DecisionWhy
An internal knowledge assistant that, when you ask 'why did we do X?', surfaces the actual decision memo, ADR or Slack thread where the call was made — and clearly explains why the other 20 'related' documents are not the answer.
Mid-market operations and product managers digging through old decisions in Notion, Confluence, Drive and Slack
- Decision-classifier that ranks retrieved docs by whether they actually make the call, not just mention the topic
- 'Why this and not the other docs' explainer that names which documents look similar but contain no decision
- Connects to Notion, Google Drive, Confluence, Slack and Linear with permission boundaries
- Audit trail showing which doc was cited so reviewers can replay how the answer was produced
The Dev.to post shows a frustration widely felt inside real companies: bigger context windows make confidently-wrong answers more dangerous, and leadership teams keep asking 'why did we choose X?' for decisions that are now years old.
Real, frequently-discussed pain (multiple 2024-25 articles on AI for ADRs/decision logs), but the specific Dev.to '12 reactions' claim could not be verified via search, so the 'widely felt' framing is unconfirmed.From ADRs to Action: Embedding Architectural Decisions into Dev ↗Documentation Architecture: Using ADRs and Decision Logs for AI-Generated Systems ↗
Adjacent players exist (Glean for general enterprise search, LLM-ADR creation assistants) but none I found specifically answer 'why did we choose X?' with explicit non-answer reasoning across Slack+Notion+Confluence+Drive.Using generative AI as an architect buddy for creating architecture decision records ↗Architectural Decisions: A Human-Led, AI-Powered Approach ↗
Enterprise AI-search category (Glean, Hebbia, Guru) has demonstrated $20-100+/seat willingness to pay from mid-market buyers; PM/ops tooling routinely lands $15-40/user/mo with strong retention.From ADRs to Action: Embedding Architectural Decisions into Dev ↗
Decision rationale debt compounds forever — every quarter makes 'why did we pick vendor X?' harder to answer, so the need is durable and grows with org size and tenure of staff.Agentic AI for Living Architecture: Enhancing ADRs with LLMs and Decision ↗
Doable but non-trivial: requires connectors + auth for 4+ SaaS sources, per-source permission mirroring, and reliable exclusion/negation reasoning — a known weak point for current LLMs.