Who Recruits the Machines?

For the entire history of recruiting, one assumption has gone almost completely unquestioned: talent is human.

That assumption is becoming optional.

Agentic AI is creating the possibility of digital workers that do not simply help an employee perform a job, but can increasingly accept responsibility for portions of the job itself. If that continues, companies will eventually stop automatically translating every capability requirement into human headcount.

They will start one step earlier by asking:

• What outcome are we trying to achieve?
• What capabilities are required to produce it?
• What is the best available source of those capabilities?

Sometimes the answer will be a person. Sometimes it may be a digital worker. Increasingly, it could be a combination of both.

That creates a problem much larger than recruiting technology. If a digital worker performs the daily work of an employee but is purchased like software, who owns the decision to acquire it, manage it and evaluate it? Talent Acquisition? Procurement? IT? The business itself?

For decades, manufacturing has already dealt with a earlier version of this problem on the shop floor. Companies learned to divide physical work between humans and machines according to what each could do best. Agentic AI could bring that same economic logic to cognitive work in the front office.

And if it does, that leaves Talent Acquisition with a choice about what business it is actually in:

Today: Find people for jobs. Tomorrow: Source capability for outcomes.

The Battle for Sourcing Outcome Capacity

If a business leader says, “I need more electrical engineering capacity,” who decides whether the answer is a human engineer, a digital worker, a human engineer managing ten digital workers, traditional enterprise software, or an outsourced firm operating an entirely digital workforce?

Today, corporate boundaries are neatly drawn. Talent Acquisition owns the human option. Procurement buys software and vendors. IT governs technology and infrastructure. Business leaders define operational needs. HR thinks about organizational design.

Digital labor completely blurs those boundaries.

If a digital worker is purchased like software, perhaps procurement owns it. If it requires API access to enterprise systems, perhaps IT owns it. If it performs the daily work of an employee, perhaps the business function owns it directly.

But if you are evaluating its specific skills, operating history, reliability, performance, cost, collaboration, autonomy, and fit with the existing workforce, suddenly that looks a hell of a lot like workforce planning.

Traditionally, workforce planning determined how many people a business needed, and Talent Acquisition went out and found them. But what happens when the answer isn't necessarily a person?

There is an uncomfortable possibility buried inside this shift.

Talent Acquisition could remain exclusively responsible for human labor. Workforce planners, business leaders, IT, and Procurement would determine the optimal architecture first, evaluating automation, digital workers, outside providers, and human labor against the required outcome.

TA would enter the process only when that analysis concluded that a human being was still required. The function would survive, but it would move downstream, becoming the sourcing option used after other forms of capacity had already been considered. TA would effectively drop to the bottom of the corporate sourcing hierarchy.

The alternative is much more consequential. Talent Acquisition could move upstream and evolve from sourcing people into sourcing capability. That would require participating in workforce architecture itself, helping determine not simply who should we hire? but what combination of human and digital capacity gives the business the best outcome? If TA doesn't claim this territory, another function will emerge to design the increasingly important part of the company that isn't human. That is both the threat and the opportunity.

Cost will inevitably influence these decisions. But an interesting question emerges when digital labor cannot perform a particular job as well as a person can. What is that human output worth then? If a capability is difficult to automate, difficult to find in the labor market, and critical to the business, the economics of that person may change dramatically.

Companies may employ fewer people to perform certain types of work, while paying considerably more for the people whose judgment, expertise, relationships, or decision making cannot be easily replicated. In that environment, the value of human labor will not simply be determined by what a person can produce. It will increasingly be determined by how scarce that uniquely human capability is, how important it is to the outcome, and how effectively that person can leverage the digital capacity around them.

Intelligence Was Never the Product

To understand why this could become a Talent Acquisition responsibility rather than simply an IT or procurement decision, you have to look past the technical hype and ask what companies actually buy when they hire.

A lot of the conversation about technology centers on the sudden abundance of intelligence. I think that overstates what is actually new here. Access to knowledge has been getting cheaper for centuries through libraries, search engines, and early generative AI. But there is a massive psychological and economic difference between knowing how to do something and actually doing it.

For instance, I can watch a YouTube video that shows me exactly how to replace my car's alternator. I can ask a large language model which part to buy. But I still don't want to change my alternator. I want someone else to change it for me.

Companies operate the exact same way. They don’t hire an accountant because that person knows accounting. They hire them to close the books correctly. Knowledge simply enables the outcome. The outcome itself is where the economic value lives.

That is why the real turning point for the workforce isn’t that machines can answer difficult questions via a chat prompt. The real shift happens when agentic AI allows systems to reliably execute multi step goals.

A generative model that can explain electrical engineering is just a tool. But an autonomous AI agent operating as a digital worker that can take an assignment on Monday, log into company systems, collaborate with the team, recognize its own limitations, escalate complex issues, and hand over a flawless finished product on Friday? That is a viable candidate for a job. At that point, we are no longer talking about software procurement. We are talking about labor.

Sourcing for Capability and Compatibility

This is where the shop floor serves as our exact blueprint. When a manufacturer automates an assembly line, they do not build a robot from scratch, nor do they download generic code and hope for the best. They turn to systems integrators and specialized automation fleet providers.

The future of front office digital labor will mirror this B2B staffing model. Anyone may eventually have access to roughly the same underlying intelligence. The scarce capability is not the foundational model itself. It is training a digital worker to reliably perform a specific occupation. Turning raw intelligence into a digital electrical estimator that understands complex drawings, operates specialized software, follows specific corporate workflows, and recognizes bad input requires extensive training infrastructure, domain expertise, and rigorous evaluation.

I expect a new supplier category to emerge around this need: Agent Fleet Providers. These organizations will act as specialized digital staffing firms, building, fine tuning, and continuously maintaining fleets of digital workers engineered for highly specific occupational classes.

Consequently, the recruiter’s evaluation of a digital candidate must split into two distinct categories: Capability and Compatibility.

Capability is the objective, technical fit. On the shop floor, you evaluate a robotic arm by its payload capacity and speed. In the front office, vetting a digital worker's capability means auditing its operational specifications:

• Token efficiency and compute costs to determine financial overhead.
• Documented error and correction rates across thousands of past assignments.
• Software competence to ensure native integrations with legacy enterprise tools like SAP or Salesforce.

Compatibility is the operational fit. In manufacturing, you cannot simply buy the fastest robot arm and bolt it to the floor. It has to safely mesh with the human environment. You must test its sensors, its emergency stops, and how easily a human supervisor can override it.

When recruiting digital labor for the front office, compatibility becomes an organizational safety check centered on Outcome Autonomy: how much human involvement is required for the digital worker to produce an acceptable outcome?

Imagine two digital engineers turning in similar technical work. The first requires a human engineer to review 30 percent of its decisions. The second only requires human involvement 2 percent of the time, and more importantly, it reliably flags exactly which 2 percent needs escalation. Those two workers are not economically equal. The second hasn't just replaced an engineer; it has multiplied them, introducing a new evaluation metric: the Human Leverage Ratio.

Beyond autonomy, recruiters must vet digital candidates on collaborative organizational fit:

• The Communication Burden: Can it take direction from a human manager using messy, ambiguous language, or does it require rigid prompt engineering that creates an extra task for the team?
• The Black Box Dilemma: Can it explain why it made a specific decision when challenged, or does it demand blind trust from human executives?
• The Failure Boundary: Does it know when it doesn't know, flagging boundaries safely before making a catastrophic, undetected mistake?

Reliability, tool competence, security, collaboration, and Outcome Autonomy will all become standard candidate characteristics. Recruiters will not just be interviewing an individual digital worker; they will be evaluating the infrastructure of the fleet provider backing them up.

Telemetry Is the New Work History

But you cannot claim a digital worker requires only 2 percent human intervention without evidence. If companies are going to evaluate digital workers the way they evaluate human candidates, they will need a reliable record of what those workers have actually done and how well they performed. For digital labor, that record will come from performance telemetry.

Telemetry would track the assignments a digital worker completed, where humans intervened, what errors were corrected, when it escalated, what tools it used, how long the work took, and whether the finished outcome held up over time.

Telemetry becomes the digital equivalent of a work history. It turns claims about capability into an auditable performance record. Even if digital workers eventually become capable of improving themselves, companies will still need an independent way to determine whether that improvement actually produces better outcomes. Otherwise, the system is effectively grading its own homework.

The Unbundled Candidate and Experience Custody

This data layer reveals an even bigger structural transformation: the digital labor market is about to separate intelligence, training, experience, and employment into entirely different economic layers.

Human talent has always arrived as a package. Intelligence, training, experience, judgment, and work history all come attached to the same person. Digital talent may not.

For digital labor, those pieces could potentially be unbundled and fractured completely. One provider supplies the underlying foundational intelligence. An Agent Fleet Provider trains the occupational capability. A third party auditing system maintains the verified performance telemetry. The client contributes proprietary organizational experience.

Different entities could potentially own different layers of what would all be inseparable in a human candidate. This unbundling completely supercharges the problem of Experience Custody.

If a digital worker spends five years executing projects inside Company A, who owns what it learned? The question is no longer merely who owns the software. The question is: Who owns the underlying intelligence? Who owns the occupational training? Who owns the accumulated workflow experience? Who owns the verified telemetry log? And who gets to take those things somewhere else?

Can a digital staffing firm make 1,000 copies of an agent whose expertise was partially developed on a client’s proprietary engineering systems? Can Company A prevent its accumulated institutional knowledge from becoming part of a digital worker subsequently deployed directly to a competitor?

Human employment law and intellectual property law were built on one convenient assumption: the worker and the worker's experience are inseparable. Digital labor completely smashes that assumption, turning experience into a reproducible, legally contested corporate asset.

Sourcing the Humans Who Leverage Digital Labor

Automation on the shop floor didn't just change the equipment companies used. It also changed what made certain workers valuable. The people who could operate, program, troubleshoot, and supervise automated machinery could produce far more than they could through their own labor alone.

The same dynamic could emerge in the front office. If Talent Acquisition takes a role in sourcing digital labor, it will also need to identify the people who are particularly good at directing it.

The most valuable engineer may not be the one who can personally produce the most drawings. It may be the engineer who can direct the work of multiple digital engineers, recognize when their output is wrong, and know when human judgment is required. A project manager may increasingly be evaluated on the ability to coordinate a mixed workforce of people and digital workers.

This changes the economics of human talent as well. Digital workers will become more valuable as they require less human intervention, while certain human workers will become more valuable because of how effectively they can direct digital capacity. The question will no longer be only what a person can produce individually, but how much productive capacity that person can responsibly manage.

The Automated Displacement You Can't See

This shift also changes how we must view job displacement. Most discussions on the topic focus on layoffs at existing companies. That will happen. But the most consequential scenario involves the companies that don't even exist yet.

Consider a future startup. The founders will start fresh with the exact outcome they want to produce, map out the required skills, and source the most effective delivery method for each. They won't automatically build a 50 person accounting department just because companies traditionally have them. Instead, they will simply source accounting capacity. Some will come from people, some from specialized digital workers, and some from standard automation.

And that creates a form of displacement that will be much harder to see than standard layoffs. Nobody gets fired from the 500 person company that never employs 500 people in the first place. There is no layoff announcement. The jobs simply never exist.

Instead, you will see a company run by 50 exceptionally capable human leaders directing a massive digital workforce. This happens because each human became capable of directing vastly more productive capacity.

So Who Recruits the Machines?

I've been in recruiting for more than 20 years. I've watched newspapers, job boards, Applicant Tracking Systems, and LinkedIn reshape how we find human candidates. But all of those past shifts were ultimately about technology changing the way we recruit human beings.

This is fundamentally different.

We already have the blueprint for mixed workforces because we’ve been building them on the shop floor for decades. What changes now is that the economic logic of physical production is reaching the front office.

If digital systems become capable of accepting real responsibility for cognitive work and reliably delivering outcomes, companies will have to figure out which skills they need, where to acquire them, and whether the best source is human, digital, or a mix of both.

Maybe recruiters will recruit digital workers. Maybe they won't. Maybe procurement or IT will wrestle control of the workforce away from HR entirely.

But the recruiter's job has always come down to a fundamental question: Where do we find the talent needed to produce the desired outcome? Until now, the answer has always been people. That is what is changing.

Talent Acquisition may not ultimately own digital labor, but it should have a seat at the table when companies decide how work gets done. And perhaps TA has an advantage right now. The organizational lines have not yet been drawn, which gives the function time to develop the expertise and help define where the acquisition of digital labor belongs within the organization.