Re: AI and meetings. The value propn is in having the AI *attend* the meeting for you, make your points and build your alliances and understandings and get your decisions, and after the meeting, letting you know outcomes if they are surprising. That's what an agent does.
All of those micro-tasks in your diag and Claude's are pretty much automatable now. None of them really make sense in a fully AI-supported world.
(With picking fantasy football teams, using AI is defeating the purpose. It's entertainment, a test of YOUR skill and insight against others'. Using AI makes as much sense as using AI/robots to actually play the games and win Olympic medals.)
AI agents, if they live up to even a sixteenth of their billing, are a bigger change to How We Do Things than the transition from steam engines to electric motors for organising factories, and from gaslight to electric light in homes and on streets, and from messenger boys to telephones, and from horse and cart to diesel trucks and tractors. All put together. Possibly even bigger than the change from the miasma theory of disease to the germ theory, and the change from manuring to artificial fertiliser for crops.
Calendar conflicts and doctor appointments and payment reminders and tax returns and food ordering are going to be quaint historical practices that exist only in textbooks, like morris dancing, faxes, and running out of gas on the freeway.
Doordash is far too expensive for me to tell my phone to order me a burger on a whim. The AI agent can't save me time there.
I have three calendar items for the rest of the year, so nothing needs to be sorted. The AI agent can't save me time there either.
Corporate mandates we attend online meetings and have our cameras on so middle management can boost their bs metrics. The AI agent can't help me because the entire point of corporate meetings is to waste employees' time. If time saving was an option, it would just be an e-mail instead.
It can however, save me 20 minutes one day a year by doing my taxes. However, if it screws that up, the ramifications aren't worth it.
It could, realistically, apply to jobs for me. However, since I'm not unemployed, it also offers me no value.
Wake me up when AI can mow my lawn, clean my bathroom, do my laundry, wash my car, or do the dishes. I want a robot that does domestic labor, not a glorified Google Assistant.
Phenomenal read. I think you’re dead on about agents for personal use - I’d be very interested to know your view on agents for business use, which I suspect is going to be critical in the playbook of “first steps for agents”. As you mentioned you can pay an intern $20 an hour to screw up that task, or you can pay an agent $20 a week to screw up the task but get to put “powered by agents” in your marketing material, and OpenAi (or meta, or anthropic, or…) get practice and training data; the agents slowly get better, people slowly get used to seeing them work and understanding their capabilities, but just as importantly their limitations and how to work around or supervise to still get value - google maps doesn’t work perfectly all the time, but it’s close enough that it’s a tremendous value add when it does work, and we know where it fails (new roads, no service, traffic) and we have to fall back to old skills like reading the street signs and figuring out which way north is or just driving around till we figure get back into an area it works at its tremendous value.
All of this does not lend itself to a hard takeoff or “the year of the agent”. But it does lend itself to adoption.
Re: AI and meetings. The value propn is in having the AI *attend* the meeting for you, make your points and build your alliances and understandings and get your decisions, and after the meeting, letting you know outcomes if they are surprising. That's what an agent does.
All of those micro-tasks in your diag and Claude's are pretty much automatable now. None of them really make sense in a fully AI-supported world.
(With picking fantasy football teams, using AI is defeating the purpose. It's entertainment, a test of YOUR skill and insight against others'. Using AI makes as much sense as using AI/robots to actually play the games and win Olympic medals.)
AI agents, if they live up to even a sixteenth of their billing, are a bigger change to How We Do Things than the transition from steam engines to electric motors for organising factories, and from gaslight to electric light in homes and on streets, and from messenger boys to telephones, and from horse and cart to diesel trucks and tractors. All put together. Possibly even bigger than the change from the miasma theory of disease to the germ theory, and the change from manuring to artificial fertiliser for crops.
Calendar conflicts and doctor appointments and payment reminders and tax returns and food ordering are going to be quaint historical practices that exist only in textbooks, like morris dancing, faxes, and running out of gas on the freeway.
Doordash is far too expensive for me to tell my phone to order me a burger on a whim. The AI agent can't save me time there.
I have three calendar items for the rest of the year, so nothing needs to be sorted. The AI agent can't save me time there either.
Corporate mandates we attend online meetings and have our cameras on so middle management can boost their bs metrics. The AI agent can't help me because the entire point of corporate meetings is to waste employees' time. If time saving was an option, it would just be an e-mail instead.
It can however, save me 20 minutes one day a year by doing my taxes. However, if it screws that up, the ramifications aren't worth it.
It could, realistically, apply to jobs for me. However, since I'm not unemployed, it also offers me no value.
Wake me up when AI can mow my lawn, clean my bathroom, do my laundry, wash my car, or do the dishes. I want a robot that does domestic labor, not a glorified Google Assistant.
Phenomenal read. I think you’re dead on about agents for personal use - I’d be very interested to know your view on agents for business use, which I suspect is going to be critical in the playbook of “first steps for agents”. As you mentioned you can pay an intern $20 an hour to screw up that task, or you can pay an agent $20 a week to screw up the task but get to put “powered by agents” in your marketing material, and OpenAi (or meta, or anthropic, or…) get practice and training data; the agents slowly get better, people slowly get used to seeing them work and understanding their capabilities, but just as importantly their limitations and how to work around or supervise to still get value - google maps doesn’t work perfectly all the time, but it’s close enough that it’s a tremendous value add when it does work, and we know where it fails (new roads, no service, traffic) and we have to fall back to old skills like reading the street signs and figuring out which way north is or just driving around till we figure get back into an area it works at its tremendous value.
All of this does not lend itself to a hard takeoff or “the year of the agent”. But it does lend itself to adoption.