AI Hiring Tools: Great on Paper
At Meetsmore, AI is part of how we work. Engineers use whatever AI tools help them do their jobs efficiently, customer service and business teams have broad access to AI tools and LLM chat systems — with the condition that any tool handling sensitive or personal data must not use that data for model training. Managers use AI to create meeting slides, take automatic notes from transcripts, and better organize our operations. Even this article was proofread by an LLM before asking for internal review.
Hiring is no exception. We use AI where it improves the experience for both us and the candidate — providing clearer introductions to the company and the role early on, streamlining logistics, and cutting down on the dead time that makes hiring feel impersonal.
Where we draw the line is in the evaluation. Reading resumes, assessing technical ability, deciding whether someone is the right fit — those stay human. We've tested the tools. They don't do it well enough.
If you've interviewed with companies that lean heavily on AI in these areas, you know what that feels like. Resume screening algorithms can be powerful, but they create a competition where you must craft a resume to rank higher on these algorithms while possibly stretching the truth about your proficiencies. Pre-recorded interviews can be more flexible, but they lack the organic chemistry between two people and leave very little room for personal touches — leaving you feeling like just another number on the list. And then there's technical testing: machine-verified, overly challenging, and leaving little room for error — which leaves you wondering what the actual job will even be about.
Here's why we've made the choices we have.
Case 1: Resume Screening
Let's be honest, if you ever started work and saw that you had 163 resumes come in over the weekend after a new job posting on Friday, you'd feel the coffee wasn't strong enough to help you with this ordeal. Why not have some AI burn through these resumes in 30 seconds and report to you the cream of the crop so you can manually evaluate them? Especially when many hiring managers report that more than half of resumes received don't meet the bare minimum requirements. Why not have AI filter them out?
It comes down to nuance.
Take a job posting that requires "5 years work experience in Typescript." Now consider two candidates: "Bob is a software engineer with 7 years of experience, specialized in Typescript/Javascript for over 2 years" and "Mary is a web developer with 5 years of work experience."
An AI screening bot would look at these and likely rank Bob higher — he has more years and specifically mentions Typescript. Mary might get flagged lower because while she passes the minimum years, she doesn't specify the skill early on. Bob passes the screening.
But here's the issue. We specified "5 years work experience" — and that distinction matters. Read further on Bob's resume and he might realistically have 2 years of applicable work experience in Typescript and 5 years from university as an IT major. Combined, that's 7 years of programming experience but only 2 years of work experience. Mary, on the other hand, is a web developer — a role that works closely with Typescript/Javascript daily — and she specifically noted work experience. Mary would be the better candidate between these two, but AI wasn't tuned to make that connection.
This is just one example of many in the complicated battle between candidates building resumes to compete and hiring managers trying to find the right people. And it's getting harder. There's a greater rise in candidates using AI to craft their resumes and cover letters, and even using programs to seek out job postings and automatically apply. The downside to manually combing through resumes is that they're all starting to read the same: "Full-stack software developer with 5+ years of experience focused on building and maintaining codebases with an emphasis on clean code and maintainability in mind" — perfectly matching the job posting. Previous job details reading "Created backend APIs with Typescript, Node, and PostgreSQL" — targeting terms for machine scanning. Cover letters opening with "I am thrilled to come across your job posting and feel that my skillset and experience match perfectly with [COMPANY NAME]" — just copy and paste.
For people fresh to reading resumes, these might sound quite good. For those who've read hundreds with the same exact wording, no one starts to stand out. But that's no fault of the candidates — they're all trying their best to compete, and creating a truly unique resume that's also optimized for automated screeners is extremely hard, almost a skill in itself. This is where we at Meetsmore try to be the better filter by manually checking through resumes to find anyone who fits our hiring criteria.
Case 2: Testing
One of the most important aspects to interviewing for highly technical positions is the testing round — or perhaps multiple rounds. Speaking as a software engineer, we often undergo two or more rounds of testing, both coding and theory, and many of us can tell you that the tests we perform are often not reflective of the actual job.
For software development, there's Leetcode and HackerRank which offer automated, customized testing that the interviewer can send off to the candidate to finish within a week. These platforms have all kinds of safeguards — keylogging, webcam requirements, AI detection — and they sound amazing at first. But the pre-created tests are often high-level problems that aren't reflective of any real work processes. It's well known that candidates study specifically to pass these tests. It's become almost an infamous secret handshake — "if you can jump through the flaming hoops the same way I did, you too should be hired."
The issue with this approach is that it treats a person like a math problem. You either pass or you don't. But that's not how work actually functions — nobody writes code in isolation, nobody gets everything right on the first try, and nobody is expected to. So why would we test people that way?
At Meetsmore, we try to treat our technical tests more like a working session than an exam. For mid-level and senior software engineers, we typically have two rounds — a programming test in Typescript, and a higher-level architecture design interview. Each test can have multiple prepared versions with multiple challenges that help us assess a candidate's skillset. But there's no "passing score." There's a minimum bar we expect the candidate to clear, but beyond that, there's no score sheet.
What we care about is how the person works — and that includes how they work with us. We proctor our tests with the understanding that nothing is done perfectly from the start, and some tests lean more into this than others depending on the interviewer and the role. Think of it like a PR review: sometimes a few comments and suggestions are what help someone get to a better solution. We're not going to give anyone the answers, but if a candidate starts overbuilding or planning too deeply when the goal is straightforward, we'll help rein in the scope. If they're stuck debugging, we'll work through it with them. If they're not deeply familiar with every Typescript type or Javascript method, that's fine — we want to see how they think and program, not what they've memorized.
It's a lot like how musicians judge other musicians — you can often tell within the first few seconds of someone playing where their ability sits, and then you watch how they handle the hard parts. Programming works the same way. We get a feel early on, and then we watch how candidates handle the peak challenges. The difference is that we're not just watching — we're in the room, and we can push and pull the test to find out what someone is really capable of.
This approach also happens to make cheating a lot harder. AI-assisted tools exist that can overlay invisible prompts on a candidate's screen and walk them through building a solution — including deliberate mistakes to seem genuine. But when the proctor is actively collaborating, asking about thought processes, and adjusting the test on the fly, those tools break down quickly. A knowledgeable human can rationalize and reject a misleading suggestion; an AI overlay can't adapt to a conversation it wasn't expecting.
That said, not all of this is cut and dry. We also trust our instincts. A candidate might have passed and seemingly done well, but if there are oddities we can't explain, we discuss it with other hiring managers and come to an agreement together.
In Summary
Hiring is about people, and evaluating a person is very different from evaluating a mathematical solution. We show different strengths and weaknesses depending on the angle you're looking from — and that goes for both the interviewer and the candidate. That's what makes it hard to automate the parts that matter most, and that's what makes us selective about where we bring AI into the process.
We're not standing still. We're looking for the right tools, testing them, and bringing them in where they make sense. But when it comes to reading your resume, sitting across from you in a technical test, or having a conversation about who you are and how you work — that's us, not a machine.
If you're someone interested in applying to Meetsmore: we'll be evaluating your resume with real human eyes, grading your test answers with human rationality and understanding, and conducting our interviews with genuine curiosity about who you are.
You may find our current open positions here: https://hrmos.co/pages/meetsmore
