I have been listening to a lot of Sherlock Holmes recently. Stephen Fry reading Conan Doyle is, as it turns out, an exceptionally good way to occupy your mind during activities that require your body but not a great deal of your brain. At the gym, driving, refitting windows on a shed...although that did prove to me that maybe I'd miscalculated just how much of my brain I would need to use for that one. But that is a very different story. One of the most famous moments in the canon is Holmes drawing a crucial clue not from what happened, but from what conspicuously did not. A dog that should have barked at an intruder stayed silent, which told Holmes everything he needed to know about who that intruder really was.
I keep thinking about that story in the context of conversations I have been having repeatedly over the past few months with people who are north of 35 and who are actively participating in the job hunt game. I have been looking for a while now, which means I have rather more first-hand experience of what follows than I would like.
The conversations usually begin with someone venting about a job application. They applied, they waited, they got the standard rejection. No feedback, just the familiar "we've decided to move forward with other candidates who more closely match what we are looking for". If they try to ask for more information as to where their application, which more than covered every requirement, missed the mark, they either get complete silence or they get the explanation that has been trotted out for years now...
"With so many applicants, it simply isn't possible to give feedback to everyone"
That may well have been true in the past. I'll grant that.
But let me describe something that has been mentioned by many people and has also happened to me. You finish an application form for a role at 23:21 on a Friday night. By 00:12 on Saturday morning, forty-three minutes later, you receive the rejection email.
Now, either there is a very dedicated hiring manager who knocked off a long week, poured themselves a drink, and thought "actually, do you know what, I fancy looking through a handful of application forms before I go to bed", or the first filtering stage was handled by an ATS using AI.
It was the ATS. Nothing ever happens so quickly in the recruitment world when a human has the action to perform.
Now, let’s picture a scenario. Take a list of reasons an automated system might have for rejecting a candidate. Feed them into an LLM. Ask it to produce a personalised, constructive feedback message. How long does that take? Seconds. Now imagine that same ATS, for every applicant it decides does not meet the criteria, doing exactly that before it fires the rejection. The applicant learns something. They can refine their CV, reconsider how they are positioning themselves, understand what the system is looking for. It is not perfect feedback, but it is something.
They don't do that. Instead, you are left in this dystopian void. The equivalent of trying to learn how to parallel park while blindfolded. What would be easy if you could see, becomes very expensive with the damage it causes to your car when your only feedback is a thud! That damage to the car in that dystopian void reflects the serious damage carried out on you, ever feeding your imposter syndrome.
But here's the thing...they really COULD give this feedback. Not just theoretically. The addition of "AI-powered" natural language explanations bolted onto existing workflows has been, in many products across many industries, almost always the first thing added when a company wants to stick the "AI-powered" label on the tin. It is often not terribly sophisticated. It is an LLM reading structured data and producing a human-readable summary. It is everywhere. It is practically the default move.
So why not here? Why do we all see the standard "we've decided to move forward with other candidates who more closely match what we are looking for"...even when the very next day they are reposting the role? These systems are making life-altering decisions about people's careers, yet the entities who build and manage these systems seem wilfully unconcerned about the situation.
It has not been forgotten. It has been purposely not added.
The only logical reason I can arrive at, after sitting with this for a while, is that producing that feedback would create a paper trail. It would make the decision legible. If the decision is legible, the biases embedded in that decision become legible too.
I want to be clear that I do not think this is purely an ATS problem. The rules fed into these systems come from somewhere. Recruitment teams set the criteria, and those criteria carry bias. The ATS encodes/learns what it is given. Ageism is real in recruitment, and I know this not just as a general observation but from direct experience.
Let me tell you something I am not particularly proud of having felt the need to do. I "de-aged" my CV and my LinkedIn profile. Removed certain roles, obscured the graduation year, trimmed the timeline. I did it because, for the first time in a career spanning twenty years, I began to suspect that my experience was being held against me. Not my lack of experience. My experience. I have operated at director level, I've worked on every continent but Antarctica, I've launched new consulting practices on the other side of the world, I have built and scaled developer communities, I predicted the AI world we have moved into now and gained a First Class degree in Computer Science with AI 20 years ago. The idea that any of that should be a liability is absurd.
But the de-aging worked. Well, in so much that interviews followed that had not been following before. Now, I am aware that there was no control in this test, but as Holmes famously noted in The Adventure of the Second Stain, it is an error to argue from data that is not absolute, yet when the coincidences multiply, they become a proof.
So, where are the other coincidences?
On the automated side, I have been rejected without being spoken to for many roles where I met every single criterion in the job specification. I have gone further than most would bother to try to understand where I may have missed the mark. I have tracked down via LinkedIn who actually got roles I was really well suited for, and found profiles of people significantly less experienced than me and significantly younger. One example had a theology degree and their first tech role was in 2021. Another had held several tech roles...about three in just over a year. This one stood out because each of the roles they’d held in that period were roles I had applied for and had not even been spoken to about. Draw your own conclusions about what the systems were optimising for. It clearly was not experience, relevant qualifications or staying power.
On the human side, the picture is no better. In one interview, during what I thought was a genuinely strong conversation where the hiring manager kept it going for longer than the 45 minutes that were booked, I made a single passing reference to Apple's marketing approach. I said something like...
"I firmly believe that we need to look at how Apple promote their content when thinking about B2B selling. They talk to customers and customers make it very clear that they want to use Apple products. I imagine the majority of laptops used by XXXXXXX are Macs"
The rejection I received cited this as the reason I was not moving forward...
"During our conversation, you shared your passion for driving B2C-forward strategies and engaging with broad consumer audiences. While your skills are excellent, our current roadmap for the UKI region remains deeply rooted in enterprise B2B ecosystems"
That was the totality of the feedback. One sentence, constructed from a mischaracterisation of a single passing remark that represented no more than 30 seconds in the hour. It wasn’t even brought up at the time. But I guess when the meetings are fully transcribed by an agent, it is very easy to select a good “reason”.
The hiring manager was around a decade younger than me. The person who got the role graduated in 2019.
Now I want to be measured about what I am saying here. I am not claiming every rejection is age discrimination. I am not claiming that I should be getting every role I apply for and I am not looking for sympathy. What I am doing is trying to apply the same analytical rigour to this problem that I would bring to any other. Gathering data, testing hypotheses, examining the evidence, and following the logic where it leads. The problem is that there is no actual data to work with, we are left with nothing but circumstantial evidence.
What I find is a system that has every technical capability required to provide meaningful feedback to rejected candidates, chooses not to, and benefits directly from that opacity. A system where the absence of explanation is not an oversight but a feature. A system that is, in ways both algorithmic and human, making decisions that would not survive scrutiny, and those behind that system know it.
I am someone who has spent a career solving hard problems, communicating complex ideas clearly, and building things that work. The fact that I am sitting here having had to disguise my own timeline to be taken seriously is not a reflection of my value. It is a reflection of a broken process that the people running it have very little incentive to fix.
Holmes would have spotted this immediately. It is not the feedback that exists that tells you something is wrong. It is the feedback that so easily could exist, in a world of AI-powered everything, but doesn't. That is the dog that isn't barking.
The question worth asking, loudly and repeatedly, is WHY?