ISACA AI Fundamentals Certification Sample Questions
The purpose of this Sample Question Set is to provide you with information about the ISACA Artificial Intelligence Fundamentals (AI Fundamentals) exam. These sample questions will make you very familiar with both the type and the difficulty level of the questions on the AI Fundamentals certification test. To get familiar with real exam environment, we suggest you try our Sample ISACA AI Fundamentals Certification Practice Exam. This sample practice exam gives you the feeling of reality and is a clue to the questions asked in the actual ISACA Artificial Intelligence Fundamentals certification exam.
These sample questions are simple and basic questions that represent likeness to the real ISACA Artificial Intelligence Fundamentals exam questions. To assess your readiness and performance with real-time scenario based questions, we suggest you prepare with our Premium ISACA AI Fundamentals Certification Practice Exam. When you solve real time scenario based questions practically, you come across many difficulties that give you an opportunity to improve.
ISACA AI Fundamentals Sample Questions:
01. A retailer's model predicts which orders will be returned, working from the customer's order history and the product category. The operations team has recently started recording the delivery method on every order, and believes it bears on returns.
What has to happen before the model can take delivery method into account?
a) The delivery method has to be present in a training set and the model retrained on it
b) The training settings have to be adjusted so that it will accept one more input
c) The delivery method can be sent with each order at prediction time, and the model will begin weighing it once enough orders carrying it have been scored
d) The delivery method can be shown alongside the model's output, so that operations staff read the prediction and the method together
02. A parcel logistics operator runs twelve regional depots. Each depot books casual staff a week ahead, and managers currently pick the number from last year's roster. The operator holds five years of daily parcel counts for every depot, together with the calendar, local event listings and weather records for the same period.
Which AI capability most directly addresses the staffing decision?
a) Anomaly detection across the depots' scanning logs
b) Quality inspection of arriving parcels
c) Demand forecasting of daily parcel volume
d) Document processing of the depots' booking forms
03. An agricultural insurer uses a model to flag which crop-damage claims are likely to be disputed. Whether a claim is actually disputed becomes known only after an assessment process that runs for nine to twelve months. What difficulty does this create for monitoring the model?
a) Its input fields cannot be monitored at all until the matching outcomes have been recorded
b) Its accuracy cannot be confirmed until long after the predictions were made
c) Its predictions become less precise the longer an individual claim stays open in the system
d) Its training data must be rebuilt from scratch whenever the assessment process changes length
04. A municipal housing department has deployed a model that scores benefit applications and automatically issues a denial letter whenever the score falls below a set threshold. An applicant who is denied loses assistance for the following quarter. Department staff currently see only a monthly summary of denials.
Which change would best put the principle of human oversight into practice here?
a) Lower the score threshold so that fewer applications fall into the automatic denial band each quarter
b) Publish the model's overall accuracy figure on the department's website each quarter
c) Retrain the scoring model more frequently and publish its denial rate each quarter
d) Require a caseworker to review and confirm each proposed denial before the letter is issued
05. An agricultural cooperative built a model to forecast weekly grain moisture from silo sensor readings. At evaluation it performs worse than the spreadsheet method the cooperative already uses. Several weeks of project time remain, and the sensors and historical readings are all still available.
What is the most appropriate response?
a) Find out why the model is performing badly before changing it
b) Conclude that grain moisture cannot be forecast from sensor readings and close the project
c) Retrain with different settings repeatedly until the evaluation numbers beat the spreadsheet method
d) Deploy the model beside the spreadsheet and let operators choose which forecast to trust each week
06. A public employment service runs a model that predicts which job seekers will find work within six months, using fields its case management system has collected for a decade. Remote hiring has reshaped the regional labor market, and the fields the model leans on — commuting distance, counts of local vacancies — now say little about who gets hired.
What is the most likely limitation of simply retraining the model on the last year of data?
a) Retraining rewrites the model's structure, and the service would have to revalidate every one of the fields it collects
b) A single year of records is too small a sample for a model of this kind to be retrained on
c) The available fields no longer carry the information the outcome depends on
d) The model will simply relearn its old behavior
07. An agricultural cooperative runs an image model that flags diseased crops from photographs its members upload. The model is retrained each season on newly submitted photographs. Accuracy on one particular disease collapses after a season in which a large batch of submissions arrived from a single unverified account, all of them labeled as healthy plants.
Which concern does this pattern most directly raise?
a) Model drift, because conditions in the field have gradually moved away from the data the model learned from
b) Data poisoning, because corrupted records entering the retraining set have taught the model a wrong rule
c) Overfitting, because seasonal retraining has fitted the model too closely to recent photographs
d) Prompt injection, because uploaded content reached the model without being checked
08. A manufacturer's predictive-maintenance model flags pumps that are likely to fail. Its warnings became unreliable three weeks ago. Investigation shows that a vibration sensor on one production line was recalibrated and now reports in different units, and that this line supplies about a third of the model's input readings.
Why would retraining on recent data be the wrong first response?
a) The recent data carries the miscalibrated readings, so retraining would learn the fault
b) A recalibrated sensor changes the physical meaning of its readings, so the model would need a new architecture rather than recent data
c) Retraining corrects errors in the model's outputs, never errors in its inputs
d) The remaining production lines still report correctly, so the overall input distribution has barely moved at all
09. A university is assembling training data for a model that will route student support requests. Some of the data comes from sources the team did not create and cannot see the history of. Which general safeguard most directly reduces the risk of poisoned training data?
a) Encrypt the training data while it is stored and while it is transferred between systems
b) Add more records from the same sources, so that weak examples are outweighed by good ones
c) Increase the number of training rounds so that the model settles on stable weights
d) Check where each batch of training data came from before it is used
10. A construction firm pays overtime under a written agreement: every hour worked beyond forty in a week is paid at one and a half times the base rate, and every hour on a public holiday at double. The payroll team enters hours from timesheets and wants the calculation done automatically.
Why is machine learning the wrong choice here?
a) Machine learning cannot handle arithmetic on hours and rates, because that needs an exact calculation
b) The firm would first have to convert each timesheet into one numeric score before a model could read it
c) The rule is already known exactly, so it can simply be programmed
d) Payroll data is confidential, and machine learning models cannot be applied to confidential records
Answers:
|
Question: 01 Answer: a |
Question: 02 Answer: c |
Question: 03 Answer: b |
Question: 04 Answer: d |
Question: 05 Answer: a |
|
Question: 06 Answer: c |
Question: 07 Answer: b |
Question: 08 Answer: a |
Question: 09 Answer: d |
Question: 10 Answer: c |
Note: For any error in ISACA Artificial Intelligence Fundamentals certification exam sample questions, please update us by writing an email on feedback@edusum.com.
