Ask HN: What are the BIGGEST Problems you'd face with a full-AI employee? | Hacker News Reader

Ask HN: What are the BIGGEST Problems you'd face with a full-AI employee?

10 pointsdontoni5 comments

Imagine an AI agent that’s indistinguishable from a remote worker, something like Devin but for anything and whose medium of contact are email, Slack, phone calls, etc.

What would your main pain points and concerns be? I’ve come up with the following list but can’t think of any more problems:

Dream outcome (Donsir Mechas) - Create anything big & of high quality and best practices with a prompt. - Be like God: you ask for something, immediately get the full & perfect response. - 1,000 employees in your pocket. - People will look at you and say, how does he go so fast? How does he get that much done? - Like Devin, but for everything

Problems - Create an account - I will have to copy-paste stuff into my project (code) - I will have to run tests myself - I’ll have to add refinements myself - I’ll have to find bugs and fix them myself - I can’t trust what ai makes, it’s just slop - I will have to use something like Cursor to review and accept/reject the code - I will have to read and internalize what the AI did and how it links with the rest of the code - I will have to give the same large instructions every time in every prompt - I will have to learn about the platform - I’ll have to manage secrets and api keys, and know which of them to give the agent

This question is so artificial that it’s hard to answer. If we assume the employee really is perfect and indistinguishable from a real person then the issues are all related to the company running it.

Instead of hiring an autonomous individual, you’re contracting out to a specific company. You incur risks related to that company. If the company is breached, all companies using their employees are breached. If law enforcement becomes involved, the rules might be different.

Having your “employees” be bots owned by another company also leaves you vulnerable to being squeezed. What if they get a new CEO who raises prices 5X? Or decides they no longer want to serve your industry? Or some laws or geopolitical dispute means they get cut off one day?

So all or at least main risks of such a service is security and compliance, rather than quality of output?

Couldn’t you make the same argument for Midjourney, for example, which heavily depends on Discord?

Single point of failure, single vendor.

All the CEOs are dreaming of getting rid of workers, but I wonder how many of them think clearly about what they are getting. Depending on a single, or few vendors for full-AI employees is the concrete equivalent of unionizing their workforce. If the vendor decides on a price hike (think about the recent history of VMWare), or their full-AI employee gets dumber or lazier (Claude Code taught us a good lesson here just lately), then your whole AI-powered company is at risk. Our current human-based organizations are much more resilient to single point of failures.

Hm. What if it’s a startup that has just a couple of people and really needs more operational capability even though can’t hire that many people?

From what you said, you won’t accept such a product if it isn’t open source? Or at least, where you can choose the models and pay per use or something similar to that.

My guess is that big, powerful companies won’t use a pay-per-use Devin-like service because they just have so much people they can delegate tasks on that also use Cursor or the like; and startup companies that really need the product’d prefer a fixed rate like Cursor does.

Maybe they’ll actually use specialized agents for multiple tasks, but will also try to build their own solutions or sue frameworks rather than hire another company (like big companies that use Crew AI because it’s open source).

And are you ok with the quality that those full-fledged AI agents (like Devin) deliver? Do you have more pains that the single vendor dilemma?

I am generally very AI-positive. I would accept such a product, if the second best alternative is close enough that I can substitute it in case of emergency.

The question I was replying to is very specific: "Imagine an AI agent that’s indistinguishable from a remote worker, something like Devin but for anything and whose medium of contact are email, Slack, phone calls, etc. What are the BIGGEST Problems you'd face with a full-AI employee?"

My answer: I think the major biggest problem is vendor lock-in, because such a solution would be too juicy to resist.

Reply on news.ycombinator.com