Chore vs Craft: When an AI Agent Replaces a Tool, and When You Still Need a Design Engineer
AI agents absorb chores well — drafting, filing, monitoring. They don't replace the craft decisions that make a product shippable. Here's a simple test to tell the two apart.
Shaheer Malik
Framer Designer & Developer
Quick answer
AI agents reliably absorb chores where 'correct' already exists somewhere — renaming components consistently, drafting changelogs, monitoring competitor pages, reproducing bugs. They don't replace craft work where someone has to decide what correct means right now: visual hierarchy, real edge cases, and the last-mile judgment between a working prototype and something a customer can rely on. Test any task by asking which category it falls into before delegating it.
AI agents are genuinely good at absorbing reversible chores — drafting, monitoring, filing, renaming, research. They are not a substitute for the craft work that makes a product shippable: visual systems, edge cases, and the judgment calls in the last mile between "it works in the demo" and "a customer can rely on it." Here's the test for telling the two apart.
Chores an agent can genuinely take
- Renaming a component consistently across a codebase
- Drafting a changelog entry from a merged PR
- Producing QA notes on a build
- Monitoring a competitor's pricing page for changes
- Reproducing a reported bug and documenting the steps
These share a trait: the definition of "correct" already exists somewhere (a spec, a diff, a prior example), and the task is applying it consistently rather than deciding what it should be.
Craft that still needs a design engineer
- Deciding the visual hierarchy of a new screen from scratch
- Handling the actual edge cases — what a form looks like with 200 characters of overflow text, what a dashboard looks like with zero data
- The last mile from a working prototype to something that survives real usage: accessibility, responsive behavior across real devices, the states nobody thinks to demo
- Judging when a design decision that "looks fine" is actually inconsistent with the rest of the product
These share the opposite trait: there's no existing spec to apply, because the decision itself is the work.
The chore-vs-craft test
Ask one question about the task in front of you: does "correct" already exist somewhere, or does someone have to decide what correct means right now? If it already exists — a spec, a design token, a prior pattern — it's very likely a chore an agent can absorb, following the same reversibility discipline covered in our one-reversible-job piece. If someone has to decide, it's craft, and today that decision still needs a human who can be held accountable for it.
A Framer-specific tell
If a page assembly follows existing CMS structure and existing components, that's chore territory — an agent applying a pattern. If a page requires deciding a new information hierarchy, a new interaction pattern, or how a brand-new section should feel next to the rest of the site, that's craft — and per our publish-authority piece, it still needs a human to both design and approve it before it ships.
Frequently asked questions
Can an AI agent replace a design engineer?
It can absorb the chores around the work — drafting, monitoring, filing — but not the craft decisions: visual hierarchy, edge cases, and the judgment calls in the last mile to a shippable product.
How do I know if a task is a "chore" an agent can handle?
Ask whether "correct" already exists somewhere — a spec, a token, a prior pattern. If yes, it's a chore. If someone has to decide what correct means right now, it's craft.
What's an example of "craft" work in a Framer site?
Deciding the information hierarchy or interaction pattern for a brand-new section — as opposed to assembling a page from existing CMS structure and components, which is closer to a chore.
FAQ
Frequently asked questions
It can absorb chores around the work, but not craft decisions: visual hierarchy, edge cases, and last-mile judgment.
Need this kind of work for your product?
I design and build websites, products, and brands for SaaS & AI startups — design and code under one roof.