What AI Can and Cannot Do – A Practical Guide By the end of this lesson, students should be able to: The core decision framework – four questions. Question 1: Is this task language-intensive? AI's strengths are primarily language-based – writing, synthesis, explanation, code. If the task is primarily about judgment, physical action, interpersonal relationship, or real-time data, AI adds less value. If it is primarily language-based, proceed to question 2. Question 2: Does this task require current or highly specific factual information? If yes – current events, live data, highly specific technical specifications, regulatory details – the answer is either (a) supplement AI output with verified current sources, or (b) do not use AI for the factual core of this task. AI handles well-established, general information reliably; specific and current information requires verification. Question 3: How severe are the consequences of an error? Low-severity, easily-correctable tasks (internal drafts, brainstorming, first outlines) tolerate AI's error rate well. High-severity tasks (legal filings, medical guidance, financial disclosures, safety protocols) require more stringent verification or human expert input, even if AI is involved in drafting. Question 4: Does this task require a judgment call that belongs to a human professional? If the task output is a professional recommendation, advice, or decision that carries accountability – and where your expertise is what the client, employer, or audience is paying for – AI is an input tool, not a decision maker. Human professional judgment is the required output. Quick reference: high-value AI use. Quick reference: use with caution. Quick reference: exclude or use minimally. A financial advisor uses the four-question framework to audit her practice. She maps fifty task types she performs regularly and classifies each. She discovers: (1) client communication drafts – high AI value (language-intensive, general, low severity, easily edited), (2) research on established investment principles – high AI value with verification for specifics, (3) specific regulatory compliance guidance – use with caution, verify from official sources, (4) investment recommendations to clients – exclude AI from recommendation itself; human professional judgment required and clients are paying for her expertise, not an AI's. The map takes two hours. It replaces six months of ad hoc decisions about when to use AI. The AI decision framework is also useful in reverse: when deciding whether to stop using AI for something you currently use it for. If a task has drifted into the "use with caution" or "exclude" category without you noticing – because AI started doing it conveniently – the framework helps you identify the drift before an error makes it obvious. The AI decision framework is a judgment tool, not a compliance checklist. The right answer for a specific task depends on your organization's policies, your professional obligations, your client relationships, and the specific stakes involved. Use the framework as a guide and apply your own professional judgment to the conclusion – especially in regulated industries where specific rules may govern AI use. Apply the four-question framework to your ten most common professional tasks. For each, go through all four questions and classify: high-value, use with caution, or exclude. Note any task where your current AI use does not match the classification. Start with the highest-consequence mismatches. You should be able to apply all four questions of the AI decision framework to any task, explain what each question is testing for, and have a personal AI use map for your most common professional tasks. Log in and enroll to access lesson quizzes.
Lesson 5: Choosing When to Use AI and When Not To
Lesson Objectives
Lesson Content
Practical Example
Lesser-Known Tip
Safety Notes
Practice Task
Completion Check