Most AI in hiring is still a button you press. That's changing: orchestrated, proactive work that shows up inside the assistants your team already uses, all standing on a system of record that keeps it accountable.
Most AI in hiring still works the same way the rest of your software does.
You go to a screen, you find the feature, and you press the button. Summarize this interview. Score these resumes. Draft that rejection email.
Each one is a small, useful thing you ask for, one at a time, by hand. Sure, it's better than doing the work yourself, but it's still you doing the asking.
However, that's about to change, and the change is bigger than just anotherfeature.
The Josh Bersin Company put it well in its recent work on where talent acquisition is heading. It describes a function that has been fundamentally reactive, waiting for a requisition, waiting for an application, waiting for someone to click, becoming a proactive force, with AI working in the background and “digital twins” handling routine work so people don't have to. That's the shift I want to talk about in this article. Not smarter buttons, but the work that happens.
When I say 'work that happens', I don't mean the software runs off and makes hiring decisions while you're at lunch. I mean the difference between a tool that waits to be told and a system that already knows what needs doing, surfaces it, and gets the routine parts moving, with you deciding what matters. The first is a feature. The second is closer to a colleague.
And in my opionion, three things are moving at once to get us there:
The first shift is from features to orchestration.
Today, the AI in a hiring system tends to be a drawer full of separate tools. One summarizes. One scores. One drafts. Each is a single step you trigger and then carry the output of to the next step yourself. You're the orchestration layer, ferrying things between buttons.
Orchestrated work strings those steps together. “Find me the strongest candidates for this role, draft the outreach, and get the first-round interviews scheduled” is not one feature. It's a sequence of them, coordinated, with the system carrying its own output from one step to the next instead of handing it back to you each time.
The individual capabilities barely change. Instead, what changes is that they stop being isolated buttons and start being a job of work the system can run end to end.
The second shift is from reactive to proactive. A reactive system sits still until you go and ask it something. A proactive one watches what's happening and tells you what needs attention without being asked.
This role has three candidates stalled at offer stage and nobody has chased them. That hiring manager hasn't left feedback on four people who interviewed last week, and the pipeline is jamming behind them. This requisition is pulling almost no qualified applicants and probably needs rewriting.
None of that requires you to remember to look. The system already knows, because it can see the whole process, and it brings the thing forward so you can act on it. The work of noticing, which is most of what eats a recruiter's day, becomes something the system does and you direct.
The third shift is about where this work lives. Right now, AI in hiring mostly lives inside the hiring system. You go to the ATS to use it. That's fine, but it isn't where most people spend their day.
Increasingly, the same work becomes available through the assistants your organization already runs, Copilot, Claude, Gemini, the ones sitting inside Teams and Slack and the tools your team lives in. A hiring manager could ask their company's own assistant, “where are we with the two candidates I interviewed Tuesday,” and get a real answer, because that assistant can reach into the hiring system and pull it. The recruiter doesn't have to be the go-between. The work comes to where people already are, instead of making them come to it.
This is the part that draws the most excited, and the loudest, predictions. So it's worth being clear about what it does and doesn't mean.
There's a popular version of this story that goes much further. It says that once your company has a capable enough assistant, the assistant simply becomes the software. Ask it to handle hiring and it handles hiring.The agent is everything, and the specialist systems underneath, yes, the ATS included, quietly disappear because you don't need them anymore.
Unsurprisingly, I don't agree.
In my previous article, I highlighted one of Chermaine Dufosse's LinkedIn posts, which reads "Real agents don’t replace RPA, they sit on top of it."
And that's exactly how I believe this will play out.
An agent is brilliant at reasoning, deciding, and adapting, but it isn't the right place to keep the things that have to be reliable, permanent, and accountable. And hiring is full of those things.
Candidate data, defensible hiring, the whole structured record of every application, every stage, every note, that you'll need if a hiring decision is ever challenged. Not to mention the set up workflow that guarantees every candidate actually gets a reply rather than usually getting one.
None of that should live in a conversation. It has to live in a system built to hold it: with the permissions, the audit trail, the structured data, and the workflow.
And that system is the ATS. The agent is the interface you reason and act through. The ATS is the layer underneath where the work lands and stays accountable. Make the agent the whole building and you've moved your most consequential, most regulated data into the one place that was never designed to keep it safe.
As Chermaine says: real agents don't replace the system of record. They sit on top of it, and they're more useful for having something solid to stand on.
Of course, the system of record holds more than the permissions and the audit trail. It also holds a person in charge, and that matters just as much.
Every output from the AI arrives as a suggestion for a person to act on. It surfaces strong candidates and the recruiter decides who moves forward; it drafts the outreach and a person chooses whether to send it. The Josh Bersin Company puts recruiter oversight at the very top of its model, above the automation and the AI, and that's where I'd put it too.
You can't switch that off. When an agent works on top of a system of record, every action it takes runs through its permissions and lands in its audit trail. Keeping a person in control and keeping an accountable record turn out to be the same thing, so build for one and you get the other.
Some of this is already real. We ship an MCP server today, the piece that lets an outside assistant reach into Pinpoint and do real work, securely and within the same permissions a person would have. So the third shift, the work showing up inside the assistants your team already uses, isn't a someday idea. The plumbing exists now.
The rest is direction, and I'd rather be honest about that than dress a roadmap up as a promise. Where this is heading is a world where you can say "find me the strongest candidates for this role and schedule the interviews" once, and have it happen, whether you say it inside the hiring system or through your company's own assistant calling out to it.
The steps run in sequence, carried from one to the next without you ferrying the output between them. The system surfaces the stalled offer or the missing feedback before you'd have thought to check, and it does that work wherever your team already spends its day. Underneath all of it sits the record that holds the permissions and the audit trail, with a person deciding the things a person should decide.
That's the future I find genuinely exciting, and it's a more grounded one than the version where the software vanishes into a chatbot. When a recruiter asks their company's agent to do hiring work, that work shouldn't disappear into a thread, or into an email and a spreadsheet nobody can audit later. It should land inside the system that holds the permissions and keeps the record straight.
The agent is how you ask. The system of record is where the work lands and stays accountable, with the person who matters still making the call. This is the job the ATS was always for, and it's a bigger one than the software has had until now, finally with something worth standing on.
Tom is the CEO at Pinpoint, he's passionate about building world-class teams and world-class products for organizations around the world.
