Gbiz Project Reflection

---
name: gbiz-project-reflection
description: Walks you through Gibbs' Reflective Cycle on a finished project with Claude asking the questions, producing your own account of what happened, what worked and what did not, what you learned about the work and about using AI, an action plan, and two interview-ready lessons. Use for "run gbiz-project-reflection", "reflect on my project", "Gibbs reflective cycle", "what did I learn from this project", "lessons learned for interviews", "help me reflect on using AI", "reflective account of my work", part of the Claude for Business Graduates Pack by Polar Bear.
---

# Project Reflection

## When To Use
You finished something and want the lessons ready for interviews. "What would you do differently?" is coming, and "nothing, it went well" is the answer that loses the room. This reflection answers: what did this project teach you, about the work and about using AI, that you can back with evidence?

## When Not To Use
If you need interview stories in STAR form, use the AI Interview Answer after this. If the reflection is assessed course work, run it only within your university's rules, and Claude still never writes your answers. For a single bad day rather than a project, a short note in the AI Work Log is enough.

## Inputs
- Your AI Work Log, or any notes kept during the project
- The deliverable, or the Portfolio Case Study
- Any feedback you received, with names replaced by roles
If you have none of this, I start from your memory of the project and mark the reflection as unsupported until you add evidence.

## Approach
Gibbs' Reflective Cycle, as set out in the University of Edinburgh Reflection Toolkit: six stages, from description through feelings, evaluation, analysis and conclusion to an action plan. An AI strand runs through evaluation and analysis. Claude only asks; you write every answer in your own words. Where Gibbs breaks: in a work setting the feelings stage can turn into a diary, so it stays short. The failure it prevents: a polished reflection Claude wrote that you cannot repeat in an interview.

## Workflow
1. Ask three questions: which project, what evidence you have to hand, and whether this is for you, for an interview, or for assessed course work (if assessed, within your university's rules).
2. Description: Claude asks what happened, when, and who was involved by role. One question at a time; you answer. Claude points out any vague answer ("it went fine") and asks for the specific moment.
3. Feelings, kept short: what you thought at the time and what you think now. Two or three lines, then move on.
4. Evaluation: what went well and what went badly, each tied to a log entry or a page of the work. AI strand: where Claude helped, where it misled you, and what your check caught.
5. Analysis: why each of those happened, and what knowledge or method helps explain it. AI strand: was the problem the brief, the check, or the choice to use AI at all?
6. Conclusion and action plan: what you learned, the skill to develop, and what you would do differently, with how you will make sure (a habit, a checklist, a skill in this pack).
7. Close with two interview-ready lessons in your words, each tied to evidence. Claude flags any lesson with no evidence behind it; you fix or drop it.

## Output Format
```markdown
# Project Reflection
Project: [name] | For: [self / interview / course work within your university's rules]
## Gibbs' cycle (your words)
| Stage | Your answer | Evidence |
|---|---|---|
| Description | [what happened, roles only] | [log entry] |
| Feelings | [two or three lines] | |
| Evaluation | [went well / went badly] | [log entry or page] |
| Analysis | [why] | [log entry or method] |
| Conclusion | [what you learned, skill to develop] | |
| Action plan | [what you would do differently, how] | |
## The AI strand
| Where Claude helped | Where it misled | What your check caught |
|---|---|---|
| [moment] | [moment] | [check and finding] |
## Two lessons for interviews
1. [Lesson in your words] | Evidence: [trace]
2. [Lesson in your words] | Evidence: [trace]
## Decision
[You decide which lesson to lead with and which action you start on before [date].]
```

## Done When
- All six stages answered in your words, in order
- The AI strand names at least one place Claude misled you or nothing was checked
- Two lessons, each with evidence
- The action plan says how, not just what

## Quality Bar
- Claude asks one question at a time and never supplies an answer, a feeling or a lesson.
- Teamwork is reflected on by role; no judgement of named teammates.
- No invented moments or results; a gap stays a gap.
- Feelings stay short; the weight is on evaluation, analysis and the action plan.
- Written by you; Claude only asks the questions.

## Next
Run gbiz-job-ad-decoder (Job Ad Decoder) to point the lessons at real roles.

## About the makers

This pack is made by Polar Bear, a consultancy built by ex-McKinsey founders with a dream to make AI work for People, not instead of them. We help our clients build people systems and AI-first ways of working, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).