Your Network Is Already in Your Inbox

Download your contacts from Gmail, Yahoo, Apple, Facebook and LinkedIn, let an AI assistant merge them into one database, keep it current, and reach out for your business the way Canadian law allows

A small-business guide to turning your Gmail, Yahoo, iCloud, Facebook and LinkedIn data into one contacts database with an AI assistant: exporting, merging, enriching from your own email history, keeping it current, and using it for outreach within CASL and PIPEDA. Tested on fictional data.
Author

Behzad Samadi

Published

October 31, 2026

A paper-craft desk: envelopes and paper cards of different colours drift into a wooden card-index box of tabbed cards; above it, round paper portrait tokens joined by threads, most with an amber tick and one with a stop sign; a fountain pen beside the box. Title: Your Network Is Already in Your Inbox.

Ask a small-business owner where their next customer will come from, and the honest answer is usually: someone they already know. A past client, a friend of a client, the architect from the last job, the property manager who calls when something leaks. Those people are already in your life, and in your data. They’re just scattered across a Gmail account, an old Yahoo address, an iPhone, Facebook and LinkedIn, each with half the picture.

This article shows how to bring them into one place with an AI assistant: download your data from each service, merge it into a single contacts database, enrich it from what you already know, keep it current, and use it to reach out. We tested every step on a fictional business, and the assistant did the hard parts well. The part it can’t do for you is decide what’s allowed: in Canada, a LinkedIn connection is not permission to send a sales email, and that shapes everything after the database is built.

NoteIn short
  1. Every major service lets you download your own data. Contacts come as CSV or vCard files, LinkedIn connections as a spreadsheet, Gmail as a mail archive. It takes minutes to days.
  2. An AI assistant can merge them well. In our test it turned 71 records from five exports into 25 people, merged a contact listed under three different names, and flagged the two matches it wasn’t sure of instead of guessing.
  3. The best enrichment is your own history: when you last emailed someone, whether they bought, asked for a quote, or asked you to stop. Avoid tools that scrape LinkedIn or Facebook; the platforms forbid it, and Canada’s privacy commissioner has warned about it.
  4. Your network isn’t your mailing list. Under Canada’s anti-spam law (CASL), you need consent for commercial email, and a connection or friendship doesn’t count. A purchase gives you two years; a quote request, six months.
  5. Maintained, it compounds. Re-run the assistant when you download new exports; in our test it added a new person, picked up a job change, settled a merge, renewed one person’s consent and kept every note the owner had added.
WarningThis is a practical guide, not legal advice

The rules below are summarised from the statutes and the federal regulators’ guidance. For a large list, a regulated industry or anything unusual, check with a lawyer.

Step 1: Download your data

Each service has its own way, and they change their menus from time to time. What their help pages said on 31 October 2026:

Service Where What you get for contacts Notes
Google Google Contacts → Export; or Google Takeout for everything Contacts as CSV or vCard; Gmail as an archive file Takeout is ready in “a few minutes to a few days”; the download link lasts about a week. Gmail also saves people you’ve emailed as “Other contacts”.
Yahoo Yahoo Mail → Contacts → Export A CSV file Up to 5,000 contacts. Yahoo’s help says accented letters and dashes can make the export fail, so check French names. Full account data comes from Yahoo’s privacy dashboard.
Apple iCloud.com → Contacts → Export vCard; or privacy.apple.com One vCard with all selected contacts The privacy download gives contacts as vCard and mail as individual messages; you have 14 days to collect it.
Facebook Accounts Center → Your information → Download your information Your friends list, in HTML or JSON Meta’s help pages don’t say whether friends’ email addresses or phone numbers are included; plan on names and the date you became friends. The file stays available for 4 days.
LinkedIn Settings → Data privacy → Get a copy of your data A connections spreadsheet: name, profile link, company, position, date connected, and email Email appears only if that connection allows it in their own settings, so most are blank. A single category arrives in minutes; the full archive within 24 hours.

Download them to your own computer, into one folder, and keep the originals untouched: they’re your record of where each fact came from.

Step 2: One database, built by an assistant

We built a test the same way as in our earlier articles: a fictional business, realistic data, and an AI assistant (Claude Code) working from a plain request, with no one at the keyboard. The business is a small renovation company in Toronto; the five exports hold 25 invented people, written in the services’ formats, with the mess real exports have: the same person as “Ben” in one place and “Benjamin” in another, an email address typed in capitals, a job change only LinkedIn knows about, a Facebook friend listed under a nickname, and a Gmail history in which some people bought, some asked for quotes, and one asked to be removed.

The request, in short: merge everything into one database, one row per real person, keep track of where every field came from, flag uncertain merges instead of guessing, add what can be learned from the email history, work out who can be emailed under Canadian law and until when, and write a script that can be re-run later without losing my notes. Work only on this computer; send nothing.

Result
Time and cost 9 minutes, about C$3.60 of AI time
Records in 71, from five exports
People out 27, which became 25 once its two flagged questions were answered
Hardest merge “Ben Tremblay” in Google, an upper-case address in Yahoo, and “Ben” on LinkedIn: one person, merged correctly
Flagged, not guessed “Ella Mac” on Facebook (probably Ella MacDonald, but nothing else links them), and a LinkedIn “Grace Li” with a new employer
Source for job titles LinkedIn first, because people keep it current; older sources kept as history

It stored everything in a single SQLite file, a database that is just a file on your computer and that any spreadsheet tool or AI assistant can read. Every field remembers its source, so “why does it say Operations manager?” has an answer: LinkedIn, last month.

This is the same idea as our article on keeping one notebook that holds the truth about your business: one place, sources attached, a person in charge of the judgement calls.

Step 3: Enrich it from what you already know

“Enrichment” usually means buying data: services that find a person’s work email, title or phone number from the web. Be careful with them. The platforms forbid it: LinkedIn’s user agreement bans scraping and bots that download contacts, and Meta’s terms bar automated collection. And Canada’s privacy commissioner, with other regulators, has warned that personal information on public websites is still protected by privacy law, that scraped contact details feed spam, and that this applies to small businesses too.

The richer source is the one you already have: your own history. From the Gmail archive alone, our assistant added to each person:

  • When you last wrote to them, and when they last wrote to you, and how often you’re in touch.
  • Business events: an invoice, a final payment, a quote request, found by reading the subject lines and text.
  • Withdrawals: one person had asked not to receive promotional email, and the assistant put them on a do-not-email list.
  • Current role, from the newest source that has one.

It also noticed something we’d got wrong in the fictional data: one invoice was numbered “2024-117” but dated 2025. It didn’t pick one; it asked the owner to check the books, because the date decides how long the customer can be emailed.

Step 4: Know who you can email

Canada’s Anti-Spam Legislation (CASL) applies to any commercial electronic message: email, text, and messages on social platforms, LinkedIn and Facebook messages included. You need the recipient’s consent, and you have to be able to prove it. The essentials:

  • Express consent: they agreed to receive your messages (a sign-up form, a ticked box, a signed paper form). It lasts until they withdraw it.
  • Implied consent from a purchase or contract: two years from the transaction.
  • Implied consent from an inquiry or quote request: six months.
  • A published address (on their business website, say, with no “no solicitations” note) allows messages relevant to their business role, not general promotions.
  • A referral allows one first message to the person you were referred to, naming who referred you.
  • Unsubscribes must take effect within 10 business days.
  • A connection or a friendship is not consent. The personal-relationship exemption is for one individual writing to another with whom they’ve had real two-way contact; it isn’t a basis for a business mailing.
  • Asking for consent by email is itself a commercial message. Ask in person, by phone or on paper.

Here’s how the assistant sorted the 25 fictional contacts on 31 October 2026, checked against the answer key we’d written before the run:

Status Who (fictional) Why
Can email Ben, until Oct 2028 Paid invoice, two weeks earlier
Can email Katherine, until Jun 2027 Paid for a kitchen, about 17 months earlier
Can email Olivia, until Mar 2027 Asked for a quote 40 days earlier
Consent expired Victor, Maya, Quinn Quote requests over six months ago; a purchase over two years ago
Do not email Uma Asked to stop; to be honoured by 3 November
Your call Sofia; Dev and Farid A personal friend; two people met at a trade show (a business card implies consent only for messages relevant to their job)
No consent found 14 people, including most LinkedIn and Facebook contacts Knowing them isn’t permission

Every entry was right. And it’s the honest picture of most networks: a handful of people you can email today, many you know but can’t send promotions to, and a few who’ve said no.

Step 5: Keep it current

A contacts database is only useful if it stays true. We tested the upkeep too: we added a note to one contact and a signed newsletter sign-up for another, as an owner would, then dropped in a fresh LinkedIn export and a new email, and asked a new assistant session, with no memory of the first, to update the database and say what changed.

In three minutes, for about C$2:

  • It backed up the database first.
  • It kept the owner’s note and consent record, and applied the signed form: Farid became emailable, with “keep that paper form” as advice.
  • It settled a flagged merge on its own evidence: the new LinkedIn file showed Grace’s email, matching her other records.
  • It found a new connection and a job change, and kept the old job as history.
  • It renewed Victor’s consent to April 2027, because he’d written that morning asking for an updated quote.
  • It noticed the new LinkedIn file was partial (4 connections, not 14) and didn’t treat the missing 10 as removed.
  • It found and fixed a bug in its own script that could have let an older export overwrite a newer one.

A good rhythm for a small business: download fresh exports every quarter, ask the assistant to update, and read its summary of what changed.

Step 6: Reach out the right way

The database is most valuable for what it tells you to do by hand:

  • The people you can email: past customers and recent inquiries. Send them something useful, from you, with your business name, contact details and an unsubscribe option in every message.
  • The people you know but can’t email for business: call them, see them, comment on their posts, send a personal note that isn’t a pitch. If they want to hear from you, ask in person or on the phone and record the answer and the date in your database.
  • Referral partners (the architect, the realtor, the inspector): your message about their line of work is often allowed if their business address is published for that purpose; check before you send.
  • Follow-ups: our free Follow-up Finder reads a CSV export and lists who is owed a reply and who you can still follow up with under CASL.
  • No automation on LinkedIn or Facebook. Bulk invitations and automated messages break both platforms’ terms, and the messages themselves are covered by CASL.

Look after other people’s data

Your contacts’ details are personal information, and once you use them for your business, Canada’s private-sector privacy law (PIPEDA) applies: keep them accurate, keep them safe, and delete what you no longer need. Prospecting is a new purpose for most of those details, so identify it, and get consent where it’s needed, before you start. In practice:

  • Keep the database on your own computer, encrypted, and backed up.
  • Know where the assistant sends what it reads. Claude Code, like the desktop assistants in our guide to desktop AI, works on files on your computer, but the model runs on its provider’s servers, so every record it reads is sent there. Check your plan’s data settings, give it only the folder it needs, and try things out on fictional or anonymised data first, as we did.
  • Prune it. People who’ve said no stay on the do-not-email list; people you no longer deal with can go.

Try it yourself

The fictional exports, the requests we gave the assistant and our answer key are in collegica-code, so you can rerun the test before you try it on your own data. The rules about consent are in more detail in our small-business AI guide.

Sources