Here's a plan, Marlene

 Do you really want a plan, Marlene?

Here's your plan: You can imitate a land-based food chain or an ocean-based food chain; a rain forest or a coral reef.
Don't be put off by the term 'food chain'. It's just the best terms science can provide. In the end, the top predator returns to the microbes, after all. There really is no top or bottom.

In fact, research indicates the small things and their weak interactions do matter. To quote a first primary principle:
"Weak interactions are the stabilizers and strong interactions the destabilizers of food webs.
In some circles in the late twentieth century, small may have been considered “beautiful,” but in the early twenty-first century, we can say that small is “important.”

A species or interaction that appears insignificant today might be crucial tonight or next year."
endquote

Your decision for developing any process to attain any goal will determine which of the two food chains you want to emulate.

Things don't have to 'die' to be consumed though. Everything produces products; even waste is considered a product. That's the core concept behind sustainability and reliability.
Products contain value and information. That's the second core concept.

Even the top predator leaves behind products that affect the reliability of other producers: carcasses, offal, and finally their own carcass. These products provide value and information to different levels of other processes.
Ignoring this concept is the reason so many processes today don't seem reliable. Put very simply, it means no process or goal is independent of others.

How stable or unstable the process you create will be depends upon these choices in your design. You'll find you have to combine the concepts within each design to achieve different types of reliability.
If you want sustainability, you have to choose between the coral reef and the rain forest though.

If you think about it, the systems in place today are not that far away from these concepts. All this really means is adding a few priorities in design and implementation; and interpreting more accurately the products of all levels.

There. There's your plan.
To implement it, just run a network analysis and interpret the translations correctly.