AI Guides › Creative & Design
Using AI To Finish The Boring 20% Of A Creative Project, Not The Interesting 80%
By Nigel Guy · 2 min read
Most creative projects have a shape: the interesting decisions get made
early and with real attention, and then there's a long tail of necessary
but mechanical work — resizing for six platforms, writing alt text,
formatting captions, cleaning up file names, producing the fourth colour
variant nobody will remember choosing. It's exactly the kind of work that's
easiest to justify handing over entirely, and exactly the kind of work
where handing it over causes the least damage if you do.
The rule: the right place for AI to take over completely is the
mechanical tail of a project, not the front of it — the test is whether the
task has a creative decision left in it at all, not whether it's tedious.
Sorting the 20% from the 80%
- List the remaining tasks on the project honestly. Most lists split
cleanly once written down: a handful of tasks that still involve a real
choice, and a longer tail that's execution of choices already made.
- For each task, ask if a decision is still being made or just applied.
Choosing the hero image is a decision. Producing the same image at nine
required aspect ratios is applying one. That's the actual dividing
line, not how tedious the task feels.
- Hand the "applying" tasks over fully, and expect to hand them over
fully — spot-check the output rather than treating this the way you'd
treat the front-of-project decisions.
- Keep the "deciding" tasks, even the tedious-feeling ones. Writing
thirty product descriptions where each one still requires knowing that
specific product's actual differentiator is a decision task wearing a
boring costume — it belongs on the other side of the line.
What to skip
Skip the reasoning "this is tedious, so it must be safe to fully
automate" — tedium and lack of a real decision aren't the same thing, and
some of the most decision-heavy work in a project is also the most
repetitive-feeling. And skip re-reviewing every mechanical output at the
same intensity you'd give a creative one — that defeats the point of
sorting them in the first place.
Guardrails
- Spot-checking, not skipping review entirely — mechanical output still
fails sometimes (a wrong crop, a broken file, a formatting error), and a
light check catches most of it cheaply.
- This sorting is per-project, not universal — a task that's pure
execution on one project might carry a real decision on another,
depending on what's already been settled going in.
- If you're unsure which side of the line a task falls on, treat it as a
decision task until you're sure — the cost of over-reviewing a
mechanical task is small; the cost of under-reviewing a decision task
usually isn't.
All 751 AI guides · JulieMango plans from £17/mo