Which repetitive development tasks are good to delegate to AI?

Question

Accepted Answer

AI assistants shine at **repetitive, well-defined, low-risk tasks** where the correct answer is easy to verify. Delegating these frees you to focus on design and tricky logic.

## Good tasks to delegate

- **Boilerplate / scaffolding** — a new React component, an Express route, a config file. "Generate a typed `useDebounce` hook."
- **Regex** — easy to get wrong by hand: "regex matching an ISO-8601 date."
- **Format conversions** — JSON to TypeScript types, a SQL table to a class.
- **Docstrings / comments** — feed it a function, ask for a JSDoc block.
- **Renames / mechanical refactors** — rename a variable consistently, extract a function.
- **Drafting tests** — "write Jest tests covering the empty, single, and error cases."
- **Summarizing logs** — paste a stack trace, ask what failed.
- **Repetitive SQL / CRUD** — generate the insert/update/delete for a known schema.

```ts
// Ask: "docstring + edge cases for this"
/** Clamp `n` into [min, max]. Returns min if n < min, max if n > max. */
function clamp(n: number, min: number, max: number): number {
  return Math.min(Math.max(n, min), max);
}
```

## What NOT to blindly delegate

- **Security-sensitive code** (auth, crypto, SQL built from user input) — review every line.
- **Business logic you don't fully understand yet** — you can't verify it.
- **Anything you'll ship without reading.** AI can hallucinate APIs that don't exist.

## Why it matters

Delegating repetitive work is the fastest, lowest-risk way to get value from AI: you save real time on tedious tasks while keeping judgment on the parts that matter. The rule of thumb is **delegate when you can verify the result quickly** — and always read what comes back before committing it.