AI-Powered Candidate Sourcing: How to Build a Shortlist in Minutes

Why Candidate Sourcing Is Broken (And Where AI Helps)

Traditional candidate sourcing has two failure modes:

Too slow: A good sourcer might evaluate 20-30 candidates per day manually. For competitive roles, a thorough search requires 200+ profiles to find 10 worth pursuing - a week of work before you've sent a single message.

Too shallow: Speed pressure leads to shortcuts. Titles get skimmed. Profiles get judged by company name instead of actual experience. Good candidates at unfamiliar companies get passed over.

AI sourcing doesn't replace your judgment about who's a good fit. It removes the bottleneck between your criteria and a qualified shortlist.

Step 1: Write Your Candidate Profile Before You Search

Before opening LinkedIn or any sourcing tool, write down:

Must-haves:

• Minimum years of experience in the specific skill area
• Specific tools, languages, or methodologies required
• Geographic constraints (remote vs. on-site)

Strong signals:

• Experience in a similar industry or company stage
• Evidence of scope and ownership (not just titles)
• Side projects, open-source contributions, or writing

Disqualifiers:

• Currently at a direct competitor
• Role mismatch (manager profiles when you need individual contributors)

Step 2: Choose Your Sourcing Channels

LinkedIn is the broadest and most up-to-date source. The right starting point for most roles.

GitHub is essential for engineering roles. Contribution history and repository quality reveal more about a developer's actual work than their job title does.

Portfolio sites and personal blogs surface candidates who invest in their professional presence - exactly the kind of initiative you want to hire.

Job boards show you who's actively looking. Lower barrier to engagement, but higher competition.

Community sources (Slack groups, Discord, newsletters) surface candidates who aren't looking but are engaged in their field. Often the highest-quality hires.

Step 3: Research Each Candidate with AI

For each candidate, you want to answer:

1. What have they actually built or accomplished? (Not just title - evidence of scope)
2. What's their trajectory? Growing, stagnant, or declining responsibility?
3. What signal exists beyond their profile? Posts, GitHub activity, conference talks, articles.
4. What's the likely pitch? What about your role is meaningfully better than their current situation?

Doing this manually for 50 candidates takes most of a day. With Strawberry, it's a workflow: the AI opens each profile, extracts relevant signals, evaluates them against your criteria, and returns a ranked shortlist with notes.

Step 4: Write Personalized Outreach That Gets Replies

Recruiter outreach has a reputation problem. A message that demonstrates you've actually looked at their work stands out immediately.

Line 1: Reference something specific. "Saw the Figma plugin you published last year - the focus on keyboard navigation caught my attention."

Line 2: One sentence on why the role is relevant to them specifically.

Line 3: Low-friction ask. "Would you be open to a 20-minute call to hear more?"

Keep the whole message under 150 words. With Strawberry, you generate this first line at scale by feeding each candidate's profile and your job description to the AI, which surfaces the most relevant connection point.

Step 5: Track Everything in Your ATS

Every sourced candidate should flow into your ATS when you first identify them - not after they reply. With Strawberry, you can build this into the sourcing workflow: the AI evaluates the candidate, adds them to Ashby or Greenhouse with notes, and creates a follow-up task if they don't respond within 7 days.

A Full AI-Powered Sourcing Run

1. Define criteria: 5+ years B2B marketing, experience scaling content or demand gen, Series A-B company, open to Stockholm or remote.
2. Source: Pull 200 profiles from LinkedIn Sales Navigator matching your filters.
3. Research loop: Strawberry opens each profile, pulls last 3 job descriptions, recent posts, public content. Flags top 20 against criteria.
4. Draft outreach: Generate a personalized first line for each of the 20 flagged candidates.
5. Send and log: Send via LinkedIn InMail or email, create candidate record in ATS, set follow-up for day 7.
6. Follow-up sweep: Day 7 - one-line bump. Day 14 - move non-responders to passive watch list.

The full run for 200 profiles takes about an hour with AI research versus a full day manually.

The Recruiter's Role in an AI-Assisted Process

AI handles reading, filtering, and first-line drafting. The recruiter's job shifts toward:

• Judgment calls on culture fit and team dynamics
• Building genuine relationships with candidates over time
• Negotiating and closing
• Representing the company voice accurately

The best recruiting teams in 2026 have figured out exactly which parts of their process AI should own and which parts need a person. Sourcing and initial research is firmly in the first category.