Incentivizing the Man on the Floor
I work in a project X style department at a mid sized European company. I have designed and built (with the help of an external software consultancy) a software system that updates global users on market prices within our industry, doing various types of calculations with the input I provide. The software is great, but the input of market prices feeding into the system is too little for the software to really kick off and be the best.
In order to increase the quality of the market prices I therefore need to engage a global broker within my industry to get them to update prices globally, which would make my software not only the best in the world, but build a moat around the idea as no one can make something much better.
There are two problems: 1. What do I offer the partner in return? 1. Answer: equity, money, exposure, etc. I am not so bothered about reaching an agreement with the c suite of the company as the idea is good and relatively risk free for them. Also, management love big data project and “digitalization”. 2. This is a very manual process, so the 50-100-200 employees of my partner would need to daily go into a dashboard and plug in 10 numbers each. Not a huge effort, but I have a hard time seeing what could incentivize them, after all they are paid on comission they make and not inputs. This is where I feel it would fall flat, as no inputs make the software less valuable - the whole idea is to sell “live” prices anywhere in the world.
How would you guys think about such a problem? I wonder why this manual process can’t be automated? Put cameras everywhere and then use AI to update prices. Problem solved. My thought was AI, but I believe that there is to much room for AI to misunderstand the communication (which happens via instant messaging). However, a hybrid approach could perhaps be the way to go: I.e. monitor everyone’s chats and then scrape what looks like pricing info, which a human then translates into prices. Is it realistic to expect this to be fully automated anyway? It can be done with a language model very easily. My gut feeling says no, as there are: 1. different maturities
2. 24 different regions
3. Four different pricing categories So the complexity of assigning the right input value to the right pricing curve is quite high. That's not complex at all. Well not for humans, but I believe it is for ChatGPT given the incomplete information that is often found in chat when users text each other this information (and which is why it is valuable to structure it, package it and make it transparent for all globally).