Companion piece to Constitutional Guardian article - for LinkedIn engagement
By Stephen Hope — edited with AI
I started my job search the way most experienced executives do.
A resume built on 35 years of actual outcomes. No keyword stuffing. No optimization games. Just a clear record of what I had built, fixed, and delivered.
The rejections came fast. Automated. Impersonal.
So I did what the market told me to do. I used AI to analyze fit between my experience and job postings. I subscribed to tools that promised to improve my ATS score. My resume grew from 2 pages to nearly 6.
And somewhere in that process, I stopped recognizing myself in it.
The accomplishments that defined my career — the board-level obligations I delivered without board access, the systems I rebuilt under impossible constraints, the teams I led through transformation — got buried under a list of processes and tasks designed not to represent me, but to satisfy an algorithm.
That is not a personal failure. That is a governance failure. And it was entirely predictable.
In physics, every action produces an equal and opposite reaction.
When organizations deployed ATS systems to automate candidate screening, they created pressure on the candidate side of the pipeline. That pressure produced a predictable and proportional response: an entire industry of tools, services, and AI applications designed to game the screening system. Keyword optimization subscriptions. AI-generated resumes and cover letters. Automated mass applications with near-perfect ATS scores.
This was not a corruption of the process. It was the process responding to the conditions that were created for it.
The failure was never the automation itself. The failure was the absence of any governance structure capable of managing what the automation would inevitably produce. Organizations automated a consequential human process, removed the human authority layer that gave it integrity, and then expressed surprise when the system began optimizing for the wrong outcomes.
You cannot automate both ends of a human process and expect the purpose to survive.
Here is what is actually happening inside most hiring pipelines right now.
On the employer side: Applicant Tracking Systems filter thousands of inbound applications down to a manageable shortlist. The assumption is that the algorithm surfaces the most qualified candidates. The reality, confirmed by HR professionals and technical recruiters I spoke with directly, is the opposite. Things are going slower, not faster. Organizations are receiving hundreds to thousands of nearly identical, algorithmically polished resumes. HR teams can realistically review only a small fraction of even the filtered pool.
On the candidate side: A cottage industry of resume optimization services has emerged in direct response — subscription tools that rewrite your experience to pass the scanner. AI tools now submit hundreds of applications simultaneously with near-perfect, keyword-saturated documents. Candidates are not gaming the system out of dishonesty. They are responding rationally to the incentives the system created.
The result: AI-generated applications are overwhelming AI-powered filters. The signal — actual human capability, judgment, and career-long demonstrated outcomes — is being destroyed on both ends simultaneously.
Everyone loses. And the failure gets attributed to everything except the governance decision that caused it.
ATS systems were designed to screen out unqualified candidates.
What they are actually screening out is candidates whose careers are built on outcomes rather than keywords.
A 35-year executive career does not read like a job description. It reads like a series of complex problems solved, organizations transformed, and decisions made under pressure with incomplete information. That does not score well against a keyword matrix. It does not fit neatly into a 6-second ATS parse.
So those candidates either hollow out their professional story to fit the machine — as I did, briefly — or they disappear from the pipeline entirely.
Meanwhile, candidates who have learned to optimize for the algorithm move forward. Not because they are more qualified. Because they are more legible to a system that cannot tell the difference.
The pipeline is not broken. It is doing exactly what it was built to do. The problem is that what it was built to do is no longer aligned with what the organization actually needs.
From July 1999 through November 2025, only six months of my employment and job transitions did not involve a recruiter.
That is not luck. That is a model that worked.
Recruiters — the ones who worked both sides of the relationship — understood something that an algorithm cannot. They knew the candidate. They knew the organization. They knew the difference between a resume that looked right and a person who was right. They could advocate for a candidate who did not fit the template because they had context the template could not capture.
That was not a soft human touch added on top of the process. That was the governance layer. The human authority structure that gave the pipeline its judgment, its integrity, and its ability to produce outcomes that matched organizational need to human capability.
That layer has been systematically minimized in the name of efficiency.
What replaced it is volume without judgment. Speed without signal. A process that feels modern and produces outcomes that are measurably worse. New hire failure rates have not improved with ATS adoption. By most measures they have gotten worse. The automation did not solve the matching problem. It removed the layer responsible for solving it and called that progress.
Robert A. McTaggart, in his field manual HITL: The Human Layer, makes an argument that is precise and uncomfortable in equal measure: a human near a system is not oversight. Sampling is not supervision. Proximity is not governance.
His framework defines Human-in-the-Loop not as a person reviewing outputs, but as a structured authority layer — with defined roles, escalation thresholds, intervention power, documentation discipline, and institutional backing — capable of stopping a consequential decision before harm occurs.
The phrase "human-in-the-loop," McTaggart argues, has become a reassurance that masks a dangerous ambiguity. Organizations say a human reviews the process. They rarely ask whether that human has the authority, the context, or the structural independence to actually change the outcome.
The recruitment pipeline is a case study in exactly what he warns against.
ATS systems are making consequential decisions about people's economic futures — decisions that McTaggart would classify as Tier 4: rights-impacting. Yet the human oversight layer has been reduced to an HR reviewer skimming the top fraction of an algorithmically filtered list, under volume pressure, with no structured escalation authority and no accountability mechanism when the system selects the wrong person or rejects the right one.
That is not Human-in-the-Loop. That is a human beside the loop, occasionally glancing at it.
McTaggart identifies five requirements for real oversight:
- Authority to intervene
- Defined escalation thresholds
- Structured documentation
- Institutional backing
- Feedback into system correction
Strip away any one of these and oversight becomes symbolic. The modern ATS pipeline has none of them. It has a dashboard and a shortlist — and calls that governance.
The skilled recruiter was functioning as exactly the structured human authority layer McTaggart describes. Not a passive monitor. An active gate. One with context, judgment, and the standing to say: this person is worth a conversation the data cannot capture.
That gate has been removed. And the organizations that removed it are measuring the cost in failed hires, extended vacancies, and capability gaps they cannot explain — attributing none of it to the governance decision that started the chain.
Here is what makes this more than a hiring problem.
ATS systems are making decisions that affect millions of people's livelihoods, careers, and economic mobility. They operate at scale. They operate continuously. And they operate with no governance framework, no escalation architecture, no transparency requirements, and no accountability when they fail.
The EU AI Act classifies employment-related AI systems as high-risk by definition, requiring human oversight and transparency. Most organizations deploying ATS technology in the United States are operating under no equivalent obligation. The human layer was cut for efficiency. The governance framework was never built at all.
This is not a technology problem. It is a leadership problem. Executives approved the automation. Executives set the efficiency metrics. Executives accepted the reassurance that a human reviews the shortlist — without asking whether that review constitutes real oversight or just proximity to a process that has already made its decisions.
McTaggart's warning applies directly: "Reassurance is not governance. Oversight is not proximity."
If you are a CEO, CHRO, or board member reading this, here is the question worth sitting with:
Your hiring process is a technology system making rights-impacting decisions about human beings. It is making those decisions at scale, automatically, with a human nearby but not structurally in control.
- When did you last ask whether your ATS is identifying the most qualified candidates — not the most optimized documents?
- When did you last measure new hire performance against ATS match scores to see whether the correlation holds?
- When did you last ask your HR team not whether the pipeline is moving fast, but whether it is producing the outcomes it was intended to produce?
If your ATS is filtering out experienced leaders who built careers on delivery, while surfacing candidates whose primary skill is optimizing a document for an algorithm — that is not a pipeline problem. That is a governance problem. And the cost shows up later, quietly, when the hire that looked perfect on paper cannot deliver in the room.
The most effective hiring process I ever participated in had three elements: a clear definition of what the role actually required, a skilled recruiter who understood both sides of the relationship, and a human being willing to have a real conversation before making a decision.
That is not inefficient. That is just hard to automate — and harder to walk back once you have built a process that replaced it.
AI has a place in the hiring pipeline. But that place requires governance — structured human authority with the standing, the context, and the accountability to ensure the process serves its original purpose.
The question is not whether AI belongs in your hiring pipeline.
The question is whether you have built the governance structure that keeps a human actually in control of it.
Most organizations have not. And the talent they are losing because of it will not show up in any dashboard they are currently watching.
Stephen Hope | Managing Principal, Rusted Gate Advisory
Rusted Gate Advisory works with organizations navigating the gap between where they are and where they need to be — in technology, leadership, and organizational design. Sometimes the gate is rusted shut for a reason. Sometimes it just needs the right key.
Robert A. McTaggart's field manual HITL: The Human Layer is a first primer on Human-in-the-Loop leadership in the agentic age. It is recommended reading for any executive deploying AI in consequential decision systems. https://trustedbyheroes.com/
Primary Post: Constitutional Guardian technical article (LINKEDIN_ARTICLE.md)
This Comment/Thread: Post as a follow-up or standalone piece about governance in hiring systems
Hashtags: #Governance #AIethics #Hiring #ATS #HumanInTheLoop #Leadership #TechEthics #AIAct #EUAIAct #TalentAcquisition
Connection to Guardian: Both pieces share the theme: automation without governance produces predictable failures