Day 79šŸ’» EngineeringIntermediate

Architecture Decision Recorder

Document architectural decisions in ADR format from design discussions and meeting notes.

1 hoursaved
Development • Technology
ConfluenceSlackGitHubClaude

The Challenge

Architectural decisions happen in meetings and Slack threads but rarely get documented. When questions arise later, teams can't find why decisions were made.

The AI Desk Solution

AI Desk creates Architecture Decision Records from discussions, capturing context, options, and rationale.

The Workflow

Step 1: Decision Detection


Trigger: Design discussion or meeting

Sources: Slack threads, meeting notes, docs

Step 2: ADR Generation

  • Extract decision context
  • Document alternatives
  • Capture rationale

Step 3: ADR Document


šŸ“‹ ADR-047: Message Queue Selection

STATUS: Accepted

DATE: Apr 12, 2026

DECIDERS: Platform Team

CONTEXT

We need a message queue for the new

event-driven architecture. Current volume

is 10K messages/day, projected to grow

to 100K within 6 months.

Requirements:

ā”œā”€ā”€ At-least-once delivery

ā”œā”€ā”€ Message ordering within partition

ā”œā”€ā”€ Horizontal scalability

ā”œā”€ā”€ Cloud-native (AWS preferred)

└── Team familiarity consideration

OPTIONS CONSIDERED

Option 1: Amazon SQS

ā”œā”€ā”€ Pros: Simple, managed, cheap

ā”œā”€ā”€ Cons: No ordering, limited features

└── Team experience: High

Option 2: Amazon Kinesis

ā”œā”€ā”€ Pros: Ordering, replay, scalable

ā”œā”€ā”€ Cons: More complex, higher cost

└── Team experience: Medium

Option 3: Apache Kafka (MSK)

ā”œā”€ā”€ Pros: Full-featured, industry standard

ā”œā”€ā”€ Cons: Operational overhead, cost

└── Team experience: Low

DECISION

We will use Amazon Kinesis because:

ā”œā”€ā”€ Ordering requirement is critical

ā”œā”€ā”€ Replay capability valuable for debugging

ā”œā”€ā”€ Managed service reduces ops burden

ā”œā”€ā”€ Team can skill up (training planned)

CONSEQUENCES

ā”œā”€ā”€ Additional training needed (2 weeks)

ā”œā”€ā”€ Higher cost (~$500/mo initially)

ā”œā”€ā”€ Better debugging capabilities

ā”œā”€ā”€ Easier future migration to Kafka

└── Design pattern documentation needed

RELATED DECISIONS

ā”œā”€ā”€ ADR-042: Event-driven architecture

ā”œā”€ā”€ ADR-045: AWS as primary cloud

└── ADR-046: Microservices boundaries

Value Proposition

  • Time Saved: 1 hour per ADR
  • Institutional Memory: Decisions captured
  • Onboarding Aid: New engineers understand why

Part of the 100 Days 100 Usecases campaign. View all usecases

Ready to automate this workflow?

AI Desk connects your enterprise tools and models to execute this usecase in your organization.