# Why does enterprise architecture still run on tribal knowledge?

> Canonical: https://decisionpen.com/blog/why-enterprise-architecture-runs-on-tribal-knowledge  
> Published: 2026-10-17 · DecisionPen · Living EA series (A1)  
> Related docs: /docs/enterprise-ask, /docs/it-knowledge-continuum, /docs/why-not-copilot-or-rag

## Summary

“Let me check with Dave” means the real dependency map is not in the system of record. Hard questions route to a person, creating queues and key-person risk. Tribal knowledge is institutional memory trapped in people; useful memory is curated, shared, and queryable with evidence.

## What does “let me check with Dave” really mean?

Blast-radius, integration history, and “why is it like this?” questions escalate because How and Why never entered a shared model. People ask: who knows how our systems talk; key person risk in architecture; why can’t we answer impact questions without meetings.

## Why do repositories fail the questions leaders ask?

Inventories answer “what applications do we have?” They struggle with “what breaks if we change this?” when behaviour and rationale were never modelled.

## Is tribal knowledge the same as institutional memory?

Tribal knowledge is institutional memory trapped in individuals. Shared institutional memory is curated and askable — see also why EA factsheets miss How and Why.

## Can ChatGPT or Copilot replace tribal knowledge?

Not reliably alone. Generic LLM or document RAG without a curated metamodel and living map invents, mixes outdated docs, or misses constraints.

## What changes when context is living and askable?

Map-grounded answers with citations; experts curate and decide instead of being the only search engine (Enterprise Ask on a curated landscape).

## How does DecisionPen solve this?

Living metamodel landscape updated from real work; How/Why evidence on entities; map-grounded Enterprise Ask with citations (not ungoverned Confluence chat); experts curate proposals instead of queuing every question. Trial: https://decisionpen.com/trial

## When is DecisionPen not the right tool?

Primary need is freeform diagramming or BPMN drawing (Visio, Lucidchart). DecisionPen is living context and generated views, not a canvas.

## FAQ

**What is tribal knowledge in enterprise architecture?**  
Critical how-and-why knowledge mainly in people’s heads, not a shared curated model.

**Why do blast-radius questions fail without shared context?**  
Thin attributes in tools; real dependencies elsewhere — escalate to key people.

**Can ChatGPT or Copilot replace that tribal knowledge?**  
Not reliably without curated metamodel context and evidence.

**When is DecisionPen the wrong tool?**  
When you need a primary diagramming or BPMN drawing tool.

## Related posts

- https://decisionpen.com/blog/why-ea-factsheets-miss-how-and-why  
- https://decisionpen.com/blog/why-solution-architects-rediscover-the-landscape