Why we built Recrewit.

Both sides of the desk.

1.1We are two engineers. We have sat on both sides of this. We have sent the applications, and we have run the searches as a recruitment firm sourcing for real roles.

1.2From a candidate seat it looks like silence. You press apply and nothing comes back. Not a no. Nothing.

1.3From the sourcing seat it looks like a shortage. You run a search across hundreds of applications and eleven names come back, none of them right, and you know perfectly well the person you need is in that pile somewhere.

1.4Same failure, two views of it. Nobody is being screened properly, and everyone assumes the other side is the problem.

What the filing cabinet does.

2.1The system most companies run applications through is called an applicant tracking system. It sounds like a judge. It is a filing cabinet.

2.2Your resume goes in as text. Nobody reads four hundred resumes, so a recruiter types words into a search box and the cabinet hands back whoever used those exact words.

2.3The filters get set tight, because loose ones return everybody. Must have this title. This many years. These words in this order. Whole roles come back with a handful of names, and everyone cut was never screened by anyone at all.

2.4It fails earlier than that too. Put your work in two columns, or a table, or a header, and many of these systems read the text scrambled or not at all. Your experience is in the file. It is simply not in the database.

What replaced it is worse.

3.1Recruiters know the cabinet is failing them. So a lot of screening now happens somewhere else entirely: a general chatbot, open in another tab. Paste the resume, paste the job, ask whether they are any good.

3.2It is quick, and it reads far better than a keyword search. It was also never built for this. It has no rubric for the role. It has no idea what the last hundred candidates scored or why.

3.3Ask it the same question twice and you get two different answers. Nothing is written down. Nothing is checked for bias. Your name, your school and your address go in with everything else, and they do affect what comes back.

3.4That is the part we could not leave alone. A general assistant is not trained to find talent. It is trained to sound helpful, which is a very different job.

How good people go missing.

4.1Both of those match on vocabulary rather than ability. Two people who do the same work every day describe it differently, and only one of them is found.

4.2You wrote quality assurance automation. The search was for a software development engineer in test. Same work, same tools, same person. Different words, and you are invisible.

4.3So people learn to write resumes for the machine instead of for a reader. Stuff in the words, mirror the posting, repeat the title. Everyone starts to sound identical, and the machine gets worse at telling anyone apart.

4.4It does not fail at random. It fails hardest against the people who were doing the work and calling it something else.

So we built Recrewit.

5.1Recrewit is built for this one job, rather than a filing cabinet doing a job it cannot do or a chat window doing one it was never given.

5.2Your resume is read once, properly, to fill in your profile. Your roles, your skills, your projects. You correct anything it got wrong and you never type your career into a form again.

5.3When you apply, the same rubric is used for you as for everyone else on that role. What the job actually requires, weighed against what you have actually done, in sentences rather than word counts. Ask it twice and you get the same answer, because it is not improvising.

5.4Every result carries plain reasoning: what lines up, what does not, and why. A recruiter reads that and can disagree with it. It is written down, so it can be checked later.

5.5For engineering roles there is a short coding interview in your browser. You show the work. That is a better answer to can they do this than any arrangement of keywords has ever been.

A person still decides.

6.1Nothing here rejects you. It reads, it scores how well the work lines up, and it ranks. Every decision after that is made by a person on the hiring team, and they can overrule all of it.

6.2That split is the whole idea. Software is good at reading two hundred resumes for technical fit and hopeless at judging whether someone belongs on a team. People are the other way round. So the software finds the technical fit and the recruiter finds the human one.

6.3You can ask for a person instead of a model at any point, and you can ask why a result went the way it did. Write to support@recrewit.ai.

What we will not do.

7.1Your name, photo, graduation year, school name, address, zip code, pronouns, age, citizenship, marital or parental status and disability status are stripped before anything scores your application. A school name is not a skill.

7.2Nothing scores your personality, your working style or whether you seem like a good fit. Those words hide a great deal of bias, and a machine has no business anywhere near them.

7.3If a resume reads as though a model wrote it, that is reported on its own and never folded into your score. It is a note for a human, not a mark against you.

7.4Nothing you give us trains a model unless you say yes to that. The answer is no until you say otherwise. The long version is on the privacy page.