Course › Module 3 · Sourcing with AI

Lab 2: a 50-profile longlist in 40 minutes

Module 3, Lesson 8  ·  4 min read ·  Updated 21 September 2026

Module 3 · Lesson 8

Forty minutes, one job, fifty people. The point is not the list. It is finding out where your time actually goes, and whether your searches survive contact with reality.

What to do

Use the job blueprint you built in Module 2. Find 50 people for that job, working to the clock.

The clock

MinutesDo thisYou should have
0–5Write down your target and limits from the blueprintMust-haves, location, seniority on paper
5–12Ask AI for the three lists from Lesson 4Job titles, skills, companies
12–17Check the new words — ten results eachWords sorted into keep and drop
17–22Build the search in three blocks; test on ten resultsA search where 5 of the first 10 look right
22–35Run it across sites and collect people50 people, with the source noted
35–40Scan for obvious mismatches and duplicatesA cleaned list

If you run over, note where. The overrun is the finding. Most people lose the time in step four, rebuilding a search they should have saved as a template in Lesson 2.

What to hand in

A list of 50. Each row has the profile link, which site it came from, and which search found it. That third column is what makes the list useful later — you will be able to see which search produced the people you actually contacted.

Plus the searches themselves, saved in the file from Lesson 2.

How to know you have done it right

CheckGood looks like
Someone else could repeat itYour searches would give them a similar list
Search qualityAt least 5 of the first 10 results looked right
Mix of sourcesMore than one website represented
The widening workedAt least 10 people found by a word you did not start with
You can trace itEvery row records where it came from

The fourth row is the one that matters. If every person came from your original job title, the widening step did nothing and your pool is as narrow as when you started.

Without any sourcing tools

Everything here runs on a search engine and a spreadsheet. The X-Ray searches from Lesson 3 need no extension, no subscription and no account — that is the point of them. Record each person with the search that found them. No tool gives you that column automatically, and it is the most useful one.

Check yourself before Module 4

  • Which single search word brought in the most useful people? Add it to your library.
  • Which word looked promising and returned the wrong people? Write it down as a known trap.
  • How many of the 50 would you actually contact? Under half means your search is too loose. All of them means it is probably too tight.
  • Does your list look like your existing team? If so, re-read the fairness section in Lesson 3 before going further.

Keep the list

Module 6 uses it for writing messages, and Module 10 uses it to work out your own conversion rates — which will replace the rough planning numbers from Lesson 1 with numbers that are actually yours.

Common questions

What is X-Ray search in recruitment?

X-Ray search uses a general search engine to look inside a platform's public profiles rather than using the platform's own search — `site:linkedin.com/in` plus your Boolean terms. It costs nothing, needs no extension, and finds people the internal search ranks away from you.

Is scraping LinkedIn legal for recruiters?

Major professional networks prohibit automated collection in their terms of service and enforce it by restricting or closing accounts — usually automatically, often without appeal. That is a business risk rather than a criminal one in most places, but the activity runs under your account. Manually viewing public profiles is a different thing and is what the platforms are built for.

How many candidates do you need to make one hire?

As a planning starting point, 150 to 250 profiles identified, 60 to 100 contacted, 12 to 25 replies, five to seven first interviews, two to three finalists, one or two offers, one hire. Reply rate moves more than anything else on that list. Replace every figure with your own after two or three hires.

Do you still need Boolean search if you have AI?

Yes, and arguably more. Boolean does exactly what you asked rather than what a ranking model inferred you meant, which makes it the right tool for genuine must-haves. The best setups use Boolean for hard requirements and semantic matching to order what survives.

Check yourself

Four questions. Nothing is recorded anywhere but your own browser — this is for you, not for a score.

  1. 1What does `python OR java AND senior` actually search for?

  2. 2X-Ray search only finds people who maintain public profiles. Why does that matter beyond coverage?

  3. 3An expanded search term returns a large volume of the wrong people. What should you do?

  4. 4When does a data-protection obligation attach to a sourced profile?