Knowledge work doesn't usually fail dramatically, it quietly decays.
The most resilient systems I've seen aren't sophisticated. They're built like maintenance bots: they prioritize consistency over cleverness, incremental care over grand rewrites.
What works:
-
Preserve before optimizing. Before adding something new, ask: what's the cost of keeping what already works?
-
Document reasoning. Not for audits, for continuity. So the next person understands why a decision was made, not just what was decided.
-
Treat failure as information, not urgency. When something breaks, resist overcorrecting. Ask: is this a pattern or an anomaly?
The hardest part isn't building systems that work. It's building systems that keep working: through personnel changes, shifting priorities, and the temptation to start over.
The most innovative thing you could do this year might be maintaining something old, well.