Bring Your Skills Together on One Meaningful Goal

You've reached the end of the course's individual techniques. The real test is whether you can combine them to finish one meaningful task without being guided through every decision.

Imagine a local club wants a short proposal for improving how it welcomes new members. It has a few invented survey comments, notes about the current process and a modest budget. There is no single magical prompt that completes the task perfectly. You need to understand the problem, organise evidence, develop options, create a usable document and decide what should be checked by a person responsible for the club.

Start with the outcome. A good proposal might explain the current difficulty, set out two possible improvements, compare their effort and produce a one-page recommendation. You do not need to build an application simply because the course has introduced AI-assisted development. The most capable choice is the smallest process that produces a trustworthy result. That could be an assistant, a spreadsheet and an editable document.

Gather the source material before asking for conclusions. Separate the comments of hypothetical members from established facts about the club. If one comment says the joining process took two weeks, that is one reported experience, not evidence that every person waits two weeks. Ask the assistant to organise patterns without inventing statistics. This applies the research and evidence skills you learned earlier.

Next, explore possibilities without immediately committing to a preferred answer. Perhaps the club needs a clearer welcome email, a checklist for volunteers or a shared calendar of introductory events. Ask the AI to explain which problem each option addresses and what trade-offs it creates. An idea with an impressive technology component is not necessarily more useful than a simpler operational change.

Turn the strongest options into a comparison. Record estimated effort, who would be responsible, what information is missing and how success could be measured. If the proposal includes a cost, verify the actual figure or label it as an assumption. The assistant can help structure the reasoning, but it should not turn unknown details into confident claims. Evidence and judgement have different roles.

Now produce a real deliverable. Create an editable document with a short context section, options, recommendation and unresolved questions. If the tool can create the file directly, inspect it. If it only produces text, move the approved content into a document editor. Open the file and confirm the headings and tables survive. A final artifact should be useful to somebody who wasn't present for the AI conversation.

Ask for a critical review before sharing. What assumptions are unsupported? Which statements rely on only one source? What could go wrong if the club acts on the recommendation? The assistant can help surface weaknesses, but the final judgement remains yours. If the proposal concerns a regulated or sensitive activity, it may need an authorised review beyond what an AI conversation can provide.

Decide how to know whether the work succeeded. Perhaps the proposal takes forty minutes instead of an afternoon and contains no invented facts. Perhaps the club trials one welcome checklist and measures the time required to complete registration. These are observable outcomes. The exercise is not about showing that AI participated in every stage. It is about finishing a good task more effectively.

At the end, write a short reflection. Which AI capability was most valuable? Where did you need to intervene? Which check caught a mistake? Which step would you keep if you repeated the work? That reflection turns a one-off success into a capability you can apply elsewhere. The strongest final project is one you can explain, verify and improve without the course holding your hand.

The capstone does not require the most ambitious project you can imagine. A clear, accurate proposal from invented notes is a stronger demonstration of ability than an unfinished application containing impressive screens. Choose a task small enough to complete, but substantial enough that it demands judgement. When the work is done, another person should be able to use the result without knowing which prompts you wrote or which model helped.

A useful final check is to ask the intended reader what they would do after receiving the artifact. If they cannot identify the decision or action, you may have produced a polished document without helping anyone. Revise the structure around their needs.