Sign in
Home/Salesforce/Certified AI Associate/Free questions

Certified AI Associate — Free Practice Questions

10 free sample questions from a bank of 77, with the correct answers and explanations. No signup required — start practising right now.

1Which type of bias imposes a system's values on others?
  • Association
  • Automation
  • Societal
Answer: B

The short version

B — Automation bias imposes the system values. Salesforce Trailhead defines automation bias with exactly the stem wording, so banked B is kept over the C challenge.

Key concepts in this question

  • Automation bias: system values imposed on others through biased training outcomes.
  • Association bias: data labeled according to stereotypes.
  • Societal bias: past prejudice toward social groups reproduced, such as redlining and zip-code proxies.

Why B is correct

Trailhead states automation bias imposes a system's values on others and illustrates it with the 2016 AI beauty contest trained mostly on one group, translating training bias into real-world winners. That matches the stem word-for-word, while societal bias is defined separately as reproducing past prejudice like redlining. Vendor prose naming C is therefore an error; banked B plus 70% community for B are confirmed on docs.

Why the others are wrong

  • A. Association bias is stereotyped labeling such as toys-for-girls versus toys-for-boys, not imposing system values on others.
  • C. Societal bias reproduces past prejudice toward social groups such as redlining and zip-code proxies, not the stem definition; Trailhead reserves the stem wording for automation bias.

Certified-AI-Associate exam tip

Memorize the Trailhead pair: automation bias equals imposes system values; societal bias equals reproduces past prejudice like redlining.

2A data quality expert at Cloud Kicks wants to ensure that each new contact contains at least an email address or phone number. Which feature should they use to accomplish this?
  • Validation rule
  • Autofill
  • Duplicate matching rule
Answer: A

The short version

A — Correct. A validation rule is the right feature to require email or phone on Contact.

Key concepts in this question

  • Data quality enforcement at point of entry
  • Validation rules vs autofill vs duplicate rules
  • Cloud Kicks Contact completeness requirement

Why A is correct

Validation rules enforce business logic on save and can require that at least one of Email or Phone is populated, blocking incomplete records.

Why the others are wrong

  • B Autofill: Autofill suggests values but does not enforce required data entry.
  • C Duplicate matching rule: Duplicate rules detect potential duplicates; they do not enforce field completeness.

Certified-AI-Associate exam tip

If the question asks to ensure, require, or block bad data on save, choose validation rule.

3In the context of Salesforce's Trusted AI Principles, what does the principle of Empowerment primarily aim to achieve?
  • Empower users of all skill levels to build AI applications with clicks, not code.
  • Empower users to solve challenging technical problems using neural networks.
  • Empower users to contribute to the growing body of knowledge of leading AI research.
Answer: A

The short version

A — Correct. Empowerment means clicks-not-code AI for all skill levels.

Key concepts in this question

  • Salesforce Trusted AI Principles
  • Empowerment principle definition
  • Democratizing AI beyond data scientists

Why A is correct

Salesforce defines Empowerment as abstracting AI complexity so people of all technical levels can build AI apps with clicks, not code.

Why the others are wrong

  • B Empower users to solve challenging technical problems using neural networks: This describes advanced deep technical work, not democratization.
  • C Empower users to contribute to AI research: This describes research contribution, not the Salesforce Empowerment principle.

Certified-AI-Associate exam tip

Memorize Trusted AI buzzwords: Empowerment = clicks not code for everyone.

4Cloud Kicks wants to use an AI model to predict the demand for shoes using historical data on sales and regional characteristics. What is an essential data quality dimension to achieve this goal?
  • Age
  • Reliability
  • Volume
Answer: A

The short version

A — Age is the essential quality dimension here. Historical sales and regional data must be current enough to predict demand, and only Age captures that timeliness aspect.

Key concepts in this question

  • Age / timeliness: how recent and current the data is for prediction
  • Reliability: general trustworthiness, not the listed quality dimension in this set
  • Volume: amount of data, a scale characteristic not a quality dimension

Why A is correct

Age reflects whether historical sales and regional characteristics are recent enough to predict future shoe demand. Stale data produces stale forecasts even if there is a lot of it. Among the choices, Age is the only recognized data-quality timeliness dimension.

Why the others are wrong

  • B. Reliability sounds important but is not the specific quality dimension being tested here; reliable old data can still mispredict current demand.
  • C. Volume describes how much data exists, not whether it is fit for prediction, so more rows alone do not ensure quality.

Certified-AI-Associate exam tip

Think timeliness = Age; if the question is about predicting current or future demand, pick recency over size.

5A financial institution plans a campaign for preapproved credit cards. How should they implement Salesforce's Trusted AI Principle of Transparency?
  • Communicate how risk factors such as credit score can impact customer eligibility.
  • Flag sensitive variables and their proxies to prevent discriminatory lending practices.
  • Incorporate customer feedback into the model’s continuous training.
Answer: A

The short version

A — Correct. Transparency means explaining how factors like credit score affect eligibility.

Key concepts in this question

  • Trusted AI Principle of Transparency
  • Explaining predictions and drivers
  • Preapproved credit card scenario

Why A is correct

Transparency requires communicating how models work and why a prediction was made, including how risk factors such as credit score impact eligibility.

Why the others are wrong

  • B Flag sensitive variables and proxies: This is fairness / bias mitigation, not transparency explanation.
  • C Incorporate customer feedback into continuous training: This is feedback loop / improvement, not transparency.

Certified-AI-Associate exam tip

Transparency = explain how and why the model decided; look for communicate, explain, disclose.

6What is a key challenge of human-AI collaboration in decision-making?
  • Leads to more informed and balanced decision-making
  • Creates a reliance on AI, potentially leading to less critical thinking and oversight
  • Reduces the need for human involvement in decision-making processes
Answer: B

The short version

B — Over-reliance on AI erodes critical thinking. Human-AI collaboration helps decisions, but the key risk is humans deferring too much to the model.

Key concepts in this question

  • Human-AI collaboration: humans plus AI making decisions together
  • Automation bias / over-reliance: trusting AI outputs without enough oversight
  • Critical thinking and oversight: human review that must be preserved

Why B is correct

Effective collaboration requires humans to question, validate, and oversee AI suggestions. When teams rely on AI by default, they exercise less scrutiny and miss errors or bias. That loss of oversight is the central challenge described by Salesforce trusted-AI guidance.

Why the others are wrong

  • A. More informed and balanced decision-making is a benefit of good collaboration, not a challenge.
  • C. Reducing need for human involvement describes full automation, not the collaboration risk of weakened human judgment.

Certified-AI-Associate exam tip

Challenge = over-reliance; benefit = balanced, informed decisions. Do not swap them.

7Which best describes the difference between predictive AI and generative AI?
  • Predictive AI and generative AI have the same capabilities but differ in the type of input they receive; predictive AI receives raw data whereas generative AI receives natural language.
  • Predictive AI uses machine learning to classify or predict outputs from its input data whereas generative AI does not use machine learning to generate its output.
  • Predictive AI uses machine learning to classify or predict outputs from its input data whereas generative AI uses machine leaning to generate new and original output for a given input.
Answer: C

The short version

C — Correct. Predictive classifies or forecasts; generative creates new original content.

Key concepts in this question

  • Predictive AI vs generative AI
  • Both use machine learning
  • Classify/predict vs generate content

Why C is correct

Both use machine learning, but predictive AI classifies or predicts outputs from input data, while generative AI generates new and original output for a given input.

Why the others are wrong

  • A Same capabilities differing only in input type: Wrong; capabilities differ fundamentally, and both can take varied inputs.
  • B Generative AI does not use machine learning: Wrong; both predictive and generative AI use machine learning.

Certified-AI-Associate exam tip

Predictive = forecast future; Generative = generate content. If both ML mentioned, pick that option.

8What is machine learning?
  • A data model used in Salesforce
  • AI that can grow its intelligence
  • AI that creates new content
Answer: B

The short version

B — Correct. Machine learning is AI that can grow its intelligence.

Key concepts in this question

  • Salesforce definition of machine learning
  • Learning/growing intelligence vs data model vs generation

Why B is correct

Salesforce Trailhead defines machine learning as the idea that machines can grow their intelligence from data.

Why the others are wrong

  • A A data model used in Salesforce: Too narrow; ML is a capability, not a single data model.
  • C AI that creates new content: This defines generative AI, not machine learning generally.

Certified-AI-Associate exam tip

Memorize the Salesforce phrasing: ML = grow intelligence; Generative = create content.

9The Cloud Kicks technical team is assessing the effectiveness of their AI development processes. Which established Salesforce Ethical Maturity Model should the team use to guide the development of trusted AI solutions?
  • Ethical AI Prediction Maturity Model
  • Ethical AI Practice Maturity Model
  • Ethical AI Process Maturity Model
Answer: B

The short version

B — Use the Ethical AI Practice Maturity Model. It is the established Salesforce model for guiding trusted AI development.

Key concepts in this question

  • Ethical AI Practice Maturity Model: Salesforce framework for maturing ethical AI practices
  • Trusted AI solutions: AI built with responsibility, accountability, and ethics
  • Development processes: how teams assess and improve AI delivery

Why B is correct

Salesforce AI Research published the Ethical AI Practice Maturity Model specifically to help teams assess and guide ethical development. The Cloud Kicks technical team would use that named model, not a generic prediction or process variant. It aligns assessment of effectiveness with trusted-AI practices.

Why the others are wrong

  • A. Ethical AI Prediction Maturity Model is not the established Salesforce model name; prediction is too narrow.
  • C. Ethical AI Process Maturity Model sounds plausible but is not the official Salesforce title.

Certified-AI-Associate exam tip

Memorize Practice — Ethical AI Practice Maturity Model is the exact exam phrase.

10What is a potential source of bias in training data for AI models?
  • The data is collected in real time from source systems.
  • The data is collected from a diverse range of sources and demographics.
  • The data is skewed toward a particular demographic or source.
Answer: C

The short version

C — Skewed training data creates bias. Data concentrated on one demographic or source teaches the model that narrow view.

Key concepts in this question

  • Training-data bias: systematic skew learned by the model
  • Representativeness: covering diverse sources and demographics
  • Real-time collection: freshness, which alone does not prevent bias

Why C is correct

If training data overrepresents one demographic or source, the model optimizes for that group and performs poorly or unfairly for others. Bias comes from lack of representativeness. A skewed sample is therefore a classic potential source of bias.

Why the others are wrong

  • A. Collecting data in real time affects freshness and latency, not whether groups are fairly represented.
  • B. Collecting from diverse sources and demographics reduces bias; it is the mitigation, not the source.

Certified-AI-Associate exam tip

Skewed = bias; diverse = fix. Pick the skewed option when asked for bias source.

Want the full bank of 77 questions for Certified AI Associate? See all practice exams.