When the team has decided that a thing must be built, where do you start?
In the old times, we would draft specifications, run it by a committee of senior engineers, revise it, check it again for correctness, and scrutinize it endlessly before our engineers would meticulously render each decision.
More often than not, we would build monuments to engineering perfection and at the end, we would find that our specifications were inadequate in one dimension or another. We built a castle but what we really needed was a Burger King. Maybe we can install a kitchen and even a drive-thru, but it’s still a castle, but it’s always going to have weird edges.
Definitely what you can not do, CAN NOT DO, is decide that the system is too complicated so what we really need to do is rebuild Castle Burger King in smaller bits but put a network between them and it’ll be easier to understand. The small bits might be individually easier to understand, but the system (aka Burger King) is actually what you still have to understand. And the system is still a Burger King defined in terms of what a castle has to offer.
So how do you build Burger King to begin with?
You don’t, you can’t, you don’t even know that’s what you need (and probably nobody does). The customer came to you and said “I need a place to eat.” So what you should do is take them very literally and rent them a table at a cheap restaurant and ask them if that solves their problem.
And then you keep going until you arrive at a Burger King, and then you keep on going, because even Burger King can’t stay the same forever.
The not-knowing bit can be uncomfortable, and we try to mitigate it with planning and thinking really hard. In the end though, all that planning and thinking is going to die on a wiki page and reality is going to show us the truth (occasionally, with violence). Given enough time and patience, even the greatest castle will eventually converge to what it must become. So don’t build castles. Build hovels. They might not look as nice, but you’ll get to Burger King at lot faster.