AI on a real project: In brief
Lesson 5 closes the track’s first arc, the four lessons on using generative AI well. It adapts the case-study session that closes the same unit of the Harvard Kennedy School course this track is built on: everything the unit taught, prompt anatomy, tailoring, the decision criteria, run end to end on a single communications campaign. That session ends with an invitation to take the method to a situation “closer to what you do in your day-to-day life,” and this lesson accepts it. Our project is a job search, with a detour to a small nonprofit near the end.
Nothing new arrives in this lesson, and that is the point. The tools from lessons 2 through 4 stop being a collection and start being a method, run on one project with real stakes. The project’s three readers set the problem: a recruiter who opens your resume with dozens more waiting and may give it less than a minute, reading a document quite possibly filtered through screening software first; a hiring manager who reads your cover letter slowly, hunting for evidence you understand the actual work; and a former coworker deciding on the spot whether to answer a two-line note. One person, one true story, three completely different formats.
The capability: after this lesson, you can name the audience, platform, goal for every piece a project produces and vary the shape while the facts stay fixed; hand the drafting to the assistant you built in lesson 3 and have the machine interview you about the task before it does the task; keep the line-by-line verification of every claim about yourself, and lesson 4’s privacy question, in your own hands; rehearse turn by turn against an interviewer persona tuned from friendly to skeptical; and hold the line between polish and fabrication under pressure.
What the lesson covers. First the method: three questions before every piece. Who will read this? Where will they read it? What should happen because they read it? Then the handoff. The task passes lesson 4’s first filter loudly, and the session’s best tip turns the context problem around: state the goal, then add “please ask me what you need to know to do this task well.” Then the checks that stay yours. The machine quietly promotes you, helped with becomes led, because confident application language is what its training text is full of; it is not lying, it is predicting, and you strike it. Then rehearsal: one question at a time, aimed at your actual seams, with the persona’s pressure tuned to the preparation you want. The course’s own classroom split on how capable that simulation is, and the session’s teacher sides with caution: use the machine coach as a complement, and keep one live mock interview with a human who can hear you. Then the method travels to a three-person food pantry, unchanged. The lesson closes on the spectrum from legitimate to hard to defend, and the bright line: after reading, does this person believe anything false about you?
Why this order. Lessons 2 through 4 delivered the skills separately, each with its own practice; a case study is how separate skills become one working method. The practice runs the whole method in Clawless on your own project: name the role, shape two pieces for two different readers, make one explicit which-pieces call, then rehearse until the interviewer finds a question you cannot answer yet. Lesson 6 opens the track’s second movement, from how do I use this well to what it all means for your world, starting with the kinds of risk this technology carries and the levers for managing them.