What works with AI for me (and what doesn't)
AI can be insidious.
It builds you a perfect landing page. It's hard not to be impressed because it has great design chops.
But does it align with your goals and the problem you really want to solve?
AI cannot reliably judge whether its output aligns with your actual goals when it isn’t fed with a lot of meaningful context, and even then ...
AI seems 'intelligent', but let’s not forget the 'artificial' part of it.
I hear this most from domain experts where specific nuance matters. Think lawyers, accountants, etc., and I experience this myself with more strategic tasks → suboptimal results.
I also see this with writing. For documentation it’s fine because it’s more mechanic, but for content it needs to be more original and genuine.
It’s very hard to emulate your own unique voice, without producing some kind of slop. AI writing can be generic (robotic), overly verbose, and just less creative.
On the other hand, coding is more exact and there are more specific ways to verify its output. Hence I see better results there.
Use AI as an accelerator, not a compass, because it can and will drift if you’re not closely steering and watching it.
I find LLMs useful for research, but even then, LLMs tend to agree with what you hint them at and they can be overly pleasing (aka sycophancy), so nothing beats human brainstorm sessions.
I think it's important in this fast-paced tech evolution to share what works and what doesn't, and train more people and companies how to use AI well. 🚀
Please reply and let me know an AI success and failure story of your own, curious to hear. 💡
Have a great rest of your week.
Bob
If you know somebody who could benefit from this message, please forward this email. 🙏