How ready are you for AI and automation in finance?

Ten real situations where AI and automation meet finance work: invoices, debtors, TDS, trial balances and scheduled emails. Most are hard. It tests whether you think like a builder, not which buttons you know.

10 questions: 2 easy, 2 medium, 6 hard · about 3 minutes
Your score in four areas: how AI works, building with AI, automation, testing and data safety
The exact concepts to learn next

No sign-up to start.

Question of

You want AI to help you build a tool that reads your clients' invoices. What do you share with the AI while building it?

A. A few real client invoices, so it learns properly B. Your accounting software login C. Dummy invoices with the same layout and made-up figures D. Nothing, because AI cannot read invoices

You automate debtor follow-ups: the ageing, the MIS and one reminder email per customer. Which step should a person still do?

A. Working out each debtor's ageing bucket B. Reviewing the reminder emails before they are sent C. Copying the dues into the MIS D. Sorting customers by days overdue

You ask an AI chatbot for the TDS rate on a payment, and it confidently gives you a wrong one. What is the most likely reason?

A. It is deliberately misleading you B. Its internet connection dropped C. TDS rates are kept confidential D. It predicts the most likely words from patterns instead of checking an authoritative source

Your collection system must email the week's overdue reminders every Monday at 9 am, even when nobody opens the sheet. What makes that happen?

A. A time-driven trigger in Google Apps Script B. An onEdit trigger on the sheet C. Conditional formatting on the overdue column D. A macro you run yourself every Monday

You ask AI to “make me a collection system” and get a messy tool that doesn't fit your data. Using the five-step method, what was missing?

A. A paid AI plan B. Clarity on the output, the input columns and the logic rules, written down before handing over to AI C. Knowing how to code D. A longer prompt asking the AI to try harder

You give AI the same prompt twice to build the same tool and get two slightly different versions of the code. Why?

A. The AI remembers your first attempt and changes it on purpose B. Your internet speed affected the answer C. The model generates text by sampling likely next tokens, so each run can differ D. The prompt was saved incorrectly

Your AI-built tool turns a trial balance into a P&L and a balance sheet. Which check best proves that every ledger was mapped?

A. Total debits equal total credits, and the balance sheet balances once the P&L result is carried to equity B. The charts look reasonable C. The AI confirms that the mapping is correct D. The Excel download opens without an error

Your AI invoice reader has pulled 200 scanned purchase invoices into Excel. What is the best control before you post them?

A. None, because AI extraction is accurate B. Re-run the extraction until the output looks clean C. Post them and fix any differences at the GSTR-2B reconciliation D. Check that each invoice's lines add up to its total and that the tax equals rate × taxable value, and send mismatches for review

Your sheet lists 300 customers and their unpaid invoices. A script must email each customer only their own unpaid invoices. Which design is right?

A. One email to all 300 customers in BCC, with the full list attached B. Attach the whole debtors ledger as a PDF to every email C. Loop through the customers; for each, filter their unpaid rows, fill an email template with them, and send it from Apps Script D. Keep one draft and copy-paste each customer's figures into it by hand

A scheduling mistake runs your Monday reminder script twice. How do you design it so customers never receive duplicate emails?

A. Only ever run it by hand B. Delete the trigger after every run C. Write a 'last reminded on' date to each row after sending, and skip rows already reminded that week D. Send fewer emails in each run