Every AI planner I have opened asks me some version of the same question. What else would you like to optimise. It is a reasonable question if your problem is ambition. It is the wrong question entirely if your problem is that you are already holding more than fits.
I have been building ReturnKit for women who are the operations layer of a household, usually while also having a job, and often while their own body is doing something inconvenient. When I started adding anything AI shaped to it, I kept hitting the same wall. Almost every feature I could think of made the app cleverer without making her load lighter. Some of them made it heavier.
So I wrote down a test, and now nothing gets built here unless it passes.
Does this reduce the amount of thinking she has to do, or does it only help her think faster? And can she verify the result without redoing the work?
Both halves matter. The first one is obvious once you see it. The second one is where most good sounding ideas quietly fail.
The execution trap
The load in a household is not mostly doing. It is anticipating, coordinating, remembering, worrying and monitoring. Noticing the bread is low. Knowing the vaccine is due. Holding the fact that Thursday has no cover, all week, in the back of your head, while doing something else.
Researchers call that cognitive household labour, and it is the part that falls hardest on women, more unevenly than the physical chores do. It is also the part connected to depression, stress and burnout.
Which means an app that books the appointment has removed the smallest piece of the job. The booking was never the hard bit. The hard bit was carrying the appointment around for three weeks before it existed.
Faster is not lighter
Here is the part the industry keeps missing. The risk with AI in domestic life is not that it fails to reduce unpaid work. It is that it raises what everyone assumes you can hold.
A woman who can now process forty school messages in four minutes does not have a lighter week. She has a household that will send her sixty, and a quiet new expectation that she is on top of all of them. The ceiling moved. She did not get more room.
That is optimisation working exactly as designed and helping nobody.
The demo that would sell the most and damage the most
There is a version of this product that demos beautifully. You open the app and it says: newsletter summarised, pickup confirmed with your husband, shopping list de-duplicated, calendar conflict fixed. Two things left for you to decide.
I am not building that, and I want to be specific about why, because it is the most tempting idea in this whole category.
First, every line of it requires reading your email, your school app and your partner's messages, plus permission to change your calendar. That is exactly the surveillance ReturnKit exists to avoid, and it would trade away the one thing I actually promise.
Second, and this is the part I think gets skipped: "I handled it" is a claim you cannot check. Being wrong once, silently, costs more trust than a hundred correct summaries earn, because the cost lands on you and you had no warning it was coming. If you have ADHD you already know exactly how expensive one silently dropped thing is.
The good half of that idea is not the digest. It is the question underneath it. Not what would you like to optimise, but what can we stop thinking about today. That question is free, and it does not require reading your inbox.
Four things this app will not do
It will not give you a score. "Your load has increased 42% this week" is not a measurement, it is a number wearing a lab coat. You cannot compute that from sleep and appointment counts. It invites an argument you cannot win, and it reads as being watched. ReturnKit says things like this week has more in it than usual, from signals you logged yourself, and lets you decide what that means.
It will not predict a bad day from your cycle. Phase information can help explain a day you have already had. It will never forecast one you have not. Telling a woman on day 24 that tomorrow will be hard is a self fulfilling instruction, and it hands a stereotype the authority of software.
It will not give you permission. There is a warm version of this app that says buying dinner tonight is a reasonable decision. It sounds kind and it is still grading you, and you already have enough voices with opinions about your choices. So ReturnKit does the quieter thing instead: on a day you have said you are running on empty, it does not offer you the two hour focus block at all. There is nothing to forgive because nothing was asked.
It will not pretend to hand work to your partner. An app cannot transfer mental ownership. It can transfer a task. The load is the wondering whether he did it, and a checkbox marked "assigned" can make that worse, because now you are monitoring the app as well as the person. What actually helps is smaller: a message you did not have to compose, already in his hands.
What subtraction looks like
Every other planner asks what you want to achieve today. ReturnKit asks how much you have got.
Say you are running on empty and the app gets smaller. Three things become one. The long focus blocks come off the list, because offering two hours to someone who just told you she has nothing would be the app not listening. Optional strips fold away. Your medication and food move up the page rather than down, because a low day means more support, not less.
Two rules hold that together. Nothing with a deadline ever folds, because a missed appointment is worse than a busy screen. And one tap brings everything back, without changing what you said about your day. You said you had less. You did not ask to be locked out of your own planner.
That is the whole idea, and it is unglamorous, which is probably why the category is crowded at the other end.
ReturnKit is built by one person in Ireland, for women who hold a household in their head. It is local first: nothing leaves your device unless you turn sync on, there are no ads and no trackers. You can look around with sample data at returnkit.life/app/?demo=1.