Home/Portfolio/Competitor Analysis
Project · n8n + AI · GTM Intelligence

Competitor Analysis: Multi-Agent Research

Send a product URL or a short description from Telegram, WhatsApp, Slack or a Google Doc. Nine AI research agents investigate the market, and a sourced competitor analysis plus a designed visual report arrive by email.

Status: built in n8n, not yet run on live data. Self-initiated portfolio project; no client data. We will add sample outputs once it has been run and checked.

The problem

Competitor research for a small team usually means a few tabs, a chatbot summary without sources, and a doc that goes stale. Pricing, positioning, ICP and objections are rarely compared side by side.

How it works, in plain words

You a person on your team   AI an AI step   Auto runs by itself

  1. You

    Send a link

    A product URL or short description, from Slack, Telegram, WhatsApp or a Google Doc.

  2. AI

    It studies the product

    Reads the page and searches the web to understand what you sell.

  3. AI

    It finds the competitors

    Small, mid-size and large players, each with sources.

  4. AI

    Six researchers work at once

    Features, positioning, buyers, pricing, objections and videos.

  5. Auto

    A report lands in your inbox

    A visual report plus the raw research, every claim with its source.

What arrives in your inboxSummary the big pictureCompetitors in three tiersPricing side by sidePlaybook objectionsSources every claimReport plus raw research file

Your part: check pricing and claims before acting on them. The agents cite sources and mark unknowns as “not public”, but AI research can still be wrong.

Without it
  • A few tabs and a chatbot summary with no sources
  • Pricing and positioning never compared side by side
  • Notes that go stale
With it
  • Nine agents research in parallel, six at the same time
  • Every claim has a source; unknowns are marked, not guessed
  • Report and raw research arrive by email
See the real build (n8n)
Workflow diagram: four input triggers feed a product-understanding agent, a competitor-discovery agent, six research agents in parallel, then a compiled report emailed with attachments.
  1. Input from anywhere: Telegram, WhatsApp, a Slack mention, or a new Google Doc in a folder. All are normalised into one input.
  2. Understand the product: an agent reads the page and searches the web to write a product brief.
  3. Discover competitors across three tiers: small/early, mid-market, enterprise/legacy, with sources.
  4. Six agents research in parallel: product and features, positioning and messaging, ICP and personas, public pricing, objection handling and buyer complaints, YouTube videos.
  5. Compile and design: all research is merged with sources, then an agent writes an email-safe visual report.

What you get

  • An email with the visual report: executive summary, tiered competitor cards, feature and positioning tables, persona cards, pricing comparison, objection playbook, video list, recommended actions and sources
  • raw-research.md: every section in full, with sources
  • visual-report.html: the same report as a standalone file

How it's built

PartDetail
Platformn8n, 47 nodes
AI9 agents, each with its own Claude model node
ToolsWeb search, page fetch, YouTube search
InputsTelegram, WhatsApp, Slack, Google Docs
DeliveryGmail with HTML body and two attachments

Reliability

  • Agents must cite sources and mark unknowns; pricing shows "not public" rather than a guess
  • Research waits for all six agents before compiling, so no section is silently missing
  • Public information only; objection handling uses public reviews

Limits and next steps

  • Several minutes and meaningful AI usage per run
  • AI research can still be wrong: verify pricing and claims before acting
  • Planned: chat acknowledgement, per-agent retries, saving to Drive/Notion, scheduled re-runs to track changes

Want this for your market?

Start with a founding GTM report, or have this workflow set up for your team.

Discuss this workflow