4.0 Operational Procedures

AI integration means using artificial intelligence inside real workflows to analyze information, automate tasks, support decisions, and improve communication. The opportunity is powerful, but so is the responsibility: AI should support professional work, not replace judgment, honesty, privacy, or accountability.

For Objective 4.10, the goal is to understand basic AI concepts well enough to recognize appropriate and inappropriate use. AI tools can process large amounts of data, find patterns, summarize content, translate language, assist with troubleshooting, and help users complete routine work faster. In a support environment, this can make technicians more efficient and help organizations make better decisions.

At the same time, AI can be misused. A tool that can generate text, code, images, audio, or video can also be used to deceive people, impersonate someone, plagiarize work, or create content that violates policy. A professional technician should treat AI output as a starting point that must be reviewed, verified, cited when required, and used according to organizational rules.

Appropriate AI Use

Appropriate AI use means applying AI to solve real problems, improve efficiency, and support better outcomes without violating privacy, policy, ethics, or ownership rules. The infographic groups appropriate use into four major areas: data analysis, automation, healthcare support, and communication. These examples show how AI can assist people by handling scale, repetition, and pattern recognition while humans remain responsible for final decisions.

Data Analysis

AI can analyze large amounts of information from multiple sources much faster than a person could review manually. In IT and security work, this can include logs, alerts, tickets, system performance data, endpoint telemetry, inventory records, or other structured and unstructured information. AI can help identify patterns, connections, trends, and anomalies that may not be obvious at first glance.

For example, AI may review massive log files to detect unusual activity or possible security threats. It might notice repeated login failures, unusual access times, abnormal network behavior, or changes that match previous incidents. This does not mean the AI has proven an attack by itself. It means the tool can help surface clues so a technician or analyst can investigate more efficiently.

1. Find patterns across data types. AI can compare information from logs, tickets, events, reports, and other sources.
2. Support threat detection. AI can help flag unusual activity that may require review by a technician or security analyst.
3. Improve decision support. AI can summarize large data sets so humans can focus on judgment and next steps.
4. Require verification. AI findings should be checked against evidence before action is taken.

Automation

AI can help automate repetitive tasks and reduce manual effort. Automation may identify problems, recommend next steps, or take corrective action when the task is well understood and approved. This can help technicians spend less time on routine work and more time on troubleshooting, planning, user support, and higher-value tasks.

Examples include automatically categorizing support tickets, suggesting knowledge base articles, detecting repeated error patterns, identifying systems that need attention, or assisting with routine workflow steps. In some environments, AI may help trigger corrective actions, but those actions should be controlled by policy, permissions, testing, and change management. Automation should not become uncontrolled change.

Healthcare Support

The infographic uses healthcare support as an example of AI assisting with clinical and administrative tasks. AI can help analyze patient data over time, check for possible drug interactions, assist with diagnostics, or support decision-making. These uses can be valuable because healthcare environments often involve large amounts of information and time-sensitive decisions.

However, healthcare also shows why AI must be used carefully. AI should support qualified professionals, not replace them. Medical data is sensitive, regulated, and high impact. Any AI-assisted recommendation must be reviewed within approved clinical, privacy, and compliance processes. In a broader IT exam context, this reinforces an important idea: the more sensitive the data and the higher the risk, the more carefully AI use must be controlled.

• Diagnostic support: AI may help identify patterns that support a professional review.
• Drug interaction checks: AI can help compare medication data against known risk patterns.
• Patient monitoring: AI can assist with watching trends over time and flagging concerns.
• Compliance matters: Sensitive data must be protected and handled only through approved systems.

Communication

AI can improve communication by translating text or speech, helping with grammar, improving clarity, summarizing long messages, and making content easier to understand. This can be useful for support technicians who need to communicate with users, write ticket notes, prepare instructions, or explain technical topics in everyday language.

Good communication support does not mean letting AI speak without review. A technician should check that the message is accurate, respectful, complete, and appropriate for the audience. AI can help draft or polish content, but the technician is still responsible for the final wording, technical accuracy, and tone.

Inappropriate AI Use

Inappropriate AI use includes actions that are harmful, dishonest, deceptive, or unethical. The infographic highlights three major misuse categories: fraud and impersonation, unethical shortcuts, and plagiarism. These are important because AI can create convincing content quickly, and convincing content can be abused.

Fraud and Impersonation

AI can be used to create realistic fake identities, generate deepfake audio or video, imitate writing styles, or produce content that appears to come from someone else. This creates risk for social engineering, scams, business email compromise, fake support requests, misinformation, and reputational harm.

A technician should be cautious when a request, voice message, video, email, or identity claim seems unusual or high risk. Verification through approved channels matters. If AI-generated impersonation is suspected, preserve evidence, avoid taking risky action, and escalate according to policy.

• Fake identities: AI can help create realistic profiles, photos, messages, or personas.
• Deepfake media: AI-generated audio and video can look or sound real.
• Social engineering risk: Attackers may use AI to make scams more convincing.
• Verification required: Sensitive requests should be confirmed through trusted channels.

Unethical Shortcuts

AI can be misused to take credit for work that is not your own. This might include generating application code instead of learning and writing it yourself, creating images and pretending they were human-designed, or submitting AI-created work without following disclosure requirements. In a professional environment, this can violate policy, damage trust, and create quality or licensing problems.

Using AI as a learning aid or productivity assistant can be appropriate when policy allows it. The line is crossed when a person hides AI use, misrepresents authorship, bypasses required learning, ignores review, or submits work they do not understand. A technician should be able to explain, verify, support, and take responsibility for the work they deliver.

Plagiarism

AI can produce content that closely resembles existing work or rewrites someone else's ideas without proper credit. Even when AI creates new wording, the user is still responsible for checking whether the output uses protected material, needs citation, or violates policy. This applies to reports, documentation, training content, code, images, scripts, and other deliverables.

Plagiarism risk is not solved by asking AI to paraphrase. If the underlying idea, structure, or source belongs to someone else, proper credit may still be required. In school, business, and certification prep environments, users should follow disclosure and citation requirements.

Bottom line: AI can help technicians analyze data, automate repetitive work, support specialized decision-making, and communicate more clearly. Use it responsibly: follow policy, protect sensitive information, verify results, give proper credit, avoid deception, and use AI to support your work instead of replacing integrity.

AI can be helpful and powerful, but it is not perfect. Its output depends on training data, model design, assumptions, and prompts, so technicians must understand bias, hallucinations, inaccurate information, and the need to verify results before using AI output for decisions.

For Objective 4.10, AI limitations are just as important as AI capabilities. A tool may sound confident, produce polished text, summarize large documents, or provide a detailed answer, but that does not guarantee it is correct. AI systems do not understand truth the same way a human expert does. They generate output based on patterns learned from data and the instructions they receive.

The technician mindset is simple: use AI as a helpful assistant, not as an absolute source of truth. When the information matters, verify it. When the output affects users, systems, safety, privacy, compliance, or business decisions, review it carefully and confirm it with trusted sources.

Limited to Training Data

AI systems learn from training data. That data may include books, websites, code, documentation, examples, images, or other information selected during the model-building process. The AI can only generate answers based on patterns it learned from that training process and any additional context provided by the user or connected tools.

If the training data is incomplete, outdated, misleading, low quality, or unrepresentative, the AI may produce incorrect conclusions. The infographic summarizes this with the phrase, garbage in equals garbage out. If the model learns from flawed information, the output may reflect those flaws. If the model lacks information about a new product, policy, vulnerability, or procedure, it may guess instead of saying it does not know.

1. AI does not truly know like a person. It predicts likely output based on patterns, not personal understanding.
2. Incomplete data creates weak answers. Missing or misleading training examples can lead to wrong conclusions.
3. Outdated data can be risky. AI may not know recent changes unless it has access to current information.
4. Verification is required. Important information should be checked against trusted sources, documentation, logs, or policy.

Bias in Data

Bias in data happens when the information used to train or guide an AI system does not fairly represent people, groups, events, devices, environments, or outcomes. If the data overrepresents one group and underrepresents another, the AI may perform better for one group than another. In high-impact areas, that can lead to unfair or inaccurate results.

The infographic gives a healthcare example: if medical data underrepresents a certain ethnic or gender group, an AI tool may provide less accurate support for that group. The same idea applies beyond healthcare. In IT, data bias could appear in support ticket patterns, hiring tools, security alerts, user behavior analytics, language translation, facial recognition, or automated decision systems.

Bias in Algorithms

Bias can also come from the way the AI is designed, programmed, tuned, tested, or deployed. Even when the data seems good, the model's assumptions and design choices can influence results. Developers may not intend to create bias, but bias can still enter through feature selection, weighting, labels, optimization goals, thresholds, or testing methods.

This matters because AI output can look objective. A score, recommendation, summary, or ranking may seem mathematical and neutral, but the system behind it may be shaped by hidden assumptions. Technicians should remember that AI results are not automatically fair just because they came from software.

• Design choices influence output: Model architecture, labels, and priorities can affect results.
• Bias may be unintentional: Developers may not realize how a system favors or disadvantages certain cases.
• Testing must be broad: AI should be evaluated across different groups, data types, and real scenarios.
• Policy still applies: AI-assisted decisions should follow organizational, ethical, legal, and compliance requirements.

Real-World Example: Amazon Hiring Tool

The infographic references Amazon's experimental hiring tool as an example of past data reflecting past bias. The tool used historical resumes to identify patterns, but the historical data came from an environment where many past resumes were from men. As a result, the system learned patterns that favored male candidates and penalized signals more associated with women.

This example is important because the AI was not necessarily trying to discriminate. It learned from historical patterns. If the past data reflects unfairness, the AI can repeat that unfairness at scale. A technician should understand the larger lesson: automated systems can continue old problems unless data, design, testing, and human oversight are handled carefully.

AI Hallucinations

An AI hallucination is when an AI confidently provides information that is incorrect, made up, unsupported, or misinterpreted. The answer may sound polished and convincing, but it can still be wrong. Hallucinations are especially risky because users may trust the tone of confidence instead of checking the facts.

The infographic gives a simple image example: if asked to identify a picture, AI might describe a cloud as a dog because it sees a snout, tail, and four legs. In text, hallucinations can show up as fake citations, wrong commands, incorrect troubleshooting steps, imaginary features, false summaries, inaccurate policy details, or confidently stated technical errors.

1. Confidence is not proof. AI can sound certain even when it is wrong.
2. False details can look real. Fake references, commands, steps, or product names may appear plausible.
3. Context can be misread. AI may misunderstand an image, log, ticket, policy, or user request.
4. Verification protects users. Important answers should be checked before being used in support or documentation.

Inaccurate Information in Real Products

Major AI products have produced inaccurate information in public examples, summaries, and answers. The infographic mentions Google Bard and Microsoft Bing Chat publishing article summaries that contained incorrect facts. The point is not that those tools are useless. The point is that even major companies' AI tools can be wrong while sounding confident.

This is a major exam concept: do not assume AI output is correct because the tool is popular, branded, or professionally designed. A technician should verify important information, especially when using AI for research, summaries, troubleshooting, legal or compliance content, security decisions, medical information, or anything that affects users and systems.

Measuring AI Accuracy

AI accuracy is often measured by comparing AI predictions or answers against known test data. If the expected answer is already known, reviewers can check whether the AI output matches. This is one common way to evaluate model performance, but it is not the same as proving the AI will always be correct in real life.

Accuracy scores can vary by task, data set, prompt, language, product version, and evaluation method. A model may perform well on one type of question and poorly on another. A technician should understand that accuracy is not a permanent guarantee. It is a measurement under specific conditions.

• AI makes predictions: It uses mathematical models to produce likely outputs.
• Predictions can be wrong: Even strong models make mistakes.
• Known test data helps measure performance: Comparing answers to known results helps evaluate accuracy.
• Scores have context: Accuracy depends on what was tested and how the test was performed.

Accuracy Study Results

The infographic includes accuracy study results from August 2024 showing fact-checking accuracy percentages across several models. The lesson is that AI accuracy can vary widely. Even the highest score shown is not perfect, and several models are much lower. That means users should expect AI to sometimes be wrong, even when it appears useful.

For an exam or workplace scenario, the exact percentage is less important than the pattern: AI output needs review. If a model is 70 percent accurate on a certain test, that still means many answers may be wrong. If a model is around 55 or 60 percent accurate, the need for careful verification is even clearer.

Key Takeaways for AI Limitations

Bottom line: AI can be impressive, but it has real limitations. It is limited by training data, can reflect bias, can be shaped by algorithm design, can hallucinate, and can produce inaccurate information even in polished products. Use AI skeptically, verify important results, and treat it as a support tool rather than a source of absolute truth.

Public AI and private AI can use similar underlying technology, but the environment changes the risk. Public AI is open to broad internet use, while private AI is built or controlled by an organization and can be designed to protect internal data, enforce policy, and limit how information is used.

For Objective 4.10, technicians need to understand that AI is not only about what the tool can do. It is also about where the tool runs, who controls it, what data it uses, what happens to user input, and whether sensitive information is protected. The same question asked in a public AI platform and a private organizational AI system can carry very different privacy and security implications.

The safest habit is to classify the data before entering it into an AI tool. Public AI may be useful for general research, drafting non-sensitive content, brainstorming, or learning broad concepts. Private AI may be more appropriate for internal documents, customer data, proprietary processes, support tickets, business records, or regulated information, but only if the organization has approved that use.

Public AI

Public AI refers to AI services that are available to the general public over the internet. Examples shown in the infographic include ChatGPT, Google Gemini, and Microsoft Copilot. These tools are designed for broad use and can help with everyday questions, research, summaries, writing support, coding help, translation, and general productivity.

Because public AI is easy to access, it can be tempting to paste anything into it. That is the risk. Information entered into public AI platforms may be processed, stored, reviewed, or used according to the provider's terms and settings. Even when a tool offers privacy controls, technicians should follow organizational policy instead of assuming the data is safe.

1. Available to the general public. Public AI tools can usually be reached from any internet-connected device.
2. No internal network required. Users do not typically need company VPN or internal access to use them.
3. Designed for broad use. They are helpful for general tasks, not necessarily for sensitive company workflows.
4. Policy still applies. Employees should not enter confidential or restricted data unless policy explicitly allows it.

Private AI

Private AI refers to an AI system built, hosted, configured, or controlled by an organization. It may run inside the organization's environment, through a private cloud deployment, or through an enterprise AI service with contractual and technical controls. The key idea is that the organization has more control over the model, data, access, logging, and acceptable use.

Private AI can use proprietary data such as internal documents, customer records, ticket histories, procedures, knowledge bases, product information, and company-specific workflows. This can make the AI more useful for internal work because it understands the organization's context. It also increases responsibility because the data may be sensitive or regulated.

Data Security

Data security is one of the biggest differences between public and private AI. Information entered into public AI platforms may not stay private in the way a user expects. Sensitive details such as passwords, encryption keys, certificates, customer records, private business plans, source code, security findings, regulated data, and internal procedures should not be entered into public AI tools unless the organization has specifically approved that workflow.

Private AI environments can help reduce the risk of exposing sensitive data because they can be built with access controls, approved data storage, retention policies, logging, and governance. However, private AI still needs careful design. If permissions are too broad, a user may retrieve information they should not see. If logs are not protected, prompts and outputs may expose sensitive details. If data sources are outdated or poorly classified, the AI may provide incorrect or inappropriate information.

• Do not enter secrets. Passwords, keys, certificates, tokens, and credentials should never be pasted into public AI tools.
• Protect customer data. Personally identifiable, financial, health, legal, or confidential customer information requires approved handling.
• Check the environment. Know whether the AI tool is public, private, enterprise-approved, or prohibited.
• Use approved tools. Follow organizational policy for where data can be entered and how AI output can be used.

Data Source

Data source means where the AI's information comes from. Public AI commonly uses public or broadly available information, depending on the tool and configuration. That may include internet content, websites, articles, documentation, public examples, and other open sources. Public data is useful for general knowledge, but it may be incomplete, outdated, inaccurate, biased, or not specific to the organization.

Private AI can use private data that belongs to one organization or entity. This can include internal files, customer records, tickets, proprietary documentation, policies, procedures, and other non-public information. This makes private AI powerful for company-specific support, but also creates access-control and privacy concerns. Users should see only the information they are authorized to access.

Data Privacy

AI tools can learn a lot about a user from the information they provide. The infographic lists examples such as location, habits, memberships, job applications, and more. Even if a user does not directly provide a full profile, repeated prompts can reveal patterns about where they live, what they do, what systems they use, who they interact with, what problems they have, and what they are trying to accomplish.

This is especially important when users enter personal data, work history, customer notes, ticket details, or application materials. AI systems may be able to infer sensitive information from small clues. A technician should consider whether entering information could create a profile of a person, customer, employee, applicant, or organization.

1. AI can collect large amounts of personal data. Prompts and files may reveal more than the user intended.
2. AI can infer details. Location, habits, memberships, job activity, and preferences may be inferred from context.
3. Profiles create risk. Data entered for one task may reveal patterns useful for profiling or targeting.
4. Minimize what you share. Use only the data needed, anonymize when appropriate, and follow approved procedures.

Choosing Public or Private AI

The decision is not simply public equals bad and private equals good. The correct choice depends on the data, task, policy, and risk. Public AI may be appropriate for general learning, drafting a non-sensitive email, explaining a generic concept, or brainstorming public-facing ideas. Private AI may be appropriate when the task involves internal records, company-specific procedures, customer support data, or proprietary information.

When in doubt, choose the more protective option and ask for guidance. Technicians should not experiment with sensitive data in public systems just because the tool is convenient. Convenience does not override data handling rules.

Technician Responsibilities

Technicians often help users understand tools, risk, and policy. If a user asks whether they can paste customer data into a public chatbot, the safest response is to stop and check policy. If a user wants AI to analyze internal logs, tickets, resumes, or customer records, the technician should confirm whether an approved private or enterprise AI workflow exists.

Technicians should also think about output. Even if the input was safe, AI output can still include sensitive data, incorrect conclusions, or content that should not be shared broadly. Output should be reviewed before being copied into tickets, reports, emails, documentation, or customer-facing messages.

Bottom line: Public AI is easy to access and useful for general tasks, but sensitive information entered into public tools may not stay private. Private AI gives an organization more control over data sources, access, and use, but it still requires governance. Know where the AI runs, what data it uses, what information you are entering, and whether policy allows it.

Objective 4.10 is about understanding basic artificial intelligence concepts and using AI responsibly. AI can be useful for analysis, automation, communication, and decision support, but technicians must understand its limits, protect data, verify output, and follow policy.

AI tools can feel impressive because they produce polished answers quickly. They can summarize long information, analyze patterns, draft messages, translate language, assist with code, and help automate repetitive tasks. The exam focus is not just knowing that AI is powerful. The key is knowing when AI use is appropriate, when it becomes risky, and what a responsible technician should do before trusting or sharing AI output.

A technician should treat AI as an assistant, not an authority. AI can help with work, but the human user remains accountable for accuracy, privacy, ethical use, and compliance. If an AI tool produces a recommendation, summary, report, image, code sample, or answer, the technician still needs to evaluate whether it is correct, allowed, secure, and properly credited.

Use AI for Appropriate Support

Appropriate AI use focuses on solving real problems and improving efficiency without violating policy or ethics. AI can help analyze large data sets, review logs, identify patterns, summarize documentation, draft user-friendly explanations, translate content, proofread communications, and assist with repetitive workflow tasks. These uses are helpful when the information is allowed to be used and the output is reviewed before action.

Avoid Inappropriate AI Use

Inappropriate AI use includes actions that are deceptive, dishonest, harmful, or against policy. AI should not be used to impersonate people, create deepfake audio or video for fraud, generate fake identities, plagiarize content, hide authorship, bypass required learning, or submit work the user does not understand. Even if AI can create something, that does not mean it is appropriate to use it.

1. Do not impersonate. AI-generated voices, images, videos, or writing styles can deceive people and support fraud.
2. Do not plagiarize. AI output may closely resemble existing work or rewrite someone else's ideas without proper credit.
3. Do not take unethical shortcuts. AI should not be used to claim credit for work, code, designs, or analysis that was not properly reviewed or disclosed.
4. Do not ignore policy. Organizational rules decide what AI tools, data, and disclosure practices are allowed.

Understand AI Limitations

AI is limited by the data it was trained on, the design of the model, the assumptions built into the system, and the information provided in the prompt. AI can produce wrong answers, outdated information, biased recommendations, inaccurate summaries, fake references, or confident statements that are not supported by evidence. This is why verification is a central Objective 4.10 skill.

AI hallucinations are especially important. A hallucination occurs when AI presents incorrect or made-up information as if it were true. The output may sound professional and detailed, but the confidence of the writing does not prove accuracy. Technicians should be skeptical when AI provides commands, procedures, citations, product details, policy summaries, security recommendations, or troubleshooting steps.

Protect Data and Privacy

Public AI and private AI may use similar technology, but they create different data risks. Public AI is available to broad internet users and is useful for general questions, drafting, learning, and non-sensitive research. Private AI is built or controlled by an organization and may use internal documents, customer records, ticket data, policies, or proprietary information. The environment matters because input data, output data, retention, access, and monitoring may be handled differently.

Sensitive information should not be pasted into public AI tools unless organizational policy explicitly allows it. This includes passwords, encryption keys, certificates, customer data, employee data, financial records, health information, private business plans, source code, support tickets, security findings, and anything regulated or confidential.

1. Identify the data type. Decide whether the prompt contains public, internal, confidential, regulated, personal, or proprietary information.
2. Choose the correct AI environment. Use public AI for appropriate general tasks and approved private or enterprise AI for sensitive or internal work.
3. Minimize data sharing. Provide only what is needed, anonymize when appropriate, and avoid unnecessary personal details.
4. Protect output too. AI-generated summaries, reports, and answers may contain sensitive or inaccurate information and should be reviewed before sharing.

Common Exam Traps

• Trusting AI output because it sounds confident or polished.
• Using AI-generated information in a ticket, report, or decision without verifying it.
• Pasting customer records, credentials, certificates, source code, or internal logs into a public AI tool.
• Assuming private AI is automatically risk-free without checking access controls, data sources, and logging.
• Ignoring bias in training data, historical records, or algorithm design.
• Using AI-generated content without citing, disclosing, or following plagiarism rules when required.
• Letting AI make high-impact decisions without human review.
• Believing public AI and private AI have the same privacy expectations.

Technician Mindset

Bottom line: AI is a powerful support tool, not a replacement for professional judgment. Use AI to improve analysis, automation, and communication, but remember its limitations. Protect sensitive information, understand public versus private AI risks, watch for bias and hallucinations, give credit when required, verify important output, and follow organizational policy every time.