Your First Contact Card
Outcome: one person turned into a complete contact record — current role, company, verified work email — from a single prompt, for about 2.3 credits.
- Surface
- MCP server
- Level
- Beginner
- Uses
- linkedin_profile_enrich · find_work_email · verify_email
- Credits
- ~2.3 per contact
- Prerequisite
- Lesson 01's working connection
The chain
Three tools, in order, each feeding the next.
linkedin_profile_enrich
Input: a profile URL. Output: name, current title, company, location, and history.
find_work_email
Input: the name and company domain from step one. Output: a work address.
verify_email
Input: the address. Output: a deliverability status. Only valid counts.
Each step’s output is the next step’s input, which is exactly what makes this worth doing in a chat window rather than across three tabs.
The prompt
Using the SyncGTM MCP server, build a contact card for this person:
https://www.linkedin.com/in/example
Steps:
1. linkedin_profile_enrich on the URL.
2. find_work_email using their name and their current company's domain.
3. verify_email on the result.
Return a table with: name, title, company, domain, email, email status.
Do not enrich anything else.Three things make this reliable, and they are the whole of the prompting guide in miniature:
- The tools are named. “Find their email” leaves the model choosing among several plausible tools.
- The steps are ordered. The chain is explicit rather than inferred.
- The scope is closed. “Do not enrich anything else” stops a helpful model from running five more lookups.
That last line matters more than it looks. Without a scope limit, an assistant asked for a contact card will often decide the company details would be useful too — and each of those is a billed call you did not ask for.
Reading the output
| Field | What to check |
|---|---|
| Title | Does it match what you expected? A stale title means the profile has not been updated |
| Company | Does the domain match the company name? Mismatches mean the wrong entity |
| Does it match the company’s actual pattern? | |
| Email status | valid or nothing. Everything else is not contactable. |
A catch-all result is not a failure — it means the domain accepts everything and the specific mailbox cannot be confirmed. Treat it per your own policy; see Waterfall Enrichment lesson 03.
When a step misses
| Miss | What to do |
|---|---|
| Profile enrichment returns nothing | Check the URL is a canonical /in/ profile URL |
| Email not found | Try the LinkedIn URL as the input instead of name plus domain |
| Email found but invalid | Do not send. Ask for the personal-email path only if you have a reason |
| Verification returns unknown | Retry once; if it persists, treat as catch-all |
Doing it for several people
The same prompt with a list, and one addition:
Do this for these five profile URLs: [...]
Cap at 5. Return one table with all five rows.
If any step fails for a person, leave the field blank and continue —
do not stop the whole run.The cap and the continue-on-failure instruction are what keep a multi-row run predictable.
Do this now
Pick someone you can verify
A colleague or a customer whose real address you know.
Run the prompt
Compare against what you know
Title, company, email. This is the only real accuracy test.
Check the status field
valid, not just “found”.
Run it for five people
With the cap and continue-on-failure lines.
Note the credits used
Compare against the ~2.3 per contact estimate.
Check your work
- The output is a table, not prose
- Every row has an email status
- You verified one row against someone you actually know
- No extra tools ran beyond the three you named
- Credit consumption matched the estimate
Where this breaks
A vague prompt costs more than a specific one, and the cost is invisible until you check the balance. “Tell me everything about this person” invites an assistant to call profile enrichment, company enrichment, tech stack, job listings and social posts — five billed calls where you wanted three. Name the tools, name the order, and close the scope.
Further automation
Once this chain works, save it as a reusable skill so you never retype it. The GTM agents library has ready-made versions, and AI-Powered GTM lesson 05 covers writing your own with caps and guardrails.
Next lesson
03 — A company brief under five credits, the same discipline applied to an account before a call.
Reference for this lesson: linkedin_profile_enrich, find_work_email, verify_email, MCP prompting guide.