Call · 30 min
MarketingMattia Esposito10 November 202613 min read

B2B lead generation in 2026. First volume, then specificity, then the sentence only you could have written.

More contacts don’t bring more customers, because a B2B buyer now asks AI and colleagues before answering a supplier. Here is the Excel sheet in three steps, with a filled-in example.

In brief

B2B lead generation is the work of finding companies with a problem you can solve and opening a conversation with them. In 2026 it works in three steps, in this order: volume, specificity, personalisation.

The buyer arrives informed. In Forrester’s survey of almost 18,000 B2B buyers, 94% say they use AI during the purchase, and a typical decision involves 13 internal people and nine external ones.

Data is used within precise rules. LinkedIn’s terms ban scraping profiles. The Garante wrote on 17 April 2026 that data published online doesn’t become usable for promotion just because it’s public.

The three-step Excel sheet can be downloaded below, with an example filled in for a made-up company and the PDF, without leaving your email.

Written by: Mattia Esposito, AI and marketing consultant, founder of Itria AI. From Puglia, based in Turin since 2023, he works remotely in Italy and abroad. LinkedIn profile.

The three-step sheet, to download

An Excel sheet with one row per company and five deciding columns: the source of the name, the public signal, the specificity criterion, what to personalise and the legal basis for contact. Each step’s counts calculate themselves. The second sheet is an example filled in for a made-up company, and the same example is in the PDF.

The three-step B2B lead generation Excel sheet, with the example filled in for a made-up company
FileContentsLink
Excelto fill in, in Italian

The blank 60-row sheet with automatic counts, and the sheet with the filled-in example.

lead-generation-b2b-foglio.xlsx

PDFthe example, in Italian

The filled-in example, to read before filling in the sheet.

lead-generation-b2b-esempio.pdf

ColumnWhat you enterStep
Company and website

The name and the website address, one row per company.

Volume

Source of the name

Exhibitor list, business register, saved search.

Volume

Public signal

The fact the company has made public: a job ad, a post, a page.

Specificity

Where and when

The link and the date it was read, because a signal ages.

Specificity

Does it pass the criterion?

Yes or no, according to the sentence written at the top of the sheet.

Specificity

What to personalise

The first line of the message, built on the signal.

Personalisation

Channel

Where you write: email, LinkedIn message, reply to a request of theirs.

Personalisation

Legal basis

Consent, their request, customer, or none: then no promotional email.

Personalisation

Status

To write, written, replied, conversation booked.

Conversation

In September a founder asked, in a post, whether gathering contacts from social media for his software could be automated. I told him it couldn’t be done in bulk: almost every social network’s API is closed. Something similar, though, could be done, and I tried it on his case.

I ran 22 searches and gathered 389 profiles. On 15 September I put it like this: «by profile 151 I’d probably have stopped reading them clearly». Of the 389, 30 were left, to open and read one by one. On three the problem was written in public: one owner asked customers to write again because he’d lost messages, two were taking bookings by hand.

That work gives this page its order: first volume, then specificity, then personalisation. It’s also the order of how far each step lets itself be automated, from most to least. The full story is in the post of 15 September.

The measured cases, from my posts

389, 30, 3. Twenty-two searches, 389 profiles gathered, 30 left after the filter, 3 with the problem written in public. September 2026, in the post of 15 September.

142 sites, 61 with no door. Reading the websites of 142 Italian companies that also sell abroad, 61 had no real way for a foreign buyer to get in touch. It’s a public signal, visible to anyone opening the site: the post of 7 September.

19 out of 121. In the study of food producers’ websites, 19 producers out of 121 offered the product spec sheet. For someone who writes spec sheets, the other 102 companies have the signal on the page.

What B2B lead generation is, and what it means

B2B lead generation is the work through which a company finds other companies that might buy from it and opens a conversation with them. A lead is a contact who has given a signal: asked for a quote, downloaded a template, written in public about a problem you can solve. The definition is ours, written for this page.

There are two routes. In the inbound one, the buyer looks for you and leaves contact details. That’s what a lead magnet is for. In the outbound one, you look for the companies and write first. This page is mostly about the second.

How a company buys today

A company buying today researches on its own, asks AI and decides as a group. Forrester, in the report «The State Of Business Buying, 2026» presented on 21 January 2026, gives its survey of almost 18,000 B2B buyers worldwide: «Nearly all business buyers (94%) report using AI during their buying process.»

The figure needs two caveats. It refers to AI in general, and doesn’t say how many use an assistant like ChatGPT. The full report is reserved for Forrester’s clients, so the numbers here come from the public blog and press release. In the same release, a typical purchase decision involves 13 people inside the company and nine external influencers.

So the first message reaches a person who will discuss it with a group, and someone will ask an assistant who you are. Forrester also writes that buyers check AI against people they trust: colleagues, experts, analysts. A true fact about the company survives that check; a template sentence doesn’t.

The three steps: volume, specificity, personalisation

The three steps narrow the list: volume gathers many companies with broad criteria, specificity keeps only those with a public signal of the problem, personalisation writes to each one starting from that signal. The order matters. Across 389 names attention fades before the end; across 30 it stays sharp.

The three steps of B2B lead generation: 389 profiles in volume, 30 after specificity, 3 personalised, then the conversation Volume389profiles from 22 searches Specificity30to read one by one Personalisation3 with the problem in public Conversation largely automatedpartly automated written in personthe person replies
From volume to conversation, with the numbers from the September 2026 case: at each step the list narrows and human reading grows.

Each step can be automated less than the one before. Gathering is done with filters and saved searches. Checking the signal means opening the page and reading it. The first line needs a person who has understood what they read.

Volume: where the names come from

Volume is the wide list, built with criteria you can filter: sector, size, area, role. Typical sources are the business register, trade fair exhibitor lists, trade associations and LinkedIn searches. Each source goes in its own column, row by row.

Sales Navigator, LinkedIn’s paid tool for salespeople, filters by headcount, company type, seniority and past roles. LinkedIn’s help sets the limits: up to 1,000 leads per list and 10,000 in total. These numbers change.

Volume keeps you from choosing by eye, and the next step removes almost all of it. The source column also has a legal use: on 17 April 2026 the Garante fined a company €1,000 for breaching the right of access of someone asking what it held on him.

Specificity: the signal that the problem exists

Specificity is the filter that keeps only the companies where the problem you solve can be seen from outside. The criterion is written in one sentence before starting, at the top of the sheet, and applied to every row with a yes or a no. In the September case, after the filter, 7.7% of the 389 profiles were left.

SignalWhat it can tell youWhere to read it
Job ad

A task done by hand or a new role, for example a maintenance plan kept in Excel.

«Work with us» page, job boards

Owner’s post

A problem told in the first person, in their own words.

LinkedIn, Instagram, Facebook

New site, line or market

A change that brings new work and new suppliers.

News, press releases, website

Trade fair

A market the company is pursuing in those months.

Exhibitor list

Website with no clear contact

Requests getting lost: 61 sites out of 142 in my September reading.

The website, read as a customer would

Missing document

A spec sheet, price list or catalogue missing or out of date: in the study of producers’ websites, few offer a downloadable spec sheet.

The site

Repeated reviews

The same complaint written by different customers.

Google, sector platforms

The signal is a fact about the company. Using the personal data of the person who wrote it to send them a promotion is another matter, with the rules explained further down. The «Where and when» column keeps link and date: a job ad from a year ago is no longer a signal.

Personalisation: the first line comes from the signal

Personalising means writing the first line from a fact the company has made public, and that the reader recognises as their own. The name in the greeting isn’t enough, and neither is the sector. The line holds up when the reader thinks: this person actually looked.

A first line built on the signal (from the sheet’s example)

I saw you’re looking for a maintenance technician to keep the maintenance plan in Excel. I work on that plan for workshops like yours: shall I send you a page with the three questions I usually ask?

The structure is short: the fact, what you draw from it, a small request. If the channel is email, the consent rules come first, gathered on the page on B2B cold email in Italy, and then the technical ones to stay out of spam, in the guide to SPF, DKIM and DMARC.

How to do lead generation on LinkedIn

On LinkedIn lead generation is done with the profile, with searches and with messages written by a person, within the platform’s rules. The profile is the first page a buyer opens after a message, and it decides whether they reply. How to build it is in the guide to personal branding on LinkedIn.

LinkedIn’s User Agreement, in section 8.2, bans using software, crawlers, plug-ins or bots to scrape or copy profiles and data from the service. The help page «Prohibited software and extensions» adds that LinkedIn doesn’t allow third-party software, including browser extensions, that scrapes or automates activity on the site.

What remains is searching with filters, saving lists, reading and writing one message at a time. In its decision of 17 April 2026 the Garante describes Sales Navigator like this: search with filters and single InMail messages, and no contact information without the member’s consent.

What you can do with the data: Italy, the EU, the UK, the US

In Italy a name found on LinkedIn doesn’t become an address to use for promotion. The Garante, in its decision of 17 April 2026, wrote that data published online «cannot be freely used for promotional purposes merely because it has been made public». An IT consultancy had sent commercial emails to a lawyer, with addresses found through Sales Navigator.

The outcome: processing declared unlawful and a reprimand for the emails sent without consent. On 14 May 2026 a second decision closed other routes: accepting a LinkedIn connection isn’t consent, consistency with the recipient’s job doesn’t count, and a data supplier’s mistake doesn’t remove the sender’s fault. Fine: €4,000.

The two amounts concern small businesses with mitigating factors, and aren’t the rule. In the United States the well-known case is hiQ v. LinkedIn. On 18 April 2022 the Ninth Circuit Court of Appeals upheld a preliminary injunction in hiQ’s favour, on the federal computer fraud law and public profiles.

The contract, however, remained. According to summaries published by law firms, in December 2022 the parties settled with a consent judgment: hiQ acknowledged breaching the terms of use, stopped collecting data and destroyed what it had collected.

TopicItalyEuropean UnionUnited KingdomUnited States
Data from public profiles

Not usable for promotion just because it’s public (Garante, 17 April 2026).

A legal basis is required (GDPR).

UK GDPR applies.

Access debated in the hiQ case (2022).

Promotional email to a person

Consent required (Privacy Code, Art. 130).

Prior consent (ePrivacy Directive, Art. 13).

Consent, except the soft opt-in for customers (PECR, reg. 22).

No prior consent: opt-out (CAN-SPAM).

Email to a company

Consent for legal persons too, as «contracting parties» (Arts 121 and 130).

Each state decides how to protect them (Art. 13(5)).

Reg. 22 doesn’t cover «corporate subscribers»; it does cover sole traders.

No separate category for companies in § 7704.

Legitimate interest for email

Ruled out by the Garante (14 May 2026).

Allowed for marketing in general (Recital 47); for email ePrivacy prevails.

Not for individuals.

Not needed: the system is opt-out.

Obligations in every message

Identifiable sender and a contact for objecting (Art. 130(5)).

Identity and a valid address for objecting (Art. 13(4)).

Identity and an opt-out address (PECR, reg. 23).

Postal address and opt-out honoured within 10 business days.

LinkedIn’s terms

Scraping banned (§8.2).

Scraping banned (§8.2).

Scraping banned (§8.2).

Scraping banned (§8.2).

B2B lead generation with AI

AI shortens the two steps that take the most time, gathering names and reading signals, and is less useful for the sentence that opens the conversation. On 5 October I wrote it in a post: for gathering signals and speeding up the collection of contacts, yes; for having it write the messages, at the start, I don’t recommend it.

The reason is in Forrester’s data too. The buyer uses AI and then checks with trusted people: a message that sounds generated gets spotted, and sometimes ends up shown in public as an example not to follow. The time AI frees up on the list is best spent on the thirty rows that remain.

The data rules don’t change when a machine does the work. In the decision of 14 May 2026 the address had been found with an app that enriches profiles, and the Garante wrote that a good-faith error caused by a supplier doesn’t rule out fault.

B2B lead generation examples

The sheet’s example is a made-up company, Calibra, selling maintenance planning software to engineering workshops with 20 to 200 employees in Emilia-Romagna. The specificity criterion written at the top: the company shows in public that it manages maintenance by hand or is looking for someone to do it.

Volume comes from a trade fair’s exhibitor list and a saved search in Sales Navigator: ten rows in the example sheet. Four pass the criterion. For one there’s consent, because it had signed up to the newsletter; another had asked for a quote. For the two with no legal basis the channel is a LinkedIn message, written by hand.

The same shape works for other trades. A web designer looks for companies with no clear contact; a spec sheet writer looks for companies that don’t publish them; a packaging supplier follows companies opening a new line. The signal changes, the sequence stays.

B2B customer acquisition: where lead generation ends

Lead generation ends when a conversation is booked; from there, sales begins. To find B2B customers what counts is how many conversations come from every hundred rows of the sheet, more than how many names go in. The «Status» column is there to count them; when the rows grow, the sheet moves into a CRM, with each contact’s history.

If you want to do it in-house, you have the sheet. If you want to hand it over, the work is on the page on contact research and qualification, and the systems that keep the flow going are on the page on B2B acquisition engines. Where AI shortens the work, and the rules it does not change, is in the guide on using AI to find customers.

Questions and answers

What is B2B lead generation?

The work through which a company finds other companies that might buy from it and opens a conversation with them. A lead is a contact who has given a signal, for example a request for a quote.

How do you do B2B lead generation in 2026?

In three steps: volume, a wide list built with filterable criteria; specificity, which keeps only the companies with a public signal of the problem; personalisation, which writes to each one starting from that signal. Each step can be automated less than the one before.

Can you use contacts found on LinkedIn to send emails?

In Italy, promotional emails require consent. The Garante, in its decision of 17 April 2026 on addresses found with Sales Navigator, wrote that data published online can’t be used for promotion just because it’s public. On 14 May 2026 it added that accepting a LinkedIn connection isn’t consent.

Is scraping LinkedIn legal?

LinkedIn’s terms, in section 8.2, ban using software, crawlers, plug-ins or bots to copy profiles and data. In the United States the 2022 hiQ ruling concerned access to public pages under the computer fraud law, and the contract with LinkedIn remained in force.

How is AI used in B2B lead generation?

To shorten the gathering of names and the reading of signals. The first line of the message is best written in person: according to Forrester, buyers check AI against people they trust.

Notes on sources

  1. Forrester, The State Of Business Buying, 2026 (blog) and press release, 21 January 2026: Buyers’ Journey Survey of almost 18,000 B2B buyers worldwide; source of the quotation, the 94% and the 13 plus nine. A research firm; the full report is reserved for clients. Read on 10 November 2026.
  2. LinkedIn, User Agreement, in force from 3 November 2025, §8.2; Prohibited software and extensions; Sales Navigator help on saving leads and filters. Read on 10 November 2026.
  3. Italian data protection authority, decision of 17 April 2026, doc. web 10252188, source of the quotation in our translation, and decision no. 349 of 14 May 2026, doc. web 10262105. Read on 10 November 2026.
  4. Privacy Code, Art. 130 and Art. 121 (Normattiva, text in force); Directive 2002/58/EC, Art. 13, consolidated text; GDPR, Art. 6 and Recital 47. Read on 10 November 2026.
  5. United Kingdom, PECR 2003, reg. 22 and reg. 23 (updated text on legislation.gov.uk). United States, 15 U.S.C. § 7704 (CAN-SPAM). Read on 10 November 2026.
  6. United States Court of Appeals for the Ninth Circuit, hiQ Labs v. LinkedIn, No. 17-16783, 18 April 2022. The December 2022 settlement is reported by law firm summaries, not by the court’s text.
  7. Mattia Esposito’s figures come from his LinkedIn posts of 15 September 2026 and 5 October 2026, and from the study of food producers’ websites. The 7.7% figure is calculated here: 30 out of 389.
·Who does it for you

A list of names takes an afternoon to find. The right companies are found by reading. Positioning is built starting from the market.

Mattia Esposito, who wrote this guide, also works with Itria on positioning: in-depth market analysis to find where competitors are exposed and take that ground. He did it on his own site first. In August, asked who had founded itria.io, ChatGPT couldn’t find him. From 11 October to 7 November 2026 itria.io pages appeared 13,803 times in Google and Bing results and 3,818 times in AI answers: Google’s AI Overviews and AI Mode, and Microsoft Copilot.