Engineering Leadership Daily Digest

01 · Business Problem

The Problem

Engineering leaders at fast-moving orgs face a signal-to-noise crisis every morning. Slack threads contain escalations and team updates buried in 200+ messages. Jira has sprint blockers and delivery risks spread across dozens of issues. The calendar holds back-to-back reviews requiring preparation — none of it synthesized. Leaders spend 45–60 minutes manually triangulating across tools before they can lead.

The Solution

A 5-agent n8n workflow that runs automatically each morning. Three specialist agents independently analyze Slack, Jira, and Calendar data. An Executive Synthesis agent cross-correlates the outputs — surfacing the connections a human would miss. A Publisher agent formats the final briefing and delivers it to the engineering-leadership Slack channel — ready before the first standup.

02 · By the Numbers

AI Agents

Data Sources

Workflow Nodes

Slack Channels

LLM Tiers

∞

Daily Runs

03 · Data Sources

Slack

Team messages, escalations, announcements, incidents from the engineering-leadership channel — read from Google Sheets Slack_Data tab.

Google SheetsSlack_Data tab

Jira

Sprint issues, blockers, bugs, delivery risks, and execution status — read from the Jira_Data tab in the same spreadsheet.

Google SheetsJira_Data tab

Calendar

Executive meetings, customer-facing events, decision reviews, scheduling conflicts — read from the Calendar_Data tab.

Google SheetsCalendar_Data tab

04 · Sample Digest Output (Slack)

📊 Executive Engineering Leadership Daily Digest

📝 Executive Summary

Engineering execution is at moderate risk today. Authentication latency degradation (P0) aligns with an open Jira blocker and dominates Slack thread volume. Sprint 14 delivery is on track overall but gated on BUG-315 resolution. The afternoon calendar is dense with back-to-back leadership reviews requiring preparation.

🚦 Overall Status

Moderate Risk Confidence: High

🎯 Top Priorities

Resolve authentication service latency — P0, escalation active in Slack
Unblock BUG-315 blocking mobile notifications — Sprint dependency
Prepare materials for Executive Product Review at 2pm

⚠️ Delivery Risks

Authentication degradation may delay Sprint 14 release
Back-to-back afternoon schedule leaves no buffer for incident response

📅 Executive Calendar

Sprint Planning — 9:00 AM (backlog ready?)
Customer Escalation Review — 11:00 AM
Executive Product Review — 2:00 PM

✅ Recommended Leadership Actions

Confirm incident commander for authentication degradation
Assign BUG-315 owner before Sprint Planning
Finalize Executive Product Review deck by noon

Agent Pipeline

⏰

Schedule
Trigger

Daily 7 AM

→

⚙

Orchestrator
Config

Role, flags

→

🛡

Guardrail
+ State

Pre-flight

→

Slack
Agent

gpt-4o-mini

Jira
Agent

gpt-4o-mini

Calendar
Agent

gpt-4o-mini

→

Executive
Synthesis

gpt-4o-mini

→

Publisher
Agent

gpt-4o

→

Slack
Channel

#eng-leadership

Layer 1

Control — Where Decisions Are Made

Pre-flight configuration, validation, and execution context

Orchestrator Configuration

Central Set node defining role, enabled sources, flags, and execution metadata. Now also sets minimum record thresholds: Slack ≥ 2, Jira ≥ 2, Calendar ≥ 10. If any source returns fewer records than its threshold, that source is skipped — graceful degradation starts here at the control layer.

Guardrail Check

IF node with real OR logic — checks slack_enabled OR jira_enabled OR calendar_enabled. At least one source must be enabled to proceed. Previously empty; now actively validates the configuration before any data is fetched.

State Management

Stores current_run_timestamp and session_id. Tracks delta_tracking_enabled flag for future runs. Foundation for comparing new vs. recurring digest items across executions.

Layer 2

Execution — Where Work Happens

Parallel specialist agents + synthesis + publishing

Slack Intelligence Agent

Analyzes all Slack records as a single executive communication assessment — not per-message. Outputs: summary, critical_issues, announcements, team_highlights, risks, recommended_actions. Model: gpt-4.1-mini.

Jira Sprint Agent

Synthesizes all Jira issues into one sprint health assessment. Identifies blockers, delivery risks, and execution trends. Outputs structured JSON: sprint_health, critical_issues, work_in_progress, delivery_risks. Model: gpt-4.1-mini.

Calendar Intelligence Agent

Identifies executive meetings, customer-facing events, decision points, scheduling conflicts, and preparation requirements. Outputs structured calendar intelligence JSON. Model: gpt-4.1-mini.

Executive Synthesis Agent

Receives normalized outputs from all 3 specialist agents. Deduplicates, cross-references (e.g. Slack incident ↔ Jira blocker), and produces one unified executive briefing JSON. Model: gpt-5-mini for deeper reasoning.

Executive Publishing Agent

Transforms the structured synthesis JSON into polished Markdown formatted for Slack — emoji headers, bullet priorities, concise executive language. Does NOT generate new analysis; formats only. Model: gpt-5-mini.

Layer 3

Reliability — Where Resilience Lives

Output schemas, structured parsing, graceful degradation

Structured Output Parsers

Each specialist agent is paired with a JSON Schema output parser (autoFix: true). If the LLM output deviates from the schema, n8n auto-repairs rather than crashing — preventing pipeline failure from a single bad response.

Two-Tier LLM Strategy

Specialist agents (high volume, structured task) use gpt-4.1-mini for cost efficiency. Synthesis and Publishing agents (complex reasoning, final output quality) use gpt-5-mini. Tiered cost without sacrificing executive output quality.

Graceful Degradation

If one Google Sheets tab is empty or unavailable, the merge nodes (combineByPosition) continue with partial data. The synthesis agent notes which sources were unavailable — the digest still publishes rather than silently failing.

Error Trigger & Error Handling

A dedicated Error Trigger node catches any workflow execution failure and routes it to a #errors Slack channel — automatically posting the workflow name, execution ID, failed node, error message, and a direct link to the execution log. Zero silent failures.

Confidential Routing

Each record is checked for confidential flags via per-record IF nodes. Confidential items are tagged and routed through a 3-input merge + JS formatter to the #confidential-leadership Slack channel. Non-confidential output goes to #engineering-leadership. Dual-channel delivery with zero cross-contamination.

Design Notes & Stretch Goals

Production Enhancements

Delta Tracking: Store each run's digest and tag items as [New] vs [Recurring] across executions using the State Management layer.

Persistent Memory: Connect Pinecone or Weaviate so the synthesis agent can reference historical sprint patterns across weeks.

RAG Integration: Add a retrieval step allowing the synthesizer to query a runbook or incident history knowledge base for context.

Architecture Decisions

Parallel Execution: Slack, Jira, and Calendar agents run in parallel from State Management — total wall time is bounded by the slowest agent, not the sum.

Single Spreadsheet, Multiple Tabs: All three data sources live in one Google Sheets doc (different tabs) — simplifies credential management and data refresh.

Normalize → Merge: Each agent output is normalized to a keyed JSON ({slack:…}, {jira:…}, {calendar:…}) before merge — prevents key collision in the synthesis prompt.