I actually do a lot of work via agents. Personally I still prefer to have actual bank transfers etc under my direct control, but I do have agents doing searches, lookups, and diverse chores for me.
If my agent(s) can't find you or are locked out of finding out what you have on offer; then I will never hear of you; and it's unlikely that we're going to be doing business.
Note that a lot of people use chatgpt or gemini instead of google search or ddg to interact with the web these days; simply because the search engines have stopped being as useful. So agents might be the only way you'll be getting found in a couple of years.
> simply because the search engines have stopped being as useful
While there's definitely some truth to this for Google (I haven't noticed with DDG), it clearly isn't the only reason. The moment Google shoehorned Gemini into the top of search results, that was enough for a lot of people to say "good enough". Because laziness and susceptibility to convenience are very real things.
I've been collecting datasets of different sources while I can so that they can be instantly and always available to my agents. Spans subreddits, vehicle forums, 3d printing, etc... Basically anytime a source can be blocked or rate limited, I look at acquiring it.
So much of AI goldrush feels like crabs in a bucket. "I'll do whatever I can to get a leg up on my neighbour today." But ultimately it's pointless - the advantage will be short lived and everyone will end up in the same spot. Only bigcorp has the money to break the laws to spin the flywheel to consume the content to train the models to rent services to you.
I don't think its gonna happen any time soon. Its too risky.
AI agents although autonomous aren't deterministic unlike compilers.
Now humans are also not deterministic and are autonomous in the human context but they have a lot to lose and can own outcome and can be held responsible unlike AI agents or their parent companies.
Unless a company is willing to guarantee outcome and bear with the losses, no one will trust AI agents for shopping and banking apps and full root access to personal mobile phones.
I think those anti-AI companies are the ones with the problem.
I have and will continue to switch from services that try and lock my agents out to ones that don't (or write my own). More importantly, this is a cat-and-mouse game they will not win.
The same way I only bank with places that YNAB (so Plaid) can talk to. If a financial institution wants to block Plaid then that's their prerogative, and I will bank/invest/whatever elsewhere.
Businesses would be smart to embrace APIs/MCPs or they are destined to fight, unsuccessfully, against agents using their web UIs which is their loss as that almost certainly is heavier on their backend systems.
Gmail MCP doesn't support downloading attachments, inserting links that it doesn't mangle, or changing the "from" field? Fine, then my agent will drive the browser.
Stripe MCP is dogshit that you have to re-auth every 5 seconds? Fine, I'll send an agent to use the browser.
ClickUp wants to put silly limits on their MCP? Fine, I'll write my own that uses their API. Still can't do some things you can do in the browser? Fine, I'll use the browser.
Agentic use is here to stay, these companies can get on the bandwagon and let people access their _own_ data or they can expect a decline.
AWS has the right idea. Using agents with it is beautiful. I just cut 45% of our monthly bill using an LLM to help dig in, track down, and make the changes. Cleaned up a ton of cruft that was never going to get done by hand. Tagged all resources for better billing reports, and did it all with 0 downtime or issues. There is a 0% chance I would have been able to make all those changes with no mistakes doing it by hand.
I actually do a lot of work via agents. Personally I still prefer to have actual bank transfers etc under my direct control, but I do have agents doing searches, lookups, and diverse chores for me.
If my agent(s) can't find you or are locked out of finding out what you have on offer; then I will never hear of you; and it's unlikely that we're going to be doing business.
Note that a lot of people use chatgpt or gemini instead of google search or ddg to interact with the web these days; simply because the search engines have stopped being as useful. So agents might be the only way you'll be getting found in a couple of years.
> simply because the search engines have stopped being as useful
While there's definitely some truth to this for Google (I haven't noticed with DDG), it clearly isn't the only reason. The moment Google shoehorned Gemini into the top of search results, that was enough for a lot of people to say "good enough". Because laziness and susceptibility to convenience are very real things.
I've been collecting datasets of different sources while I can so that they can be instantly and always available to my agents. Spans subreddits, vehicle forums, 3d printing, etc... Basically anytime a source can be blocked or rate limited, I look at acquiring it.
So much of AI goldrush feels like crabs in a bucket. "I'll do whatever I can to get a leg up on my neighbour today." But ultimately it's pointless - the advantage will be short lived and everyone will end up in the same spot. Only bigcorp has the money to break the laws to spin the flywheel to consume the content to train the models to rent services to you.
I don't think its gonna happen any time soon. Its too risky.
AI agents although autonomous aren't deterministic unlike compilers.
Now humans are also not deterministic and are autonomous in the human context but they have a lot to lose and can own outcome and can be held responsible unlike AI agents or their parent companies.
Unless a company is willing to guarantee outcome and bear with the losses, no one will trust AI agents for shopping and banking apps and full root access to personal mobile phones.
[flagged]
I think those anti-AI companies are the ones with the problem.
I have and will continue to switch from services that try and lock my agents out to ones that don't (or write my own). More importantly, this is a cat-and-mouse game they will not win.
The same way I only bank with places that YNAB (so Plaid) can talk to. If a financial institution wants to block Plaid then that's their prerogative, and I will bank/invest/whatever elsewhere.
Businesses would be smart to embrace APIs/MCPs or they are destined to fight, unsuccessfully, against agents using their web UIs which is their loss as that almost certainly is heavier on their backend systems.
Gmail MCP doesn't support downloading attachments, inserting links that it doesn't mangle, or changing the "from" field? Fine, then my agent will drive the browser.
Stripe MCP is dogshit that you have to re-auth every 5 seconds? Fine, I'll send an agent to use the browser.
ClickUp wants to put silly limits on their MCP? Fine, I'll write my own that uses their API. Still can't do some things you can do in the browser? Fine, I'll use the browser.
Agentic use is here to stay, these companies can get on the bandwagon and let people access their _own_ data or they can expect a decline.
AWS has the right idea. Using agents with it is beautiful. I just cut 45% of our monthly bill using an LLM to help dig in, track down, and make the changes. Cleaned up a ton of cruft that was never going to get done by hand. Tagged all resources for better billing reports, and did it all with 0 downtime or issues. There is a 0% chance I would have been able to make all those changes with no mistakes doing it by hand.
[flagged]