I am so tired of every tech bro telling us how AI will change the world without giving us any concrete examples

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Heres 3 examples. Yes I used chat to draft but I put in enough detail in my prompt that it came up with real world info info.

We used to have two support analysts manually categorize 5,000 IT support tickets per week. Now a model auto-tags 92% of them with 96% accuracy, and one analyst audits exceptions.

Our paralegals draft first-pass contract summaries. It used to take them hours in many cases. Have you ever read a commercial lease?? It is a monster. Now a model generates a structured summary in 45 seconds and the paralegal reviews and edits; prep time dropped from 1+ hours to under 15 minutes.

We replaced manual fraud screening of 100% of transactions with an AI risk score. Humans now review only the top 3% flagged.

Wait... are these 3 examples of things you personally did but you asked AI to write this up? Or you asked AI to come up with 3 examples?

I asked AI to come up with several example of productivity saves, and these 3 were ones I see in my immediate world so I cut and paste them. Although I’ll admit the support ticket human reduction at my co is higher than what I posted here. We had a number of support people and now we are down to one human.

You should look up ‘agentic AI’ bc that’s the next step

I can't speak to the others, but the legal one makes no sense to me. The firm probably has a template commercial lease they prefer to use and the part that AI can do (copying and pasting in the parties and numbers) could be done by a paralegal very quickly. An attorney would still need to review the final product carefully and negotiate changes with the other side whether it was generated by a paralegal or AI, and a paralegal is less likely to make mistakes. Then it says "a model generates a summary" but that's not the initial task described, which is drafting the lease. There's not really much to "summarize" in a lease other than the initial parties and terms which you would have fed into the prompt in the first place, and AI is not good at summaries and I imagine would have a difficult time cross-referencing the various subsections of the lease (as in "notwithstanding paragraph 3, and subject to the exceptions in section 5b, below, the parties shall...). Using AI seems like a huge waste of time and may create more work.

If a lawyer is getting paid, they need to review the contract.

I did not say otherwise.

[Anonymous]

My husband and I own a software development firm. We are no longer hiring pure developers; AI is better at it, given you know how to structure your prompts.

To give a very concrete example - I have a niece outside the US who runs a small medical practice. She was shopping around for software to help manage her practice, and running into a Goldilocks problem - most of the good ones were written for hospitals, and overkill/expensive, and most of the small ones were not feature rich, could not be customized for how she does her business, and support was bad/non-existent.

We stood up a hospital management system in a month - entirely customizable, feature rich, fairly easy to modify with requirements changes. And by we, I mean two people - DH who did the coding, I did the testing. A similar system would have taken a whole lot more man hours (years not months). And we can sell it to others in her situation, if we wanted to.

DH developed a study helper for DD in his spare time. It took two days, I believe. You can feed it any material, and it will give you test questions based on the material (you can customize the difficulty level), give you answers, and you can chat with it to get explanations.

[Anonymous]

Either genAI or agentic AI can now do everything our associate level hires can:
1) analyzing spreadsheets/pulling trends- I can describe 10 pivot tables/analyses to make in Google Sheets and it will do it instantly
2) first drafts of things if given a clear argument/prompt; I still have to re write a lot of it but it’s equal to a new hire quality
3) it’s getting better at building decks
4) lit review/data gathering/market mapping first passes

I used to say (like… a year ago) that genAI was as good as a dumb intern. Now it’s as good as a good associate

[Anonymous]

I have an example. I recently started a job that I know nothing about (no experience and no education in the field).

But thanks to AI, I have been able to learn everything I need to know to do my job. A few years ago, I would not have been able to do this.

[Anonymous]

Anonymous wrote:I have an example. I recently started a job that I know nothing about (no experience and no education in the field).

But thanks to AI, I have been able to learn everything I need to know to do my job. A few years ago, I would not have been able to do this.

Could you not have done research?

[Anonymous]

Seems like BS to increase stock prices.

[Anonymous]

Anonymous wrote:Right? I just listened to a long podcast on this and the main examples was that management consultancies wouldn’t be able to charge as much for making PowerPoints and doctors can take notes more easily 😂 well I am happy for the doctors.

As a doctor, writing notes is the bane of my existence. Doctors waste so much time on documentation that is required by insurance, lawyers, etc.

One of the main reasons doctors are late when seeing patients is because they are busy writing “essays” after each patient. It is extremely time consuming.

[Anonymous]

Anonymous wrote:

Anonymous wrote:Heres 3 examples. Yes I used chat to draft but I put in enough detail in my prompt that it came up with real world info info.

We used to have two support analysts manually categorize 5,000 IT support tickets per week. Now a model auto-tags 92% of them with 96% accuracy, and one analyst audits exceptions.

Our paralegals draft first-pass contract summaries. It used to take them hours in many cases. Have you ever read a commercial lease?? It is a monster. Now a model generates a structured summary in 45 seconds and the paralegal reviews and edits; prep time dropped from 1+ hours to under 15 minutes.

We replaced manual fraud screening of 100% of transactions with an AI risk score. Humans now review only the top 3% flagged.

I hope someone is still reading these. Because AI is not that great at summarizing!

I beg to differ. It is excellent at summarizing!

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:What you are saying here is you're not on GitHub, Substack, LinkedIn, or any of the places where people talk about the specific things they're building. You don't go to meetups where people demo their tools. Your version of curiosity is "posting here and demanding people tell you."

This comment exemplifies the problem. The prediction engine works very well in coding and other predictable disciplines. But tech bros are too far out ahead of their skis in so many areas they really don’t understand the lack of utility.

That’s ok. Let’s see where this all leads.

It’s funny that you insult tech bros by using one of the corniest and hackneyed bro expression.

In any event, news flash, there are brilliant people working on this and there is a reason for it. GAI is transformational and will be no matter how many of you complain, all as you don’t even realize you’re using AI constantly throughout your day.

I think we found the tech bro

I’m a woman and I’m old.

Same here. Woman in my 50s. Those who think “tech bros” are here posting about AI…are completely clueless.

[Anonymous]

Anonymous wrote:

Anonymous wrote:I have an example. I recently started a job that I know nothing about (no experience and no education in the field).

But thanks to AI, I have been able to learn everything I need to know to do my job. A few years ago, I would not have been able to do this.

Could you not have done research?

Research would have taken days. AI gives me answers in seconds.

[Anonymous]

Someone I work with was very excited to show me her AI-generated financial analysis. It was crap and she was too excited to want actual feedback, so I said “wow, interesting” and moved on with my day. It was supposed to be color-coded but all the colors were varying shades of red. The numbers were illegible. It was probably 89% wrong, but source data was missing so who knows.

To me AI has been useful for finding legal text or statutes. So far it tends to interpret them incorrectly, but it can at least point me in the right direction.

[Anonymous]

Honestly if you’re saying things like “AI isn’t good at summaries” or “someone did something with AI and it was terrible” then this is user error and/or refusal to use the paid tools. And in some cases it might be that the AI isn’t quite there yet but if you’ve seen how fast it has improved, you would know that it will be there in a few months.

[Anonymous]

Anonymous wrote:My husband and I own a software development firm. We are no longer hiring pure developers; AI is better at it, given you know how to structure your prompts.

To give a very concrete example - I have a niece outside the US who runs a small medical practice. She was shopping around for software to help manage her practice, and running into a Goldilocks problem - most of the good ones were written for hospitals, and overkill/expensive, and most of the small ones were not feature rich, could not be customized for how she does her business, and support was bad/non-existent.

We stood up a hospital management system in a month - entirely customizable, feature rich, fairly easy to modify with requirements changes. And by we, I mean two people - DH who did the coding, I did the testing. A similar system would have taken a whole lot more man hours (years not months). And we can sell it to others in her situation, if we wanted to.

DH developed a study helper for DD in his spare time. It took two days, I believe. You can feed it any material, and it will give you test questions based on the material (you can customize the difficulty level), give you answers, and you can chat with it to get explanations.

A hospital management system? Like something that interfaces with hospitals/EMR?

[Anonymous]

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:I have an example. I recently started a job that I know nothing about (no experience and no education in the field).

But thanks to AI, I have been able to learn everything I need to know to do my job. A few years ago, I would not have been able to do this.

Could you not have done research?

Research would have taken days. AI gives me answers in seconds.

So does google.

[Anonymous]

Anonymous wrote:Honestly if you’re saying things like “AI isn’t good at summaries” or “someone did something with AI and it was terrible” then this is user error and/or refusal to use the paid tools. And in some cases it might be that the AI isn’t quite there yet but if you’ve seen how fast it has improved, you would know that it will be there in a few months.

It... isn't good at summaries. And you're right, I'm not using the paid tools. Why would I pay if the free version has demonstrated it's garbage? "Oh, this is crap, so I should definitely sign up for a paid subscription to get the non-crap version!"