# Build a small internal tool with AI

Original 5HO practice material. Fictional example. Reviewed 5 October 2026.

## Useful result

You'll finish with a tested local equipment checkout prototype.

A useful internal tool does one job clearly. Start with a small screen and fictional data. If the brief already includes accounts, payments and “a quick integration”, it is probably more than a first build.

## Set up

Allow 45–60 minutes for a local prototype. Use an AI coding tool and browser. The estimate excludes production hosting, authentication and connections to real business systems.

## Practice input

Prototype: equipment checkout list. Fields: item, borrower, return date, status. Three fictional items: projector Available; remote On loan to Jo due 9 October; speaker On loan to Lee, return date unknown.

## Do the task

1. Write the one user task and four fields before asking for code.
2. Build with embedded fictional data and obvious Available/On loan filters.
3. Test blank fields, date display and whether edits survive a reload; document the storage behaviour.

## Prompt

Build a local single-page equipment tracker using the sample. Include labelled inputs and filters, and explain where changes are stored. No login, network calls or external sharing. Keep unknown return dates visible.

## Refine

Add a reset control with confirmation. Make clear that browser-local storage is not a shared team database or a backup.

## Expected checks

The two loaned items should appear under On loan. The speaker must not acquire an invented return date. Stop before using real records if storage, access or recovery is unexplained.

## Save

Save prototype-v1.html and a short limitations note. Review the missing production requirements before inviting colleagues to rely on it.

## Sources

- Writing clear prompts: https://help.openai.com/en/articles/10032626-prompt-engineering-best-practices-for-chatgpt