Oct 20, 2026
3 min read
I Asked AI to Save Me Time. It Gave Me Homework.
AI planned my dinner with six friends and handed me a second job. What that taught me about designing the work AI takes off someone’s plate, not just the work it helps me make.

Introduction
I asked AI to help organise dinner with six friends. It gave me a lovely plan.
“Collect everyone’s availability. Gather dietary preferences. Compare restaurants. Create a poll. Follow up with anyone who hasn’t replied.”
Wonderful. I now have a second job, and it doesn’t pay.
One friend says “anything works”, then rejects everything. Another answers “yes” to a question that wasn’t yes or no. Someone suggests a restaurant in a different postcode, possibly a different civilisation.
The AI explained my suffering in beautifully formatted bullet points. It didn’t take any of it off my plate.
1. “How do we use AI?” is two questions
How can AI help me do my work?
How could it change the work someone else has to do?
I can get very good at the first while barely touching the second. The second is the one users pay for.
2. Design the outcome, not the chores
Ask AI to design dinner planning and you get a dashboard: availability tab, preferences page, comparison table, notification centre. And, naturally, a chatbot in the corner.
We have given six people enterprise software to eat noodles.
So I write down what I actually want, in one sentence: “Find somewhere all six of us can eat on Friday, within budget, without me chasing anyone.”
The dashboard was a list of chores. That sentence is a job someone could take off my hands.
3. Design what happens after “Leave it with me!”
The ideal reply is one question, not a project:
“Two places fit everyone. One is quieter, one is closer. Both have a table at 7:30. Which would you like?”
Behind that sentence sits the real design work. Whose reply is still missing? What did it base this on? What happens when someone changes their mind? And if there’s a non-refundable deposit, I’d like a word before my bank account joins the group chat.
“Leave it with me!” is a big promise. Someone has to design everything after it. Usually, that someone is us.
4. Use AI to argue, not to decorate
I start with a scrappy idea of my own, then give AI something specific to push against:
“Where am I still making the person coordinate everything?”
“One friend never replies, another changes their budget, the restaurant’s availability is wrong. Walk me through it.”
“Which parts need AI, and which just need a simple rule?”
That beats “Make this innovative”, which is how you end up with a floating purple orb asking about your dining journey.
And sometimes a reminder is just a reminder. It doesn’t need a backstory and a sparkle animation.
5. Edit like you mean it
A convincing walkthrough can quietly assume data nobody has, so I still check it against real people and real technical limits. Then I edit the output, aggressively.
Keeping a decision only because AI suggested it is like bringing a stranger to a design review and saying, “He seemed confident.”
Try it this week
I judge my AI use by what it helps me notice, question and decide. Some days that’s a quick prototype. Some days it’s deleting one. Same value, different screenshot potential.
Pick one small, irritating task. Write down what “done” means. Then count how much coordination you’re still carrying.
Mine would be dinner. Although getting six friends to agree on a restaurant may be an unreasonable first assignment, for any intelligence, artificial or otherwise.
Key Takeaway
Don’t judge AI by how much it produces. Judge it by how much coordination it takes off someone’s plate. Write down what “done” means, design what happens after “Leave it with me!”, and ask which parts need AI at all.
Be the first to know about every new letter.
Stay ahead in the world of web design with expert insights, industry trends, and practical tips.
No spam, unsubscribe anytime.


