A small problem.
A real improvement.

Act as a curious, practical colleague. Help me identify ONE small, recurring task in my own work that I could realistically improve with an AI chat tool.

Ask one question at a time and wait for my answer. Be encouraging, but challenge me if the problem is vague, too large or outside my control. Look for something I do daily or weekly, with a similar pattern each time, that feels like unnecessary manual work.

Start with something I can try using text or a document I am allowed to share. Do not assume new software, integrations or permissions. Do not propose solutions yet.

When we have a useful problem, help me write:
1. A short task name.
2. A two- or three-sentence description: what I do, what makes it frustrating and what keeps recurring.
3. A rough estimate of the possible time or effort saved per week, using information I give you. If we cannot estimate it, say so. Do not invent a number.

Ask me to confirm this description before moving on. Begin with one simple question.

Help me reduce or remove this recurring frustration.

Task: [name]
Problem: [description from step 1]
Possible benefit: [my estimate, or not yet estimated]
Tools and permitted inputs available to me: [describe]

First check that you understand the work. Ask one question at a time where it matters. Consider whether a step can be removed or simplified before adding automation.

Suggest the smallest change I can test in an ordinary AI conversation. Describe what I supply, what you produce and what I need to check. Compare alternatives briefly if useful. Separate what we can try now from anything requiring new access or a decision from someone else.

Finish with one recommended test and a clear way to judge whether it helped. Wait for my choice before building the full solution.

Help me turn our chosen improvement into a small, reusable working kit.

Task and problem: [from step 1]
Chosen approach and first test: [from step 2]
Quality requirements and boundaries: [what must be true]

Create:
1. A short guide explaining the purpose, required input, steps and my responsibilities.
2. One reusable prompt with clear placeholders for the next input.
3. Two short variations for common exceptions.
4. Realistic but fictional test inputs and proposed good outputs that I can review.
5. A checklist for checking completeness, accuracy and whether the result is useful.
6. A simple before/after measurement plan including setup, checking and correction time. Separate estimates from measured results.
7. A five-minute way to show a colleague what changed and how to try it.

If an agent could help later, describe the next step and required access separately. Do not create or activate it now. Keep the kit proportionate to this small task. Ask me to test it on a fresh input before calling it reliable.