Quiz dotyczący sztucznej inteligencji w skarbcu przedsiębiorstw i prognozowaniu środków pieniężnych

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

1. Why is machine learning most useful in the direct method of cash forecasting rather than the indirect method?
   • The direct method is required by accounting standards for AI tools
   • The direct method projects actual receipts and disbursements, so models can learn from large volumes of transaction history
   • The indirect method cannot be used for horizons longer than a week
   • The direct method ignores bank data and relies only on budgets
2. A treasury team finds its AI model forecasts payroll no better than a simple scheduled rule. According to the guide, what does this suggest?
   • The model is broken and should be retrained daily
   • Payroll should be removed from the forecast entirely
   • For stable flows, simple rules often match or beat machine learning, and gains concentrate in volatile categories
   • The team should switch to the indirect method for payroll
3. How does invoice-level receipt modeling build a forecast of incoming customer cash?
   • It predicts a payment-date distribution for each open invoice and sums those distributions
   • It assumes every invoice is paid exactly on its due date
   • It forecasts only the total account balance and ignores individual invoices
   • It asks sales staff to estimate receipts each week
4. Why do practitioners decompose forecasts by cash-flow category and entity instead of modeling the total balance?
   • Regulators forbid forecasting consolidated balances
   • Totals mix predictable and noisy flows and can hide errors that cancel out
   • Category-level models require no historical data
   • Entity-level forecasts remove the need for currency conversion
5. How should known future items such as scheduled debt service or an approved payment run be handled in an AI cash forecast?
   • They should be predicted by the model like any other flow
   • They should be excluded from the forecast
   • They should be treated as fixed inputs rather than predicted
   • They should be averaged with last year's values
6. Which is the best way to judge whether an AI cash forecast is performing well?
   • Checking whether it produces a single precise number each day
   • Backtesting it against actual results by category and horizon
   • Comparing it with the forecast business units submitted
   • Measuring how quickly the model runs
7. In payment optimization, what decision might a model make about supplier invoices?
   • Which invoices to dispute automatically with the supplier
   • Which invoices to pay early to capture discounts while keeping a minimum cash buffer
   • Which suppliers to replace with cheaper competitors
   • Which invoices to reclassify as capital expenditure
8. According to the guide, what is frequently the hardest part of building an AI cash forecasting system?
   • Choosing between ARIMA and gradient-boosted trees
   • Getting executive approval for the software budget
   • Reconciling bank statement formats, ERP extracts and currency conversion before modeling
   • Training staff to read forecast charts