Free Basics of AI practice questions

Ticket summarization takes a long chat log or email thread and condenses it into a brief overview of the reported problem, prior actions, and current status. It is one of the safer, higher-value uses of AI because it works from information you provide rather than the model's general knowledge, making it especially useful at shift handoffs and escalations. Code assistance refers to generating, explaining, or debugging scripts and utilities, not summarizing tickets. Virtual agent deflection describes customer-facing chatbots on a help portal that answer common questions such as password resets before a ticket reaches a human, which is a different function than summarizing an existing ticket. Tone adjustment is a drafting and communication capability that makes a message more polite or more approachable, not one that summarizes conversation history. Even with summarization, a technician should verify that no critical detail was dropped before acting and remember that ticket contents may include sensitive customer information.

Using an unapproved tool that bypasses the organization's security review is called shadow IT, and it is discouraged because it circumvents the vetting that determines whether a tool handles data safely; an organizational AI policy lists approved tools precisely to prevent it. Provenance describes the origin and handling history of data, not the act of installing an unsanctioned tool. Overgeneralization is an AI accuracy limitation, where an answer misses the specific edge case that matters, and does not describe a technician's tool choice. Triage is an AI-assisted support capability that suggests a category, priority, or queue for a new ticket, not a policy violation. The professional expectation is to use only sanctioned tools and, if the policy is unclear or a needed tool is not listed, to ask the manager, IT security, or the policy owner before acting.

AI chat assistants predict likely, useful text based on training patterns; they are not databases of verified facts and can produce confident but incorrect answers. The correct approach is to treat assistant output as a starting point or second opinion and confirm the directions against official documentation before acting. Applying the steps directly because the assistant sounded confident is exactly the trap the objective warns about, since confidence in wording does not indicate accuracy and the model may not know your specific environment. Forwarding the response to the customer as the solution passes unverified information along and makes the technician accountable for any errors it contains. Asking the assistant to make the steps sound more technical only changes the tone and does nothing to validate whether the steps are correct or safe. The recurring theme is that a human verification step should follow every AI step, and the technician remains responsible for the final outcome.

Using AI to complete assignments or answer questions meant to reflect your own knowledge is typically an academic-integrity violation, and many training programs, universities, and certification bodies now have explicit policies about AI assistance. Beyond the policy issue, it undermines the point of studying by producing a passing grade while leaving real knowledge gaps that appear on the job. Data stewardship concerns protecting the confidentiality of customer and company information and is not what is at stake when a learner shortcuts their own practice. A hallucination is an AI limitation in which the tool fabricates information; it is a property of the output, not a concern about how the student is using the tool. Ticket triage is an AI-assisted support function for categorizing and prioritizing incoming tickets. Honesty in learning protects the technician's own competence, and disclosure or outright prohibition of AI assistance is often mandatory in academic and certification settings.

When an application uses API integration, processing can happen on a remote server even though the feature looks built-in, which means text you enter may be sent to a third-party public service. Recognizing this drives the key privacy question of where your data actually goes, and that determines most responsible-use decisions. The color scheme of the interface is a cosmetic detail unrelated to how or where data is processed. Assuming API integration guarantees offline operation is incorrect, because calling an external service specifically requires network connectivity to reach the remote processing endpoint. Assuming the AI stores no data during processing is an unfounded conclusion; how a third-party service retains or uses submitted data depends on that provider's terms, which is precisely why the location of processing matters. The good habit is to ask whether text is processed locally on the device, on the organization's own servers, or sent to an outside service before feeding it any sensitive information.

Disclosure and transparency mean letting the relevant audience know that content was AI-assisted so they can weigh it appropriately, and some organizations require indicating when a customer-facing response was generated or assisted by AI. Failing to disclose where disclosure is expected can damage trust and may itself violate policy, even if the underlying work is sound. Provenance is about the origin and handling history of data, not about telling a customer how a reply was produced. Least privilege is a security principle about granting users only the access they need, which is unrelated to being honest about how content was created. Data minimization means collecting and sharing only the data a task requires; the technician's failure here is not about how much data was used but about concealing the AI's involvement. The guiding rule is that when in doubt it is safer to disclose than to conceal AI involvement, because concealment can be treated as dishonesty.

Accountability means the human remains responsible for anything they produce or act on, regardless of how much AI helped; the tool assists, but the person answers for the result. Saying the AI is responsible is not a valid defense, and stating that the AI wrote it does not excuse errors, infringement, or policy violations. Claiming the AI vendor assumes all liability is incorrect, because the technician and their employer bear responsibility for what is published and shared. Believing review is only required for scripts and not written documentation contradicts the expectation that a human verifies AI output before it is used or shared, which applies to reports, resolution notes, and knowledge-base entries alike. Assuming AI output is always accurate ignores the well-documented tendency of AI to produce confident but incorrect answers and to lack knowledge of a specific environment. Every AI step in a workflow should be followed by a human verification step.

Relying on AI for decisions that require licensed or human judgment is inappropriate use, and a technician should recognize when AI use touches regulated data and escalate to the appropriate people rather than deciding alone. Compliance determinations tied to regulations fall squarely into this category. Letting the AI make the determination because it processes large amounts of text confuses the ability to generate plausible text with the authority and accountability required for a regulated decision, which AI cannot provide. Posting the customer's data to a public AI to get a faster answer compounds the problem by exposing regulated information to a third-party service, violating data-stewardship duties and likely policy and law. Accepting AI output as final because it sounds authoritative ignores that AI can be confidently wrong and that presenting unverified output as fact is inappropriate use. A technician does not need to be a lawyer, but should identify when a task exceeds appropriate AI use and route it to the proper authority.

Using an approved organizational AI tool is the best choice because it keeps the technician within sanctioned tools and avoids shadow IT, even for non-sensitive content. For general, non-sensitive tasks like rewording a generic internal how-to article, an approved tool is perfectly acceptable and encourages good habits. Pasting the content into any free public AI service is not the best practice; while the content is not sensitive here, defaulting to unapproved public services builds a risky habit of bypassing approved tools. Refusing to use AI because all documents are confidential is incorrect because the scenario explicitly states the article contains no customer or proprietary data, so it is not confidential. Emailing the article to a coworker for manual rewriting is unnecessary and does not address the goal of efficiently using AI for a legitimate, low-risk task. The skill being tested is matching the tool to the sensitivity of the data and following organizational policy, and an approved AI tool fits a non-sensitive task well.

A hallucination occurs when an AI tool produces information that is false, fabricated, or unsupported by real sources while presenting it as if it were true. Generative AI is designed to produce plausible text, not verified text, so it may invent a command-line switch that does not exist while sounding authoritative; fluency is not the same as accuracy, and commands must be verified against official documentation before use. Provenance refers to the origin and history of data, not to fabricated output. Overgeneralization is a reliability limit where the AI gives a broadly reasonable answer that misses a specific edge case, but the switch here is not a real option at all. A knowledge cutoff explains an answer that is out of date, not one that describes something that never existed. Recognizing hallucinations and treating AI commands as unverified until confirmed is a core competency this objective tests.