Why Maki for AI hiring: the case for hiring intelligence you can defend

Every enterprise function got rebuilt on modern software over the last twenty years. Finance did. Sales did. Support did. Hiring did not.

The applicant tracking system files applications. It does not evaluate them. So every gap in the process got patched with a point solution: an assessment tool, a video interview tool, a scheduler, a reference checker. The average recruiter now runs seven tools that do not talk to each other, and at the end of all of it the decision still comes down to a CV scan and a gut call.

AI is the first technology that can do the work rather than store the record of it. That is the opportunity, and it is also where most of the market is about to go wrong.

The problem is not speed. It is evidence.

Ask a hiring manager why a candidate was rejected at screening and you will usually get an honest answer that is impossible to defend: the CV did not look right, the profile felt junior, we had four hundred applications and forty minutes.

None of that is written down. None of it is consistent between two recruiters looking at the same pile. None of it survives contact with a works council, a regulator, or a rejected candidate who asks why.

Speed is the symptom people feel. The underlying problem is that hiring produces almost no evidence, which is why it resists improvement. You cannot tune a process that does not record what it decided or why.

Automation without science is bias scaled

Anyone can wire a large language model into a hiring funnel. A weekend is enough to build the first version, which is why a new one appears every month.

Here is the part that gets skipped. Language models are pattern matchers. Let one score interviews without a validated rubric and it will reproduce whatever bias sits in its training data, at industrial speed, with a confident tone. Faster bad decisions are not an improvement. They are a bigger liability.

Maki inverts the order. The science layer defines what is measured: the dimension, the behavioural anchor, the scoring grid, built on assessment methodology that predates the AI wave and has been validated against expert raters. The AI executes the measurement.

That inversion is the whole product argument. When a regulator, a works council or a candidate asks how a score was produced, the answer is a documented dimension and a validated grid, not an appeal to the model.

One system, from the job definition to the offer

Maki is the intelligence layer between the applicant tracking system and the recruiter. The system stays. The recruiter stays. What Maki replaces is the manual, inconsistent and undefendable work that currently sits between them.

It runs where the recruiter already works

The fastest way to kill an enterprise hiring tool is to make it a place people have to go.

Maki runs inside the applicant tracking system the team already lives in, with more than thirty integrations across the major enterprise systems. Every action writes back. There is no new login to enforce, no parallel source of truth, and no quarter spent migrating anything.

This is also the honest answer to the question every enterprise buyer asks in the first meeting, which is not whether the AI is good but how it fits the stack they already bought.

Fairness over surveillance

The category defaulted to watching candidates: webcam monitoring, real-time flagging, proctoring that treats every applicant as a suspect.

We took the other position, and published our reasoning. Real-time proctoring has a high false-positive rate, it is sensitive to environment, accessibility and culture, it is difficult to explain after the fact, and reacting live to an anomaly often introduces more bias than it prevents. A false accusation of cheating does more damage to a person, and to an employer brand, than a missed one.

So Maki does post-hoc anomaly detection instead: temporal analysis, cross-answer consistency, behavioural patterns across populations, comparison against expected distributions. Integrity is protected by evidence, not by surveillance.

Defensible, whichever way the rules go

Hiring is where AI regulation has landed first, and the picture is now a patchwork rather than a single deadline.

We think the deadline is the wrong thing to organise around anyway. The real exposure was never a fine on a date. It is a plaintiff, a works council, or a rejected candidate asking how a score was produced, and that question does not expire.

So every scored decision in Maki carries an audit trail: what was measured, against which grid, with what result. No demographic data is used in scoring. And the system is decision support by design. It ranks, it evidences, it flags. It does not make the hire.

The part that compounds

Every funnel that runs through Maki teaches the system: validation studies, scoring grids, outcome data, accumulating against real roles at real companies across sixty countries and more than eighty works council approvals.

A model-only competitor can match a feature in a quarter. It cannot start with the record of what actually predicted performance, and that record gets longer every month.

What it looks like in production

The pattern is consistent across enterprise deployments. Recruiters get back the hours they were spending reading CVs. Candidates finish what they start, at rates above the industry norm, because the experience is short, conversational and available when they are.

Time to hire falls, not because any single step got faster, but because the process stops going backwards: fewer re-screens, fewer repeated interviews, fewer debriefs where nobody can remember what was asked.

And the hires hold. When selection is anchored to a defined rubric rather than a CV impression, early tenure turnover comes down, which is the number that actually pays for the system.

Why Maki

The choice in AI hiring is not between automating and not automating. That argument is over.

It is between automation that scales whatever the model picked up, and automation that scales a measurement somebody can defend. One of those makes a recruiter faster this quarter. The other makes hiring a system the company can actually improve.

Frequently asked questions

Does Maki replace our applicant tracking system? No. The ATS stays the system of record. Maki is the evaluation layer between it and the recruiter, and every action writes back.

Does the AI make the hiring decision? No. Maki is decision support: it measures, ranks and evidences. A person decides, with the reasoning in front of them.

Where does Maki start in the funnel? Screening is the entry point for most teams, either skills-based or voice, because it replaces a step that already exists rather than adding one.

How do you handle bias and compliance? Scores are anchored to validated rubrics rather than inferred from training data, no demographic data is used in scoring, and every decision carries an audit trail. Read our position on transparent AI in hiring.

What about candidates who dislike AI in the process? Candidates are told when they are interacting with an AI system, the experience is short and available on their schedule, and integrity checks run after the fact rather than as live surveillance.