AI Recruiting Agents: Inside Adecco's Ada

The Adecco Group processes around 300 million job applications a year. Denis Machuel, its CEO, sat down with The AI Report at Dreamforce 2026 to talk through what that looks like now that AI agents have entered the process.

Denis has led Adecco since 2022, overseeing a company of roughly 40,000 employees that places people in work across the globe every single day. What he described in this conversation is not a company experimenting with AI on the edges. It is a company that has deployed it at the core of how it operates, and is learning in real time what that changes.

Meet Ada

When Adecco launched Ada, its AI recruitment agent, the early data threw up a result Denis did not expect.

Fifty percent of Ada's candidate conversations took place outside of working hours. Late evenings, early mornings, weekends. For candidates who were working during the day, it meant they could engage with a job search on their own schedule, in the evening, whenever it was convenient. Denis noted that in doing so, Adecco was providing a kind of flexibility the traditional recruiter model simply could not.

Ada started in Birmingham and Manchester in the UK before expanding further. Denis was careful about the rollout, partly because the stakes were high. A candidate's experience with an AI agent during a job search is not a neutral interaction. It carries real emotional weight, and getting it wrong has consequences. What surprised him was not just the engagement but the outcome: candidate satisfaction scores after conversations with Ada came back high. That, Denis said, was the genuine positive surprise.

Thursday Morning, Then and Now

Denis used a specific image to describe how the recruiter's day has changed.

Eighteen months ago, a recruiter would arrive on a Thursday morning to a request from a client: twenty profiles, specific skills, two weeks. To fill that order, they would spend most of the day making outreach calls (hundreds of them), sending emails, leaving messages, waiting for callbacks from people who were unavailable during working hours. Most of that time was logistics, not the actual work of matching people to roles.

Today, the agents handle that outreach over three to five days before the recruiter even arrives. Those hundreds of calls happen while the recruiter is doing other things, such as having more qualitative conversations with candidates already in the pipeline. By Thursday morning, there is a shortlist of pre-screened candidates ready for a real conversation. The recruiter's job is no longer to make the calls. It is to have the conversations that matter.

The recruiter now becomes the person who handles the human-to-human part of the process: understanding motivation, reading context, making the judgment calls an agent cannot make well. And, as Denis put it, recruiters are enjoying their work more because they are doing more of what they actually like to do.

From Reactive to Proactive

Right now, staffing companies are largely reactive. A client calls and says they need someone with a specific skill set. The recruiter goes out to find that person. Adecco has been in that model for decades.

What Denis wants is something different: a version of Adecco that does not wait to be asked. One that proactively advises candidates on where their career trajectory is heading, what skills they should be building, and what opportunities are likely to open up before those candidates have thought to look. And one that advises clients on how the workforce around them is shifting, before the gap becomes a problem.

That requires a level of data, pattern recognition, and always-on engagement that was not possible without AI. For Denis, that is what the whole shift adds up to: a reinvention of the business, from a company that responds to needs to one that anticipates them, for both candidates and clients.

27,000 People with an AI Coworker

At Dreamforce, Denis announced the next stage of Adecco's partnership with Salesforce: deploying Agentforce Coworker to 27,000 of his staff. Before this, Denis described the pattern most large organizations will recognize: employees using personal ChatGPT accounts for work, informal pilots running in pockets of the business, no clear line between what was sanctioned and what was not. His view was that the DIY phase had run its course. The step from individual experimentation to deploying agents across 27,000 people is where AI stops being something employees dabble with and starts being how the business actually operates.

Denis sat on the Dreamforce stage alongside Marc Benioff and Roland Busch, the CEO of Siemens, as part of the announcement. The conversation there was about what large enterprises need to do to make AI work at scale: governance, integration, and the kind of change management that gets 27,000 people actually using a new tool rather than just having access to it.

The Human-in-the-Loop Question

Denis is clear that AI does not replace human judgment at the moments that matter most.

His view on agent orchestration, which he sees as the next challenge for companies at his scale, is that the design question is not how to automate more. It is where humans add the most value, and how to make sure they are in the loop at exactly those moments. Getting that right requires knowing your processes well enough to identify where those moments are. Most organizations, he thinks, are still figuring that out.

For Adecco, the answer comes back to that Thursday morning. The agents have done the outreach, screened the candidates, and built the shortlist. What is left is the conversation that actually matters: the one where someone decides whether to take a job, change careers, or trust a company with something as important as their livelihood. That conversation, Denis says, is not something he is handing to an agent any time soon. It is exactly the kind of moment the whole system is designed to make room for.

Watch the full conversation with Denis Machuel on The AI Report podcast.

Stay up to date with the latest AI news and analysis. Subscribe to The AI Report.

‍