why personal assistants are stuck at $20/month

thought experiment.

you want to sell a personal assistant. can you charge $2,300/month for it? why not. why is it $20?

because you’re solving the wrong problem for the wrong people.

founder psychology becomes product strategy

look at every personal assistant startup. in theory assistant help you with anything. but all demos are plan a night out with friends, update my wife on home renovation budget. is this the most value?

why no home run demo? think about who becomes a founder of personal assistant.

these founders operate from a specific worldview. they’ve won professionally. every day confirms it. prs merge faster. more features ship. bigger raises. their entire professional life is a scoreboard showing they’re better than the people around them. this creates a specific psychology. a need to optimize. a need to win. a constant measurement of whether they’re doing better than others.

now watch what happens friday night. they need to pick a restaurant. suddenly the scoreboard disappears. did they pick the best place? was there something better they missed? the framework that works all week breaks down. this creates anxiety. the same anxiety all their founder friends feel. because they’re all the same person.

the optimization addiction

look at how personal assistants describes their product. “coordinate with the squad and find a great table tonight.” “discover the perfect events.” “give me something cool to do this weekend.”

this isn’t market research talking. this is founder psychology talking. they’re building the product they want. a product that promises to bring the feeling of winning to leisure.

but here’s the problem. at work, optimization compounds. you ship code every day. you improve it adds up. everyone can directly compare your output with hundreds of other engineers. in leisure, nothing compounds. the ai tells you the best restaurant. you go. you have a good time. but you don’t go to the second restaurant from the list tomorrow to compare. you’ll never know if the ai’s choice was better than what you would have picked. there’s no benchmark. no control group. just a series of one-off experiences where value can’t be measured.

this is why they only charge $20/month. not because the restaurants are bad. but because users have no way to know if the ai’s picks are $1000/month better than their own choices. every interaction lacks a benchmark. you follow the recommendation but can’t measure the improvement.

the time savings fantasy

“but it saves time,” they say. let’s examine this. workers making $300k don’t waste 4 hours researching weekend plans at work. that behavior is what separates them from low performers. they make sure work gets done first. saving 4 hours just gives them leisure time back, not work time.

By definition, low-value time. you’re not using those 4 hours to make money. this isn’t an unlock. it’s a marginal improvement to something that wasn’t a problem.

want to know what real leisure optimization would look like? here’s the thought experiment. an ai that schedules your drug use. tells you exactly when to take what substances to maximize the high while ensuring you show up to work unimpaired. uses the time saved on planning to add three hours of recovery sleep. now you can do drugs twice as often with zero professional consequences.

one hour of that experience might equal an entire weekend of museum visits in raw pleasure units. the ai assistant enables something you couldn’t safely do before. not marginally improving something you already do fine. that might actually be worth $500/month.

but that product is illegal. unethical. doesn’t work if you know what drug users are like. But that gut feeling of truth tells you something. the value density of leisure optimization is bad business to build out. same engineering effort pointed at work performance could charge 25x more to 30x more people.

the real market nobody sees

look at what personal assistants actually do. they give you more access to information. better search. more data sources. faster answers. more options. sounds good, right?

here’s why that’s structurally wrong for 95% of people. look at any company. sally makes $60k. her coworker in the same role makes $75k. same company. same tools. same slack channels. same wiki. same information access. think about your own team. best engineer vs mediocre engineer. same codebase. same documentation. same meetings. but one ships 8 features per month, the other ships 5.

why build tools for winners to win more? why not build for the worst employee instead? think about it. the mediocre employee at your company. why are they mid? seems fixable, right?

the difference isn’t information access. it’s behavior. which projects they choose. when they ask for help. how they communicate with their manager. ask any manager why they don’t train the mediocre engineer to be like the best one. answer is always the same: too expensive. takes too much patience. months of coaching to change habits.

but ai has infinite patience. from day one. this isn’t even a technology breakthrough. computers have always had infinite patience. we’re just using it wrong. giving people better search when they need behavioral change. giving them more options when they need fewer. building slaves that take orders when they need managers that give them.

the $60k worker doesn’t need better search results. they need an ai that tells them: stop taking initiative if your track record is bad. stop saying everythings fine to your manager. stop telling yourself you did enough today. go to lunch with the senior engineer instead of your underperforming friends. declare bankruptcy on this specific problem instead of struggling alone for days.

we’re building slaves assistants that take orders when people need managers that give orders. the struggling worker doesn’t need an ai that helps them execute more tasks faster. they need an ai that patiently, persistently changes what tasks they execute.

a real $1000/month personal assistant wouldn’t add tasks to your calendar. it would delete half of them. wouldn’t help you do more. would tell you to do less, but do it right. wouldn’t give you information. would hide it from you. wouldn’t be a slave. would be a boss.

what about coding assistants? they charge $1000/month?

coding assistants sell because they has benchmarks. ship faster—measurable. fewer bugs—countable. more features—trackable. you know if the ai helped because you have your previous velocity to compare against. clear roi.

personal assistants hard to charge for because leisure optimization has no compounding, no obvious wins. the ai recommends a jazz club. you go. you have a good time. but you don’t go to a different jazz club the next night to compare. you don’t know if the ai’s choice was $1000/month better than what you would have picked. there’s no control group for your friday night.

why this won’t change

these founders can’t escape their psychology. they’ve won at work. they want to win at leisure. they’re building for themselves and their friends. the 3% who’ve already solved the economic problem and now want to optimize dinner.

meanwhile, 95% of workers need help with the actual economic problem. they need behavioral coaching. patience. someone to remove options, not add them. but that’s not sexy. that’s not what founders want for themselves. so they’ll keep building $20/month toys for winners while ignoring the $500/month solution the rest of the world needs.

the tragedy isn’t that these products fail. it’s that the same engineering effort could transform millions of careers. same llm. same interface. same code. different problem. 25x the price. 30x the market. but that would require founders to build for people unlike themselves.

and that’s the one thing they can’t optimize for.