AI realities every professional should know before they end up burning money.
Microsoft has decided to cancel its internal team’s AI licenses because they’ve finally started to realise what works.
If you’re a professional, you need to know these realities before you end up burning your money (and time)
It was always clear from the beginning that humans have a LOT more context, intelligence and nuance than LLMs.
AI is a slot machine, not a vending machine. You cannot fire carpenters and just expect new nail guns to do all the work.
I just read a tweet that says
“Bragging about how much software you’re shipping with Al is like holding down the shutter button and bragging about how many photos you took.”
As Mo Bitar says, you might click thousands of pictures with a better shutter but you’ll waste equal amount of time shortlisting the one that’s worth uploading.
LLMs are super good at first mile problems but industry grade solutions require a lot of context, data binding, precision and judgment, which only humans can pull off with high reliability.
Tokens are getting 5x cheaper but with the new reasoning models, consumption is getting 10x higher. When you are not trained properly on how to use AI, the consumption goes 20x, productivity slumps and chaos reigns.
The correct way to adopt AI is to audit specific steps in low stake/mundane work in departments and then have the most experienced people from those teams test LLMs and validate their outputs.
Document what’s working and what is not working in isolation, before rolling out licenses to everyone.
Drive AI Adoption in phases, where only shortlisted, reliable and trained professionals are experimenting on reasoning LLMs.
Let those handful of AI champions use AI in modules, not on the entire stack with just one shot prompting.
Firing thousands of people will only lead to loss of context, increase in anxiety of those who stay and a terrible precedent for AI adoption.