Cubicles and Conference Rooms.

I started my career in staffing the way many of us do, by accident.

I was fresh out of school and looking for my first job in the want ads of the newspaper. I circled a few promising advertisements and started sending out resumes. This was the tail end of the era when that was the way it worked. Monster and CareerBuilder existed by then, but they were still curiosities to most of the market, and the newspaper classified section was where most people were actually looking for a job.

One of those circled ads got me a callback. I went in for an interview at a staffing firm and learned they were hiring for two things at once: someone part time in their office in the mornings, and someone part time at a client covering a maternity leave. I took both. That is how I started.

Taking both assignments turned out to matter more than I understood at the time. I saw the experience from the temporary employee's side and from behind the front desk in the same week. I ran the desk and handed out paper applications. Someone would walk in off the street, and I would hand them a stack of forms in a folder. They would sit in the lobby for an hour or two filling everything out, watch the VHS safety video, and then come back to interview.

That process was slow and it was expensive and nobody would design it that way today. It also had an aspect we have spent twenty-five years engineering out of the system: every single applicant talked to a person. There was no way someone would walk into that office and not talk to someone. The process would not allow it.

Learning the work, then learning the system

I moved into a staffing supervisor role, which is where I learned to recruit, interview, and place. That part was genuinely rewarding. Someone would come in looking for a career change or their first job, and the work was figuring out what they could already do and what they could learn next so they could get something better. That is the actual product of this industry, and it is easy to lose sight of when you get far enough from the desk.

I also gravitated toward the technology. We were running a DOS-based program at the time. When the company decided to move to a Windows-based ATS, I was pulled into the group of operations subject matter experts who helped design the transformation and the migration.

That timing was not unusual. Applicant tracking emerged as a real software category in the 1990s, and for most of that decade it meant something installed on your own server, built for the office, not the internet. The move I worked on was part of the industry-wide shift off character-based systems onto client/server, and it happened just before the whole category moved again to the web. Staffing was an early adopter throughout. Time collection, skills assessments, and background checks all became web services in the years that followed, and the front and back office started to converge into single platforms.

It was my first real digital transformation project and I soaked it up. I loved having an opinion about how the technology could make our jobs easier, because I had spent the previous few years doing all of it by hand. I was also fascinated by how configurable these tools were. You could shape them around a specific workflow instead of the other way around.

After we designed it, I got to travel and train people on how to use it. That was the fun part. I knew the product inside and out, not just as a recruiter but as someone who had a seat at the table when it was built.

Twenty years of rollouts

The training did not end there. Over the course of my career we rolled out more initiatives than I can count: technology changes, software implementations, process redesigns, programs to improve the sales process, programs to improve recruiting, programs to upskill our own internal people so they had somewhere to go.

I went from a cubicle to a conference room. Talking strategy, designing workflows, sitting through vendor assessments and deciding which system was actually going to get implemented. It is a fascinating process and I am glad I got to see it up close.

Having come up as an operator turned out to be the differentiator. I could tell whether a tool would actually get used by the people it was built for. That sounds obvious, but it is not. Plenty of well-reviewed software has died in the field because nobody in the selection process had recently done the job the software was supposed to help with.

The same period brought the VMS and MSP models into the center of the industry. Those systems came out of procurement, and the original goal was rate control and visibility into spend. Staffing firms had a complicated relationship with them, for understandable reasons. But they became the operating environment for a large share of contingent labor, and learning to work inside them became the norm.

There were times when decisions got made that did not make everyone happy. That is when the change management work started. Rollout strategy, communication, bringing people along even when they had not fully bought in yet. I have never seen a technology succeed on its merits alone. Adoption is a separate discipline from selection, and treating it as an afterthought is how you end up with an expensive system nobody logs into.

Owning the number

As I moved into leadership and eventually the C-suite, the technology stopped being something I had opinions about and started being something I was accountable for. Not just the tools, but the spend and the operating metrics attached to them.

If my team recommended and implemented something that increased revenue or efficiency, that was a win. If we implemented something people did not use, or that slowed the process down, or that cost more than it returned, we owned that too. That is the part of the job that changes how you evaluate technology. You stop asking whether a demo is impressive and start asking what happens in month nine.

The new technology, the old questions

Which brings me to now.

Staffing operators are facing a genuinely new technology in AI, but I would argue we are not facing new questions. We have always asked ourselves what is new in the market, whether we can adopt it, and whether we will actually benefit. Those are the same three questions I watched people ask about the Windows migration, about job boards, about VMS, about mobile onboarding.

The adoption curve looks familiar too. Bullhorn's 2026 GRID report, which surveyed roughly 2,300 recruitment professionals, found AI adoption among staffing firms at 61 percent, up from 48 percent the year before. But only about 10 percent of firms have AI embedded across their full workflow. Most of what is happening is exactly what you would expect: low-hanging fruit, one function at a time, search and screening first. That is a reasonable place to start. It is also where every prior technology cycle in this industry started, and it is not where the value is.

What concerns me is what is happening on the other side of the funnel while we optimize our side of it.

LinkedIn now takes in roughly 11,000 applications a minute, up about 45 percent in a single year, because generative AI made applying so easy. Robert Half found in March 2026 that 67 percent of HR leaders say reviewing AI-generated applications has slowed their hiring, and 65 percent say the flood makes candidate skills harder to verify. Candidates responded to the volume the way you would expect. Criteria's research, reported in March 2026, found 53 percent of job seekers had been ghosted in the past year, a three-year high, and the share of applicants who got no response at all went from 38 percent in 2024 to 48 percent in 2025.

Both sides are running the arms race and both sides are losing it. Meanwhile ASA's own Workforce Monitor found that 49 percent of employed job seekers believe AI recruiting tools are more biased than human recruiters. Whether or not that belief is accurate in a given case, it is now a market condition we have to operate inside.

The regulators have noticed. New York City's Local Law 144 has been enforceable since July 2023 and it applies to employment agencies, not just direct employers, which means a lot of staffing firms are covered whether they have worked that out or not. Illinois amended its Human Rights Act effective January 1, 2026 to require notice when AI is used in employment decisions and to prohibit ZIP code as a proxy for protected class. Other states are moving, and the specifics keep shifting. The federal picture got quieter when the EEOC pulled its AI hiring technical assistance in early 2025, but Title VII disparate impact liability did not go anywhere. Less guidance, same exposure.

So here is the word of caution, and it is the same one I would have given about any technology in the last twenty years.

AI should never replace the work a human is meant to do. Building a relationship. Sitting with someone in a hard conversation. Coaching an employee who is not performing and does not know why. Those are not inefficiencies to be engineered out. They are the job.

But almost every one of those moments is surrounded by administrative scaffolding, and that scaffolding is fair game. Scheduling. Data entry. Compliance documentation. Status updates that currently do not get sent because nobody has time. If AI gives a recruiter back the hours they are spending on that, and those hours go back into the relationship, that is the win. If those hours get harvested as headcount reduction, we will have built a faster version of the thing candidates already distrust.

The discipline that made the earlier rollouts work is the discipline we need now. Put operators in the design. Own the metric in both directions. Do the change management. Be able to say precisely where the machine acts and where a person decides, and be able to say it to a client, to a candidate, and to a regulator.

I am excited to see where the next generation takes this. They are growing up inside constant change, and the technology is going to keep evolving faster than it ever has. I am glad to be part of it. And honestly? I'm glad we no longer circle want ads in the newspaper.

Sources