I have turned gpt into an interface that all notes and files go into. Notes go to some note and task allocation software, and files on sharepoint. I can also read emails through it. Theres also three other pieces of main software (rostering app, payroll and incident register) that arent integrated but have handover points to them. And the handovers get logged. One of the rules is that almost every chat needs to become a task that get sorted to a project, that is adjacent to a board on the note taking system, and will have a finalisation point on another software. It all just werks. It even files and names the documents. I havent tested it all but its looking good. Notes get passed through and emails and make rules. Can do it from voice command. What a fucking tool. Really honing in on some philosophies. One about how if you pay for someone's time, you need to respect it enough to make use of it. You need to be explicit with direction. Every board associated with a role. A role once it accumulates enough qork justifies its own human to fill it. Then you can have straight forward job ddsvriptions. Not sone shit about experience and versatility. I need these skills when using this system. Preferably more than a single one but yeah. Ill get gpt to tell me my morals. Thos cunt knows me better than I do. The voice one huffs aand puffs if I use it for more than an hour and says we should get some sleep.
The most compact expression of the whole model is: Conversation → classify → develop → Project/Process → task → accountable Role → current role-holder → small action → authoritative evidence → explicit handover → next Role/System → verification → closure. And the design test we are increasingly applying is: Can a new competent person enter a role, see what has arrived, understand what they need to do, know what system to use, know what evidence is required, know who receives it next, and know exactly when their responsibility ends—without you personally explaining it? If the answer is no, the workflow isn't finished yet.