Build a Parallel AI Agent Pipeline

Welcome

Read before starting

Parallel AI Agent:
3 Competitors, 1 Dashboard

You built a sequential pipeline in the last workshop — each stage waited for the previous one. Now you'll build something fundamentally different: three agents that run at the same time, each researching a different competitor, whose results are merged into one dashboard.

⇉ What is AI Agent Type 4 — Parallel Execution?

A coordinator agent fans out tasks to multiple specialised agents running simultaneously. Each agent works independently — no waiting, no handoffs. A merge agent then aggregates all results into a single coherent output. Wall-clock time equals the slowest branch, not the sum of all branches.

Why Parallel Beats Sequential for This Task

Sequential (Type 3)

~90 sec

Parallel (Type 4)

~30 sec

3 competitor analyses running at once instead of one after another.

Your 3 Target Competitors

Agent	Target	Why Relevant to ANCI
Agent A	Calendly	Largest player in scheduling; most of your prospects compare you to them first
Agent B	Reclaim.ai	AI-first scheduling; overlaps most with ANCI's agent positioning
Agent C	Motion	AI priority scheduling; strong in productivity space you're moving into

Pipeline Stages

Stage 1 — Fan-Out · 3 agents in parallel

Coordinator spawns 3 independent fetch+analyse agents simultaneously. Each researches one competitor: GitHub activity, product page signals, job posts, and G2/Capterra reviews.

Stage 2 — Merge · 1 aggregator agent

Reads all three JSON outputs and produces a unified competitive analysis with cross-competitor comparison, priority signals, and strategic recommendations.

Stage 3 — Dashboard · 1 renderer agent

Renders the merged analysis as a polished HTML comparison dashboard — one row per competitor, one column per signal dimension.

What You Need

✓
Claude Code installed — run npm install -g @anthropic-ai/claude-code if not yet installed
✓
A terminal open and ready — Claude Code runs from the command line
✓
You completed Type 3 (sequential workshop) — this builds directly on that foundation

Step 1 of 4

🔵 Fan-Out Agent · ⏱ 6 minutes

Stage 1: Fan-Out Agent
3 Competitors, Simultaneously

This single prompt tells Claude Code to launch three parallel research operations — one per competitor. Claude Code will write async code that fires all three requests at the same time and waits for all to complete before saving.

The Parallel Difference In the sequential workshop, each stage waited for the last. Here, Claude Code writes Node.js async code using Promise.all() — all three competitor fetches fire at once. You see the results appear nearly simultaneously.

How to Run Stage 1

Open your terminal and create a working directory: cd ~/Desktop && mkdir anci-intel && cd anci-intel
Launch Claude Code in this directory: claude
Copy the full Stage 1 prompt below and paste it into the Claude Code prompt. Hit Enter.
Watch Claude Code write and run the parallel fetch script — you'll see all 3 competitors fetched before any analysis begins.
Wait for confirmation that competitor_data.json has been saved before moving on.

Stage 1 — Parallel Fan-Out Prompt

STAGE_1_FANOUT.txt — Paste into Claude Code terminal 📋 Fan-Out Agent

You are a parallel fan-out coordinator agent. Your job is to fetch data about THREE competitors simultaneously using async parallel execution — NOT sequentially. ## YOUR TASK Write and execute a Node.js script that uses Promise.all() to fetch data about these three competitors AT THE SAME TIME: Competitor A: Calendly (calendly.com) Competitor B: Reclaim AI (reclaim.ai) Competitor C: Motion (usemotion.com) ## FOR EACH COMPETITOR, fetch the following in parallel: 1. GitHub data (if they have a public repo): - For Calendly: search github.com/calendly — if no public org, note it - For Reclaim: search github.com/reclaim-ai — repo stars, recent issues, contributors - For Motion: search github.com/motion — repo activity or note if private 2. Web intelligence signals (use web search or direct URL fetch): - Product positioning: their main value prop from homepage H1/hero - Pricing tier count and lowest paid price visible on their pricing page - Last 3 blog post titles (signals what they're thinking about) - Top 3 job titles currently hiring (signals roadmap investment) ## IMPLEMENTATION REQUIREMENTS - Use Promise.all() or Promise.allSettled() — never await sequentially in a loop - Use the GitHub REST API for repo data (no auth needed for public) - For web data, use fetch() on their public pages or use web search - Handle failures gracefully: if one competitor's fetch fails, continue with the others - Log to console when each parallel fetch STARTS and when it COMPLETES, so we can see the parallelism ## OUTPUT FORMAT Save a single file called competitor_data.json with this structure: { "fetched_at": "ISO timestamp", "execution_mode": "parallel", "competitors": { "calendly": { "github": {...}, "web": {...}, "fetch_duration_ms": number }, "reclaim": { "github": {...}, "web": {...}, "fetch_duration_ms": number }, "motion": { "github": {...}, "web": {...}, "fetch_duration_ms": number } }, "total_duration_ms": number, "sequential_equivalent_ms": number } Include both total_duration_ms (parallel actual time) and sequential_equivalent_ms (sum of all individual durations) so we can show the time saved. After saving, print: "✅ Stage 1 Complete — competitor_data.json saved" "⚡ Parallel time: [X]ms | Sequential equivalent: [Y]ms | Time saved: [Z]ms"

Stage 1 Success Checklist

Claude Code confirmed competitor_data.json was created
You can see the parallel vs sequential time comparison in the terminal output
The JSON file contains data for all 3 competitors (even if some fields say "not available")

Step 2 of 4

🟣 Merge Agent · ⏱ 5 minutes

Stage 2: Merge Agent
Aggregate 3 Outputs Into One Analysis

The merge agent reads all three competitor JSON blobs and produces a single unified competitive analysis. It compares across competitors, flags the most important signals, and generates strategic recommendations specifically for ANCI.

Why a Separate Merge Stage? The fan-out agents are specialists — they fetch and structure data. The merge agent is a strategist — it reasons across all three. Keeping them separate means you can swap any fan-out agent (add a 4th competitor) without touching the merge logic.

Stage 2 — Merge Agent Prompt

STAGE_2_MERGE.txt — Paste into the same Claude Code session 🟣 Merge Agent

You are a competitive intelligence merge agent. Read competitor_data.json and produce a unified cross-competitor analysis. ## YOUR TASK Read competitor_data.json from the current directory. Produce the following analysis and save it to competitive_analysis.json: ## 1. COMPETITOR PROFILES (one per competitor) For each of the 3 competitors, create: { "name": "Calendly | Reclaim | Motion", "positioning": "their core value prop in one sentence", "pricing_entry": "lowest paid tier price", "github_activity": "active/moderate/low/private", "hiring_focus": ["top 3 roles they're hiring"], "content_focus": ["what their last 3 blog posts signal"], "openness_score": 1-10 // 10 = fully open source, 1 = fully closed } ## 2. CROSS-COMPETITOR COMPARISON Create a comparison matrix: { "dimension": "AI Features | Pricing Accessibility | Open Source | Enterprise Focus | SMB Focus", "calendly_score": 1-10, "reclaim_score": 1-10, "motion_score": 1-10, "anci_opportunity": "where ANCI has the clearest gap to exploit" } Produce one object per dimension. ## 3. COMPETITIVE SIGNALS (top 5, ranked by urgency) [ { "signal": "one-sentence observation across the competitive set", "source_competitors": ["which competitors this applies to"], "relevance_to_anci": "specific implication for ANCI AI", "urgency": "act_now | watch | monitor", "recommended_action": "one concrete thing ANCI should do or decide" } ] ## 4. WHITE SPACE ANALYSIS Based on all three competitors, identify 3 specific gaps none of them are filling well — where ANCI could position with minimal competitive friction. ## 5. ANCI POSITIONING RECOMMENDATION A single paragraph: given what all 3 competitors are doing, what is the sharpest, most defensible positioning statement for ANCI that avoids competing head-on with any of them? Save to competitive_analysis.json and print: "✅ Stage 2 Complete — competitive_analysis.json saved with [N] signals and [N] white space opportunities"

Stage 2 Success Checklist

Claude Code confirmed competitive_analysis.json was created
The analysis includes a cross-competitor comparison matrix
There are at least 3 white space opportunities identified
The ANCI positioning recommendation section is present

Step 3 of 4

🟢 Dashboard Agent · ⏱ 4 minutes

Stage 3: Dashboard Agent
Render the Intelligence Report

The final agent reads the merged analysis and renders a polished HTML competitive intelligence dashboard — a side-by-side comparison of all three competitors with strategic callouts for ANCI.

Stage 3 — Dashboard Agent Prompt

STAGE_3_DASHBOARD.txt — Paste into the same Claude Code session 🟢 Dashboard Agent

You are a dashboard rendering agent. Read competitive_analysis.json and generate a polished HTML competitive intelligence report. Save the output as anci_competitive_dashboard.html. ## DESIGN REQUIREMENTS - Clean, professional light background (#F8FAFC) - Inline CSS only — no external dependencies - Primary color: #1A6BFF (ANCI blue) - Positive signals: #10B981 (green) - Risk/urgent: #EF4444 (red) - Watch signals: #F59E0B (yellow) - Font: system-ui, sans-serif ## SECTIONS TO INCLUDE 1. HEADER Title: "ANCI Competitive Intelligence Report" Subtitle: "Parallel Agent Analysis — Calendly vs Reclaim vs Motion" Date: today's date Tag: "Generated by Type 4 Parallel Agent Pipeline · 3 agents ran simultaneously" 2. PERFORMANCE CALLOUT (top of page, prominent) Show: parallel execution time vs sequential equivalent Visual: two horizontal bars — parallel time (shorter, green) vs sequential time (longer, red) Label: "Time saved by parallel execution" 3. COMPETITOR CARDS (3 columns, side by side) One card per competitor showing: - Company name and positioning - GitHub activity badge - Pricing entry point - Hiring focus (top roles as tags) - Content focus signals 4. COMPARISON MATRIX (full-width table) Rows = dimensions (AI Features, Pricing, Open Source, Enterprise, SMB) Columns = Calendly | Reclaim | Motion | ANCI Opportunity Each cell shows the score (1-10) as a bar or number + the opportunity column in blue highlight 5. COMPETITIVE SIGNALS (priority-ordered cards) Each signal as a card with: - Signal text - Which competitors it applies to (as tags) - Urgency badge (🔴 Act Now / 🟡 Watch / ⚪ Monitor) - Recommended ANCI action 6. WHITE SPACE MAP 3 boxes highlighting gaps none of the competitors fill well Each box: title + explanation + "ANCI Opportunity" label 7. POSITIONING RECOMMENDATION Full-width section with ANCI's recommended positioning in large text Styled as a strategic directive, not just a paragraph 8. FOOTER "Generated by a Type 4 Parallel AI Agent Pipeline | 3 agents ran simultaneously" "Stage 1: Parallel Fan-Out → Stage 2: Merge & Analyse → Stage 3: Dashboard" "Powered by Claude Code + ANCI AI · anci.app" After saving, print: "✅ Stage 3 Complete — Open anci_competitive_dashboard.html in your browser"

Stage 3 Success Checklist

Claude Code confirmed anci_competitive_dashboard.html was created
You opened the file in a browser — 3 competitor cards are visible side by side
The parallel vs sequential time comparison is shown at the top
The ANCI positioning recommendation section is visible at the bottom

⇉

Workshop Complete!

Your Type 4 Parallel Agent pipeline ran 3 competitor analyses simultaneously.

Parallel Agents

JSON Files

Dashboard

Minutes

Open Your Dashboard

In your terminal: open anci_competitive_dashboard.html (Mac) or start anci_competitive_dashboard.html (Windows)
Review the parallel execution time callout at the top — compare the green bar (parallel) vs the red bar (sequential equivalent)
Check the comparison matrix — the ANCI Opportunity column is your strategic whitespace

What You Can Build Next

➕

Add a 4th Competitor

Add Clockwise or Doodle to the fan-out — the merge and dashboard agents don't change at all.

📅

Schedule Weekly

Add a cron job to run the full pipeline every Monday. Get fresh competitive intel without manual effort.

💬

Post to Slack

Add a 4th stage: a Slack agent that posts the top 3 signals to your team channel automatically.

🔗

Chain to Type 7

Feed the competitive signals into an Orchestrator that decides which ANCI sub-teams need to act on each signal.

Explore the Other AI Agent Types

📘 Type 3 — Sequential Pipeline (you built this)

Chained specialised agents, each passing output to the next. cal.com competitive intel workshop.

⇉ Type 4 — Parallel Execution (this workshop)

Multiple agents running simultaneously. Results merged at the end. 3 competitors in the time it takes to research 1.

◈ Type 7 — Dynamic Agent (next workshop)

Main agent reads the situation and decides at runtime which sub-agents to spawn. See Type 7 workshop →

Parallel AI Agent:3 Competitors, 1 Dashboard