we are in an insane sprint to the future. everyone wants everything to be automated. there are 100s of companies being built everyday aiming to "replace human labor doing rote manual tasks".
i've built agents for enterprises and consumers alike for the past 1.5 years. dealt with both unsatisfied customers and people who love our product.
there are a few fundamental flaws in how enterprises and consumers think about agents:
1. we expect it to work for us on day 1 and discredit it if it doesn't: there is nothing fundamentally wrong with this expectation but this expectation assumes ai works the same way traditional SAAS does, undermining the very fact that makes this technology great: it learns with time.
2. we think we know what we want: if there is one thing that came out of ai automating software creation is that, humans are terrible at describing what they really want. we have some idea but that's it. if we ourselves don't really understand the problem and possible solution, how do we expect then that a junior super smart intern(that ai is) can solve it for us.
3. we are very terrible at predicting the future of an unknown technology: when voice ai was at a very nascent stage, people complained that it doesn't feel realistic and would never be able to replace actual humans. happy to say, we were very wrong.
4. we assume that humans want to automate the rote parts of their jobs: fear. fear of being replaced is a much deeper fear than we realize. when the c-suite in an org decides that they want to use "AI" to improve their workflows, they do so in a very top-down fashion. imposing their decision onto their juniors who probably don't want the solution at all even if it helps them because they are scared.
but its not all our fault as consumers and enterprises. we have been lied to by flashy launches and demos.
then how can builders actually create great agents that help consumers and enterprises?
1. meet them where they are: why did openclaw become so great besides the fact that it was an awesome assistant on your computer? it made agents accessible through telegram and whatsapp, the place we already are. build onto that. slack. teams. emails. the technology itself should feel invisible. h/t to poke by the interaction company.
2. overcommit and overdeliver: overcommit to the whole project scope and overdeliver on the pilot scope. why? the more intertwined your product is with the complete context, the better it gets. genuinely care about the user's experience and constantly aim to change your invisible technology as per the demands.
3. spend time with people who are doing those tasks manually: literally sit beside them and watch them do it. what do they find annoying? why do they do things the way they do it? can you abstract this problem completely and let them handle the thinking?
if you think the agents problem has been solved completely, just look at what openclaw did in 3 days. openclaw for enterprises doesn't exist yet. someday it will.
reach out if you are building agents and need help in building the architecture or doing discovery with enterprises. would love to chat :)