Every recruiter hunting for a Senior Data Engineer this week is running roughly the same LinkedIn search: "("data engineer" OR "data engineering") AND (python OR scala) AND (spark OR databricks) AND 5..10 years." Some will use `-manager -lead` to filter out senior titles. A few will sharpen with specific tools. The variations are small.

The problem isn't the boolean string. The problem is that every recruiter at every competitor is running the same one. The top 200 candidates who match that search know they're in demand. They've been contacted twelve times this quarter. Response rates on that kind of outreach are in the single digits because you're not sourcing — you're standing in line.

The recruiters who consistently out-hire their peers aren't running better versions of that search. They're running different searches, in places their competitors aren't looking. AI is the fastest way to do that at scale.

Three techniques, each with a prompt pattern that works.

Technique One: Boolean Strings That Cover What You'd Never Think of

The obvious boolean string finds the obvious candidates. A better one surfaces adjacencies — people whose backgrounds overlap with what you need but don't match the surface title.

For the data engineer role above, the obvious search misses: ML engineers who built data pipelines as infrastructure work, analytics engineers at modern data stack companies who do most of what a data engineer does, backend engineers at data-heavy startups where everyone touches the warehouse, former data scientists who pivoted into engineering. Each of those profiles could be a strong hire. None of them show up in the boolean string most recruiters are running.

AI generates these adjacencies naturally because it understands the work behind the title, not just the title. You describe the role in plain language. You ask for Boolean strings that find candidates doing that work under different titles. You get back half a dozen searches you wouldn't have thought of.

Prompt pattern:

"I'm sourcing for a Senior Data Engineer at a Series C startup. The role will own data pipeline infrastructure (Airflow, dbt, Spark), work closely with analytics and ML teams, and help scale a system currently handling ~2TB/day.

Generate 6 LinkedIn boolean search strings that would surface strong candidates. For each:
What kind of candidate profile it's targeting?
Why this adjacent background transfers well to the role?
One filter to exclude candidates who probably won't fit.

Don't repeat the obvious "Data Engineer" search. Focus on adjacent titles and overlapping skill backgrounds."

Run that and you get searches for Analytics Engineers at dbt Labs-adjacent companies, Backend Engineers at data-infrastructure startups, ML Infrastructure Engineers whose day-to-day is pipeline work, and three or four others. Each search targets a pool your competitors aren't fishing in.

Technique Two: Adjacent Role Mapping

Technique one broadens the search. This one inverts it.

Instead of starting from "who has this title," start from "what does this person actually do, and who else does that?" Adjacent role mapping takes a role definition and identifies the non-obvious titles where similar work gets done, often at higher caliber, because the people in those roles are doing the work without the recognition of the job title.

The clearest example: when you're hiring a VP of Engineering at a growth-stage startup, the best candidates often aren't other VPs of Engineering. They're Senior Engineering Directors at larger companies who've been doing the VP job in everything but title. They're often underpaid, underrecognized, and hungry for the step up. They'll never appear in a VP search.

The same pattern holds across seniority levels. The best Head of Marketing for a Series A B2B SaaS company is frequently a Senior Director at Series C who's been running the function without the title. The best Senior PM for an infrastructure product is often a Staff Engineer who's been acting as the PM because there wasn't one. These profiles are everywhere, and they're invisible to keyword-based sourcing.

Prompt pattern:

"I'm hiring a VP of Engineering for a 40-person Series B startup. Candidates need: 8+ years experience, managed 15–30 engineers, scaled a team through hypergrowth, comfortable being hands-on when needed.

Map out the non-obvious candidate pools I should be sourcing. For each pool:
Titles I should search for
Company stages or sizes where these candidates are common
The signal that distinguishes a high-potential candidate in this pool from the average one
The likely reason they'd be open to this role

Avoid suggesting "other VPs of Engineering." I want the pools my competitors aren't searching."

This prompt produces a short map — maybe five pools. Senior Engineering Directors at 500+ person companies. Technical Founders whose last company sold. Principal Engineers who've been managing in practice. Engineering Managers at FAANG who've capped out at scope and want a bigger canvas. Each comes with its own signal and its own reason for motion. That's the map you use to source for the next week.

Technique Three: Passive Signal Detection

This one is harder and higher-leverage.

Most sourcing focuses on "Who matches the role?". Passive signal detection asks a different question: "Who's likely to be open to a move right now, regardless of whether they've updated their LinkedIn?" The best candidates don't have "open to work" badges. They're not applying. They're not actively looking. But they might be restless, and if you reach them in the right moment with the right message, they move.

The signals aren't mysterious. They're just hard to aggregate manually. People who've been in their current role for around the median tenure for their function — typically 2.5 to 4 years. People whose company has had a recent event: layoffs, leadership change, acquisition, down round. People whose career trajectory has been steep, and whose current role shows signs of slowing that trajectory. People who've recently started engaging with content from competitors in your space. People who've recently moved laterally rather than up.

You can't run a boolean search for "restless." But you can describe the signals, and AI can help you structure the hunt.

Prompt pattern:

"I'm sourcing for roles that require senior ICs in applied ML (research engineer / ML engineer, 6+ years). I want to build a list of candidates who are likely to be quietly open to a move, even if they haven't signaled it publicly.

Help me define the signals I should watch for. For each signal:
What it is and how to detect it (LinkedIn, GitHub, news, etc.)?
Why it correlates with openness to a move?
The kind of outreach message that would land, given this signal

I'm not interested in "recently posted about looking." I want the signals that most recruiters don't notice."

This produces a sourcing playbook that's specific to your role. It won't replace your intuition, and it isn't a magic list of candidates. What it does is make your sourcing structurally different from the recruiters running the same saved searches they've been running for years. You're watching for things they aren't watching for. That's the whole edge.

The Pattern Underneath All Three

The three techniques look different on the surface. They share one underlying shift.

Traditional sourcing optimizes for match. Who has the title, the years, the keywords? AI-assisted sourcing optimizes for "fit and reach". Who has the capability, the trajectory, the motivation — even if they don't look like the obvious hire on paper? That shift is what AgentR's career pattern analysis was built to do at scale: look past the keywords to the underlying work, and surface the candidates a keyword-first system would miss.

The candidates a keyword-first system misses are precisely the ones your competitors are missing. Which is where your hires tend to come from, if you're paying attention.

Next: Lesson 10 — How to Spot a Fake Application