Multi-Agent Market Research System
Smart Money Nigeria's manual research and reporting process was a bottleneck. Producing a single financial intelligence report took 4 to 6 hours, was inconsistent in quality, had gaps in coverage, and left no audit trail. I designed and built an end-to-end AI-powered pipeline that runs four specialist agents in parallel, synthesises the outputs into a structured report, and delivers it across three channels automatically every morning, in under 5 minutes.
< 5 min
Full research cycle, end-to-end
4
Specialist AI agents running in parallel
100%
Coverage across Trends, Competitors, Sentiment, Macro
0
Marginal cost per additional report run
The Problem
Smart Money Nigeria serves a professional investor community that depends on timely, accurate market intelligence. Their existing process was entirely manual: a researcher would spend several hours searching, reading, and writing a report before it reached the audience. The approach did not scale, coverage was inconsistent, and there was no way to verify what had been analysed or when.
- ›4 to 6 hours per report with linear scaling cost as the audience grew
- ›Inconsistent quality and coverage gaps across research dimensions
- ›No audit trail for what data was sourced or when each report ran
- ›Manual process blocked same-day delivery of time-sensitive market intelligence
The 4 Specialist AI Agents
Trend Agent
Google News via Serper, last 30 days
Top 3 market trends, growth statistics, risk flags
Competitor Agent
Serper web search, last 30 days
Key players, strategic moves, new entrants, moats and weaknesses
Sentiment Agent
Serper News API, last 7 days
BULLISH/NEUTRAL/BEARISH verdict with a 0-100 confidence score and narrative summary
Macro Agent
Serper web search, last 30 days
CBN/regulatory updates, economic indicators (MPR, CPI, FX), investment climate score 0-10
5-Layer System Architecture
Layer 1: Trigger
7AM daily schedule (Africa/Lagos timezone) initiates a new research run
Layer 2: Orchestrator
Claude Sonnet decomposes the research topic into 4 non-overlapping, targeted search queries
Layer 3: Parallel Agents
4 Serper searches fire simultaneously; 4 Gemini analyst agents process results in parallel, each returning structured JSON with a confidence score
Layer 4: Synthesis Engine
n8n Merge node waits for all 4 agents; a Code node maps outputs by agent field with fallbacks; Claude Sonnet writes the final dark-theme HTML report targeting 4,000+ characters
Layer 5: Quality Gate
Report length checked against a 2,000-character minimum. Failures trigger an admin Telegram alert and an Airtable error log entry
Layer 6: Delivery
Full HTML report sent via Gmail, a 1,000-character Markdown summary sent via Telegram, and the full report archived in Airtable per run
Workflow and Output
n8n Workflow: Full Pipeline Architecture
Report Output: Nigeria Financial Intelligence Report
Telegram Delivery: HTML Reports Sent to Subscribers
Step-by-Step Orchestration (10 Nodes)
Step 1
Trigger fires at 7AM Africa/Lagos on a daily schedule
Step 2
Set Defaults node assigns run_id, default topic, and report_date for idempotency
Step 3
Orchestrator node calls Claude Sonnet to generate 4 non-overlapping search queries
Step 4
Parse and Validate node strips markdown, parses JSON, and throws on malformed output
Step 5
Fan Out: 4 Serper API calls fire simultaneously across Trend (30d), Competitors (30d), Sentiment News (7d), and Macro (30d)
Step 6
Parallel Analysis: 4 Gemini agents process their respective search results simultaneously
Step 7
Merge and Combine: n8n Merge node waits for all 4 agents; Code node assembles the combined payload
Step 8
Report Writer: Claude Sonnet receives all 4 structured JSONs and writes the full dark-theme HTML report
Step 9
Quality Gate: report length checked; failures routed to admin Telegram alert and Airtable error log
Step 10
Parallel Delivery: Gmail (full HTML), Telegram (1,000-char summary), Airtable (archive per run)
Report Structure
Each report is a dark-theme HTML document targeting 4,000+ characters, delivered as a full email and archived per run. Every section is generated from the 4 agent JSON outputs by Claude Sonnet, producing consistent, structured intelligence every time.
Technical Stack
n8n
Workflow orchestration, parallel agent fan-out, merge, quality gate, and multi-channel delivery
Claude Sonnet
Query decomposition (Orchestrator) and final report synthesis (Report Writer)
Gemini
4 specialist analysis agents processing structured search results in parallel
Serper API
Google News and web search across 4 research dimensions
Gmail
Full HTML report delivery per run
Telegram
Real-time 1,000-character Markdown summary and error alerting
Airtable
Per-run archive of HTML reports, agent scores, and error logs