AI - Status Update

 So, he's got a point. All my joking aside against the Rest of the World Vs Me, is the idea of process automation by AI the core interest in corporate world.

And it is not gonna happen.

I was a Senior Specialist in the Knowledge Management Team of a quite sophisticated Help Desk for Clinical Trials before. By telling a Manager of the Pharma Industry that the company I worked for being hired to collect data from the participants created a software that forced the patients without telling them to having to press the Ok button below a smiley to save their data and sending that intel in an email to development and project manger I saved them few million Dollars for average German wage.

What is the process here? The missing auto save, the button or the email chain...

About 80% of all incoming emails were of much simplier requests. Back than no Ollama llama models existed. There was a quite sophisticated text recognition system that was able to analyse from an export into an excel sheet the emails, but that was hardly better than the drop down menues would have if those were targeting also the free text fields of the Specialists.

So that's how it goes: No matter if call or email a Specialist picks it and reads or listen to the issue. Then he categorises the ticket to have a record of the problem. The better the categories are structured, the better the ability to analyse problems of the product, speaking in general terms and the more automated process can be adopted into the company's Computer systems.

Iman now described the sweet dreams of every Financial Controller in every large organisation and the medicine against heart attack of all those that saved on product testing and quality control:

An AI stuffed into a leased Server Rack instead of dozent of intensively and as expensively trained humans saving a fortune.

There is a chance that a verity of the Cyberdeck I created actually can analyse being trained with company data an incoming email, if it is a known problem.

There was a joke at the Help Desk that made even those calling in laugh and that was that most problems were RTFM fixable. Read The Fucking Manual....

More humans have a self-ironical character attribute than most consider, but all humans need social interactions.

The Cyberdeck is actually designed for all other problems as a support tool to find a solution and the existing AI models are made for reasoning while process automation is only in parts in need of reasoning.

There is a huge difference between a Rule Base System and an AI made to reason, which is to check logic of a topic.

That means that most automation tasks need understanding of the processes, breaking those down into steps until actually quite simple switches and triggers do process them.

The Cyberdeck having a link at the right side can help if trained with appropriate data to understand those steps and even to write those scripts depending on which Ollama model and which information it uses, but I strongly recommend to not let it do any decisions really.