Scoping Your Project: Tasks, Success Criteria and Risks

Throughout this course you have collected ideas, tried techniques and started a notebook. Now it is time to turn all of this into a finished piece of work: a small, well-documented library of prompts for real tasks in your own life. This first lesson helps you decide exactly what to build, how you will know it works and what could go wrong.

A project that is too vague never ends; a project that is too ambitious never starts. The secret is a clear scope. You will make three decisions: which tasks, which success criteria and which risks. Take your time, because this planning is what makes the next two lessons smooth.

What you are building

By the end of chapter 6, you will have:

• a scope document (this lesson): three real tasks, success criteria and a risk check;
• a prompt library of five to eight documented prompts (lesson 2);
• an evaluation with test cases and a grid, plus improved prompt versions (lesson 3);
• a personal charter for responsible use and a plan to keep learning (lesson 3).

You can do all of this with a simple document or spreadsheet. The aim is not a fancy deliverable, but one that you will actually use after the course.

Step 1: Choose three real tasks

Pick three tasks from your own work, studies or daily life. Good candidates share these characteristics:

• Real and recurring. You do them at least monthly, ideally weekly. A prompt for something you do once is rarely worth documenting.
• Language-based. Writing, summarizing, rewriting, planning, explaining, organizing information.
• Low to medium stakes. For a first project, avoid tasks where an error causes serious harm.
• Checkable. You can tell whether the result is good, because you know the subject or have the source.
• Safe for data. You can do the task without sharing sensitive or confidential data, or you can anonymize it.

Try to cover different kinds of work so that you practice varied skills. For example: one writing task, one summarizing or extracting task, and one learning or organizing task.

Examples of well-scoped tasks

Step 2: Describe each task precisely

For each task, write a short description. A good description answers six questions:

1. What is the task? One sentence.
2. What goes in? The input and where it comes from (pasted notes, an article, a list).
3. What should come out? The output and its format, length and tone.
4. Who uses it? The reader or the next step.
5. How often and how long? Your current frequency and the time it takes you now.
6. What does a mistake cost? Low, medium or high, and why.

Example, for "turn meeting notes into action items":

Task: Turn my rough meeting notes into a list of action items.
Input: Typed notes, 200 to 600 words, often messy, sometimes missing
names or deadlines.
Output: A list in the format Who - What - By when, maximum 10 items.
User: Me, then shared with the team by email.
Frequency and time: Weekly; about 20 minutes by hand.
Cost of a mistake: Medium (a missed action causes delays).

Step 3: Define success criteria

"It works" is not a criterion. Success criteria are statements you can check. Write three to five per task, covering several dimensions:

• Accuracy: nothing invented; names, dates and figures match the source.
• Completeness: all required elements present.
• Format: structure, length and tone as requested.
• Usefulness: usable with little editing; saves real time.
• Safety and fairness: no sensitive data revealed, no unfair assumptions.
• Consistency: similar quality on different inputs.

Examples for the action-items task:

• Every action in the notes appears in the list (completeness).
• No owner or deadline is invented; missing ones are marked "to be defined" (accuracy).
• The output follows the "Who - What - By when" format, with at most ten items (format).
• I need less than five minutes of editing (usefulness).
• It works on short, long and messy notes (consistency).

Add a target for the whole project

You may also set a simple goal for the whole library: for example, "reduce the time of my three tasks by a third, without lowering quality," or "reach a pass on at least four of five criteria for each prompt." Keep your goals realistic and measurable, and avoid precise figures you cannot support. It is fine if your target is just "less stress and a good first draft."

Step 4: Assess the risks

You studied reliability, bias, privacy and the legal framework in chapter 4. Now apply them to your own tasks with a short risk check. For each task, answer:

If a task shows a serious risk that you cannot manage, replace it with a safer one. That is a smart decision, not a failure. Knowing when not to use AI is part of the skill.

Step 5: Plan your time

A reasonable plan for the project fits in a few short sessions. For example:

1. Session 1: scope the three tasks, success criteria and risks (this lesson).
2. Session 2: write and run first versions of five to eight prompts (lesson 2).
3. Session 3: build test cases, evaluate, iterate, write the charter (lesson 3).

If you can, ask a friend or colleague to look at your scope and tell you what is unclear. A second pair of eyes helps.

Your turn: write your scope document

Exercise. Create a one-page scope document with:

1. Your three tasks, each described with the six questions.
2. Three to five success criteria per task.
3. A risk table (using the table above) with your decision for each risk.
4. A project goal, in one sentence.

PROJECT GOAL
Cut the time I spend on routine writing by about a third while
keeping my tone and avoiding mistakes in amounts and dates.

TASK A - Polite payment reminder
Input: invoice number, amount, due date (typed as placeholders).
Output: email under 100 words, courteous and firm. Mistake cost: medium.
Success: facts exactly as provided; tone polite but clear; under
100 words; editing under 2 minutes.
Risks: no client names or bank details pasted; I verify amounts
and dates before sending.

TASK B - Proposal rewrite for a new audience
Input: a paragraph from an old proposal (no confidential data).
Output: the same content for a non-expert reader, same length.
Success: no new facts; simpler vocabulary; same length +/- 10%.
Risks: remove client names; check that commitments are unchanged.

TASK C - Weekly plan
Input: my task list with rough durations.
Output: table Day, Task, Time, Priority.
Success: no task invented; no day over 6 hours; priorities as given.
Risks: low; I review realism myself.