The software company that created FT Mercati Go to ftmercati.com

Case study

An AI agent that prepares trade fair visits

The CRM's Trade Fairs module: from any online exhibitor list to the visit, with an AI agent that collects, screens and matches against customer records.

In production Sales cycle: preparing for and visiting industry trade fairs Updated on

  • 3ways to load the exhibitors: paste, AI reading, scout agent from the fair's website
  • 3keys for matching to the customer records: VAT number, domain, name, with a confidence level by source
  • 1monthly cap on paid web searches, with a setting that switches them off
  • 8seconds at most for the VIES check of a VAT number on saving
  • 0internal addresses reachable: every page is downloaded by a client that blocks them

In brief

The Trade Fairs module of FT Mercati's CRM starts from the exhibitor list any fair publishes online: an AI agent collects it, screens it for relevance to the company's services, checks websites and VAT numbers within a monthly cap and matches each exhibitor to the customer records. Then priorities, tablet visits with an offline queue, AI summaries of the matched customer's CRM notes. It is part of the CRM and also available as a standalone tool.

Screenshots

The preview of the exhibitor list as read by the AI, with a priority and notes for each row
Import with AI reading: the pasted list is turned into rows with hall, stand, company, sector and priority, to be reviewed before importing.
The exhibitor list of the sample fair, with filters by priority and matching status
The fair's list: priorities, multi-select filters, search, and for each exhibitor its status in relation to the customer records.
The exhibitors matched to the CRM's customer records, with the uncertain cases to be checked
Match to CRM (Abbina a CRM): the certain matches are linked, the uncertain ones stay "to be checked" and the new companies become potential customers.
The AI summary of the CRM notes of an exhibitor that is already a customer, generated before the meeting
Before the meeting: for an exhibitor that is already a customer, the AI sums up in a hundred words the history of the sales notes in the CRM.

Screenshots from a demonstration environment with invented data: the companies, people and amounts are not real.

The starting point

Anyone who sells services to an industry goes to that industry’s trade fairs. The work starts beforehand: downloading the exhibitor list, working out who the exhibitors are, checking whether they are already customers or have been in the past, deciding whom to visit and in which hall. Done by hand, on hundreds of names, it takes days and is done badly: incomplete names, unverified websites, duplicates in the customer records.

What we built

A complete “Trade Fairs” module inside the CRM with AI agents. It works with any fair that publishes its exhibitor list online, and has three ways to load the exhibitors: pasting the list, having artificial intelligence (AI) read it, or letting the scout agent extract it from the fair’s website.

The scout agent starts from the address of the website, browses pages, embedded frames and data endpoints, uses web search and, for sites built in JavaScript, a text extraction service. It collects the list of exhibitors, with the hall and stand of each one, and screens them for relevance to FT Mercati’s services: interest, sector, the company’s role in the metal supply chain (producer, user, trader or recycler, seller of machinery and services), talking point for the visit.

A second step, “Websites and VAT number”, checks the exhibitors’ websites, which is free, and searches the web for the VAT number and a description of the company, which costs money and therefore has a monthly cap on searches and a setting that switches them off. Then comes the matching to the CRM’s customer records: by VAT number, by website domain, by name, with a confidence level that depends on the source of the data.

At the fair the work is done on a tablet, with a queue that withstands the lack of network coverage in the halls; before the meeting, the AI sums up in a hundred words the history of the matched customer’s sales notes; the notes taken at the stand become CRM notes, without duplicates; new potential customers become customer records, and the VAT data of an EU company is completed on saving with the VIES check.

A module of the CRM, and a standalone tool too

Inside the CRM the module shares everything: the customer records, the VAT rules and the VIES check, the salespeople and their roles, the assistant. A company that already has another ERP (the company’s management software) can have the module as a standalone tool: the same collection, the same screening and the same matching, connected to its own customer records. The module’s functions, in a list:

  1. loading the exhibitors in three ways: paste, AI reading, scout agent from the fair’s website;
  2. hall and stand of each exhibitor, with the way the fair is organised taken from the website;
  3. screening for relevance with written criteria: interest, sector, role in the supply chain, talking point for the visit;
  4. checking of websites and search for VAT number and description, with a monthly cap;
  5. matching to the customer records by VAT number, domain and name, with a confidence level and the uncertain cases marked “to be checked”;
  6. priorities with a multi-select filter and bulk deletion;
  7. visits from a tablet with an offline queue;
  8. AI summaries of the CRM note history of each matched customer, to prepare for the meeting; the notes taken at the stand go into the CRM as sales notes;
  9. new potential customers that become customer records, with the VAT data completed through VIES.

How we keep the agent under control

  1. Scope and caps. For each fair, one collection runs at a time, and a run already in progress is not duplicated; paid searches have a monthly cap and a setting that switches them off.
  2. Security when working on the web. Every page is downloaded by a client that cannot reach internal network addresses; the regular expressions that parse the texts are written so that they do not hang on hostile input.
  3. Matching never merges customer records. On its own it links only the certain matches, by VAT number or by verified domain; the uncertain ones, starting with those by name, stay “to be checked” with an explanation of why, and the salesperson confirms or rejects.
  4. The person chooses whom to visit. A multi-select filter by priority, bulk deletion of the exhibitors that are not needed.
  5. The rules of the customer records apply here too. A potential customer in the EU starts out without VAT data and receives it when the record is saved, with the VIES check within eight seconds.

Declared limits

  • in rare cases matching and the synchronisation of visits overlap, and the “at this fair” notice may be inaccurate;
  • a CRM name contained in a longer name may be treated as the same company: that is why matches by name need to be reviewed;
  • the language of the catalogue collected, Italian or English, depends on the path the agent took through the website.

What it means for your company

It is the first agent in our CRM that steps outside the management software and goes onto the web, and it serves as a model: a precise task, a set scope, spending caps, checks on the data found, a person who decides. The same pattern is useful for qualifying suppliers, for updating company records from websites, for preparing a round of visits; and the Trade Fairs module can also be had on its own, connected to the customer records the company already uses. We discuss it in the guide AI agents: what they are and what they can do and in the one on CRM with AI. If your sales team prepares for trade fairs with copy and paste, let’s talk.

Frequently asked questions

Trade Fairs module with AI agent: questions and answers

What exactly does the scout agent do for a trade fair?

It is given the address of the fair's website and explores it as a patient person would: pages, embedded frames, hidden data endpoints, web search, text extraction from pages built in JavaScript. From this it derives the list of exhibitors, with the hall and stand of each one, and for each one it proposes a relevance rating.

How does it know which exhibitors are of interest?

It assesses them against FT Mercati's services with a criterion written into the prompt: interest, sector, the company's role in the metal supply chain (producer, user, trader or recycler, seller of machinery and services) and a talking point for the visit. It is a proposal: the salesperson can change priorities, delete in bulk the exhibitors that are not needed and keep only the ones to visit.

What happens to the exhibitors' data?

They are data about businesses, taken from public sources: the fair's website and the companies' websites. They stay in the CRM, matched to existing customer records or recorded as potential customers; paid searches have a monthly cap and a setting that switches them off, and every page is downloaded by a client that cannot reach internal addresses.

Can the same agent be built for another company?

Yes. The relevance criteria and the ERP it matches against change; the method stays (collection, screening, matching with a confidence level, verification), and so do the limits: spending caps, a log, a person's confirmation on the uncertain cases.

Other case studies

Do you have a process you would like to see working better?

Tell us about it in a first conversation, with no obligation: together we work out whether and where AI brings a measurable benefit, and which step to start with.