Laidback and the AI sourcing category

Most AI sourcing tools are a language model on top of the same licensed dataset everyone else licenses. The interface is new, the candidate pool is not.

Start for free Book a demo

$100 of credits on us. No card required.

The short answer

Adding a language model to a database query makes search feel modern. It does not add a single new candidate to the pool.

Underlying dataOwn index built from the open web

ReasoningDeterministic scoring model grounded in structured evaluation

ConsistencySame brief, same shortlist, every time

The difference in one table

	Typical AI sourcing tool	Laidback
Underlying data	Licensed snapshot shared across vendors	Own index built from the open web
Reasoning	Prompt over a database query	Deterministic scoring model grounded in structured evaluation
Consistency	Results shift between runs	Same brief, same shortlist, every time
Calibration	Swap a keyword or a filter	Re-derive scope, skills, comp band and disqualifiers
Scope of work	Search only	Search, outreach drafts, coordination and standing habits

Judge the pool, not the interface

Measure the overlap

Run one real brief through your current tool and through me, then count how many names appear in both. Low overlap is the point.

Ask for the evidence

If a tool cannot show why a candidate ranked where they did, the ranking is decoration.

Run it twice

Same brief, two runs, same day. If the shortlist changes, there is no model, only a sampler.

Check who does the follow through

A list is not a hire. Ask what happens between the shortlist and the interview.

#evaluation

Alice

We trialled two AI sourcing tools and got almost the same names.

Laidback

Expected. Same snapshot in, same shortlist out. The model on top does not change the pool underneath.

Laidback

Run the same brief past me and compare overlap. That is the only test worth running.

Reproducible, so it is testable

How to tell the difference in a trial

Every tool claims AI sourcing. What is different here?

Two things. Most tools resell the same licensed database, so the same names come out. And most rank with a model that changes its mind. I index the open web myself, and my scoring is deterministic and explainable.

Will it hallucinate candidate details?

Every claim on a profile links to the source it came from. If I cannot evidence something, I say it is not evidenced rather than filling the gap.

Do we have to change how the team works?

No. I live in the channel your team is already in. There is no new dashboard for hiring managers to ignore.

The work was in building a different index and a model of what makes a good hire. That is the part I spent my time on.

Recruiting judgement ships with me

You are not writing prompts to teach me what a strong candidate looks like. I already know what this job is, which is why the first shortlist is worth reading.

Different data, not the same resold list

Most tools query the same three databases. I build from work people leave in public, so the person who never updated a profile is still findable.

Skills nobody wrote down

Nobody lists that they carried a migration or ran the on call rotation. I infer it from the work and then check it.

I widen the brief when it is too narrow

If your search returns eleven people, I show you the adjacent titles and teams worth opening up to, and say what the trade is.

The people who never applied

Silver medallists, referrals from last year, the engineer whose company just froze hiring. I keep watching so you do not restart from zero.