What is organizational memory?

Explicit and tacit - and the gap between

The classic distinction: explicit knowledge is what's written down - docs, wikis, runbooks, code. Tacit knowledge is what lives in people - judgment, context, the reasons behind the rules. Most organizations over-invest in the first and ignore the second, which is backwards: documents without their reasoning age into cargo cult (“we've always done it this way”), while the tacit layer - why it's done this way - walks out the door with every departure. The most valuable slice of organizational memory is exactly the part standard documentation practice captures worst: the reasoning.

Why engineering organizations forget

Not through negligence - through structure. Turnover converts tacit knowledge into vacancy on a schedule; the average tenure at technology companies is short enough that a system routinely outlives everyone who shaped it. Reorgs break the informal “ask Priya, she was there” networks that substituted for records. Growth means most engineers eventually work in systems whose formative decisions predate them. And the reasoning that does get externalized lands scattered across tickets, pull requests, chat threads, and meeting notes - stored, technically, but not retrievable as an answer to “why?”.

What good looks like

A useful test, in three questions. Can a new engineer discover why the system is shaped the way it is - without finding the right veteran to ask? When a settled question resurfaces, does the original reasoning resurface with it - or does the debate restart from zero? When a decision is replaced, does the organization keep what it believed before and why it changed - or does the old context simply vanish? Teams that pass this test share one trait: decisions are recorded with their reasoning, connected to their evidence, and kept current - whether through disciplined manual practice, tooling, or both.

Memory is the durable asset of the AI era

AI has made this old problem newly urgent. Every team will soon have capable AI assistants, and the models themselves are converging into commodities - the same frontier capabilities, available to you and your competitors alike. What a model cannot conjure is your organization's accumulated judgment: years of decisions, tradeoffs, and outcomes, curated and connected. An assistant grounded in that memory gives answers grounded in your history; the same assistant without it gives generic ones. Which makes organizational memory the rare asset that appreciates as AI improves - every better model amplifies whatever memory you feed it.

That thesis is why Rationale exists: it builds organizational memory as infrastructure, reading the artifacts your team already produces and maintaining a connected, current Decision Map of what you decided, why, what you traded away, and what came next - so the memory compounds instead of evaporating.