Yeah exactly.

Instead the tools we build need to enhance and empower people. Take away the need for manual work and doing menial tasks and instead focus on the higher level work of defining and understanding the processes that power the underlying business which can then be run in an automated way.

The promise of fully automated "AI accounting" in my opinion is not the right way.

It's similar to how so many startups have relied on Bench for all of their accounting because they did not want to have to think about it. And now they need to learn to own their numbers again.

The magic of LLMs in my opinion is in the ability to make use of data sources in really new ways without building specialised machine learning tools.

But this should often more of a supporting feature than really replacing humans in the process.

Yes, LLMs allow fundamentally new use cases and make it easier to develop features that in the past required complicated solutions - which were often delivered in clunky interfaces.

But it remains important to use LLMsin a way that makes sense. It's not enough to slap an LLM into a tool.

My opinion is that make the same mistake as I think so many others make as well which is to think that AI in software is fundamentally different than how software has worked in the past.

What I mean by that is that the problems and opportunities they see have been playing out for a while already.

A team that is growing (in capability, not number of team members) will hit different bottle necks each time as old ones are removed.

Good retrospectives really help with that.

Just read and very much enjoyed this due to your post so: thank you!

For customers or for investors?

I'm writing go professionally without having read any actual books to date. Looking forward to filling some of the gaps.

Nope but just ordered it. I'm normally not big on reading programming books but got some development budget to spend this year.

Which one?