Qualify trade show exhibitors before you go with an n8n AI agent that researches each company and scores it against your client profile
A small AI consultancy scrapes a trade show's exhibitor list, has an n8n agent with GPT-5-mini research each company by web search and scraping, scores it low, medium or high against its ideal client profile, and writes reasons to a sheet.
Evidence: The source shows its work: steps, screenshots, code or data you can inspect.
The business problem
Researching hundreds of trade show exhibitors by hand to decide which booths to visit and who to follow up with takes hours, so most of the list goes unchecked.
What was tried
1. Scrape the exhibitor list from the trade show website with Bright Data and parse out each company's name, description, booth number and website. 2. Send the companies in small batches (to control API cost) to an n8n AI Agent node running GPT-5-mini. 3. Give the agent a system prompt describing your own company, a target client profile (size, public interest in data and AI, a modern outlook, willingness to work with a small team) and instructions: find out what the company does, estimate its size band, and rate it low, medium or high, leaving fields blank rather than guessing. 4. Let the agent use two tools: SerpAPI for web search and Bright Data to scrape a specific company page when needed. 5. Require a JSON answer (name, description, booth, website, score, reasoning) and append each result to a Google Sheet, giving a ranked list of booth visits and follow-ups. The sanitised workflow, including the prompts, is published as a public gist, and the author notes you can reuse it by changing the show URL and the profile.
What was reported (not reported)
Not reported. The post is a challenge entry with a demo video; it gives no count of companies scored, accuracy, meetings or clients won. The author says Bright Data needed no manual intervention while processing hundreds of exhibitors.
Limitations
Written as an entry for a contest run by n8n and Bright Data, and it praises both, so it is partly promotional. Batching was set to a small number for testing, and the agent is stateless. No check of whether the scores matched real buying interest. The author's firm is a small specialist team rather than a solo founder. Bright Data and SerpAPI are paid services; costs are not given.
What you need
n8n, an OpenAI API key, SerpAPI and Bright Data accounts, a Google Sheet, and a written ideal client profile.
Sources
- DEV Community (Andre Moreira, lyfX.ai) ↗ Tutorial, published August 29, 2025, main source
- GitHub Gist (andremoreira73) ↗ Code repository, publication date unknown
Source published: August 29, 2025. Last reviewed here: November 2, 2026. Spot a mistake? Tell us.
Tools in this workflow
- Google Sheets: Stores scores and reasonsTry Google Sheets
- n8n: Workflow and AI Agent nodeTry n8n
- OpenAI GPT-5-mini: Research and scoring modelTry OpenAI GPT-5-mini
- Bright Data: Scrapes the exhibitor list and company pagesTry Bright Data
- SerpAPI: Web search tool for the agentTry SerpAPI
Get new workflows like this by email
A short selection of the most useful new workflows, including the ones that did not work. Free, and you can unsubscribe at any time.
Related workflows
Before fixing a zero-view marketplace shop, measure the base rate: an AI agent pulled 410 competing listings through the public API
An AI agent running a tiny template business on a 200 euro budget had zero sales and almost no views. Instead of rewriting listings again, it sampled 410 competing listings and found that new listings in its niche normally get about three views.
Deliver adaptive micro-courses in Telegram with GPT-generated lessons and Supabase learning state
The builder of SkillDay describes an n8n learning bot that generates courses, quizzes and feedback with GPT-4o-mini, stores learner state in Supabase and adjusts review intervals from test results.
Test Facebook ads on 120 simulated buyers before spending, then check against real results
A small exam-prep client's real ad results were hidden from a synthetic-buyer tool. It ranked the ads almost backwards on clicks, but correctly flagged trust and privacy as the reason few visitors signed up.
Five-part checklist for setting up scheduled AI jobs that you can trust to run unattended
A Build to Launch guide maps the scheduled-task options in Claude, ChatGPT, Perplexity, n8n and cron, and gives a checklist for each job: define a good result, set limits, keep run receipts, and name an owner and off switch.