. |
| Writing a credit memo | Key facts about a business applying for a loan. | Drafts the memo from those facts, ready for the credit officer to review. |
| Reading financial statements | A company's balance sheet and profit and loss statement. | Explains the key ratios and changes from last year in plain language. |
| Reviewing a hydropower loan | A project report for a hydropower loan. | Summarises the main risks and creates a checklist for the review. |
| Checking trade documents | A letter of credit with its invoice and shipping documents. | Compares the documents side by side and flags anything that doesn't match. |
Compliance, AML, internal audit
| Workflow | Scenario | How AI helps |
|---|---|---|
| Turning a new directive into a checklist | NRB issues a new directive. | Lists what the bank must now do and shows where current procedures fall short. |
| Comparing policy versions | A bank policy has been updated. | Shows what changed between the old and new version, in plain words. |
| Writing audit findings | An auditor has rough notes from a branch visit. | Turns them into clear findings with recommendations. |
| Checking how staff use AI | A record of how teams have used AI tools. | Checks it against the bank's AI rules and lists any gaps. |
Finance, treasury, HR, planning, IT
| Workflow | Scenario | How AI helps |
|---|---|---|
| Explaining budget differences | This month's budget vs actual figures. | Writes a short explanation for each line where the numbers are off. |
| Cleaning up a messy spreadsheet | A branch report with mixed dates, formats and duplicate rows. | Tidies the data and suggests the formulas needed. |
| Writing the daily treasury note | Today's interest rates and liquidity position. | Summarises the day's movements in two short paragraphs. |
| Drafting HR documents | A few bullet points from the HR head. | Drafts a staff circular or job description. |
| Writing an IT incident report | A timeline of what happened during a system outage. | Puts the events in order and drafts the cause and next steps. |
Department heads
| Workflow | Scenario | How AI helps |
|---|---|---|
| Turning meeting notes into minutes | Rough notes from a department meeting. | Produces clean minutes with decisions, owners and deadlines. |
| Preparing the monthly review | The department's monthly performance figures. | Summarises results and highlights what changed, for senior management. |
| Finding where AI can help the team | Staff answers from the pre-workshop survey. | Groups repetitive tasks and suggests which ones to try first. |
| Deciding on a request to use AI | A staff member asks to use AI for a task. | Checks what kind of data is involved against the bank's rules and lists what to ask before approving. |
Bring this workshop to your bank
Tell us about your organisation and who would attend, and we'll get back to you with dates and next steps.