A recruiter’s Tuesday starts with an intake call for a new vacancy. By ten there are three candidate emails waiting for a reply, a hiring manager who still owes feedback on last week’s shortlist, and a new employee’s first day on Thursday for which the welcome pack has not been updated since the last one. None of this is strictly hiring; it’s the writing, sorting, chasing and formatting that sits around hiring, and it takes up the hours that should instead be going to the conversations that actually fill the role.
That’s the work an AI agent can prepare. This post looks at what HR teams and recruitment agencies in Belgium are already doing with AI, what an agent adds on top of a chat window, five agents worth building, and the one thing an agent should never do.
What’s happening today
AI is no longer an experiment in Belgian HR. According to LinkedIn’s 2025 report on the future of recruiting in Benelux, three in four talent acquisition professionals in the Benelux agree that AI will change how organisations recruit, and across Europe the share of recruitment teams actively integrating or experimenting with generative AI rose from 26 to 37 percent in a single year. Across the wider economy, the FOD Economie’s Belgian Digital Economy Overview 2026 puts the share of Belgian enterprises using at least one AI technology at 34.5 percent, well ahead of the European average.
What recruiters use it for is consistent. Asked what AI brings them, European talent acquisition professionals name faster recruitment, a shorter time to hire, better vacancy texts, handling a higher volume of applications and a wider talent pool. The average time saved comes to roughly 16 percent of the working week, close to a full day (LinkedIn, Future of Recruiting 2025). Inside HR departments more broadly, adoption is slower than in IT, customer service or marketing, according to the Vlerick Business School HR Barometer. 58 percent of HR directors at large Belgian employers say AI adoption in their organisation is too slow, and the obstacles they name first are a lack of AI expertise inside HR, limited budget and time, and the quality of their HR data.
Speed, in other words, is the easy part. Whether the output is any good is a different question, and the evidence is mixed. Research from MIT on AI-written vacancy texts found that employers wrote them 44 percent faster and were 19 percent more likely to actually post, but those vacancies had a 15 percent lower chance of leading to a hire. A UGent study found that when ChatGPT was asked to assess CVs, candidates with a non-Belgian-sounding name were rated favourably 14 to 19 percent less often. Neither finding is an argument against AI in hiring. Both are arguments for keeping a person between the tool and the outcome, which is the thread that runs through everything below.
Most of today’s usage is also one-off. A recruiter opens a chat window, explains the company, the role and the tone, gets a draft, closes the window, and does the same again for the next vacancy. That’s where an agent is different. 
A chat window versus an agent
A chat window starts from nothing every time. An agent is set up once, for one job. It holds its own instructions: the tone your company writes in, the rules a rejection email must follow, the structure every candidate summary should have. It holds its own context: your employee handbook, your intake template, a good example of your onboarding checklist. And it keeps its own conversation, so three weeks later it still knows which vacancy you mean.
An agent can also be connected to the systems you already work in. Link it to Outlook or Gmail and it can read the candidate’s email, draft the reply and, once you have read it, send it. Link it to Salesforce and it can look up a client account before you call. Drop a CV, an intake form or an Excel sheet into the chat and it works from the document in front of it.
Then it waits. Every email, every record update, every document is shown to you as a draft first. Nothing goes out without a person confirming. Setting up a first agent from a template takes around ten minutes. 
The other side of the table is using AI too
Every recruiter who has opened an inbox this year has noticed it: applications arrive faster, in greater numbers, and they read more alike than they used to. Candidates use the same tools recruiters do. They paste the vacancy into a chat window and ask for a cover letter. They have their CV rewritten to match the keywords. Some run browser extensions that apply to dozens of vacancies in an afternoon without reading any of them.
The scale is smaller than employers think. 2026 research by Partena Professional with UGent labour economist Stijn Baert found that 53 percent of Belgian employers believe at least half of applicants use AI, and three in ten put it at 70 percent or more. Among employees themselves, one in five said they had actually used AI when applying. The gap matters because it shapes how applications are read. Nearly half of Belgian employers (48 percent) say AI has made it harder to judge how honest and motivated an applicant is, and 45 percent find it harder to tell a candidate who genuinely wants the job from one applying to keep their benefits. Interestingly, the people who do selection every day are more relaxed about it than the managers above them: 40 percent of managers and business owners say AI applications mean more work and a slower process, against 19 percent of HR professionals. The likely reason, according to Partena, is that recruiters have always checked a CV against the conversation, and look less at the writing style of a letter than at whether experience, skills, motivation and behaviour hang together.
So the picture is not AI against AI, with two machines trading polished documents while the humans wait. It is that the written application has become a weaker signal, and the conversation a stronger one. That’s why we advise building in deliberate human contact early, such as a short phone call or a video message, and interviewing around concrete situations rather than prepared answers.
That has a direct consequence for where an agent belongs. The volume problem is real, and an agent is well suited to the part of it that is about handling: structuring what comes in, replying to everyone promptly, keeping the process moving. The judgement problem is also real, and an agent doesn’t solve it. What it does is give the recruiter back the hours to have the conversation that does.
Five agents worth building
The vacancy intake agent
For agencies and in-house teams alike, a vacancy rarely arrives as a brief. It arrives as a phone call, an email thread and some notes from the hiring manager or client. Every recruiter then structures it differently, and sourcing often starts before the important questions have been asked.
Give the agent the raw notes and it returns one fixed intake structure: the role, must-haves, nice-to-haves, team context, urgency and known risks. It then lists what is still missing, such as salary range, contract type, languages or reporting line, so nobody starts searching on half a brief. From the same intake it drafts the vacancy text, the screening points for a first call and the internal briefing. The recruiter fills the gaps, adds what the notes could not capture about the team and the culture, and signs off.
The candidate communication agent
Invitations, rejections, status updates and offer follow-ups are the highest-volume writing in any hiring process, and the writing with the clearest rules. Those rules are exactly what an agent holds well: always mention the retention period, never give a rejection reason beyond what policy allows, always confirm the interview location.
Connected to Outlook or Gmail, the agent reads the incoming candidate email, drafts the reply in the company’s tone and shows it to the recruiter. Because it is working from the actual message and the actual vacancy, the draft refers to what the candidate wrote, the role they applied for and where they are in the process, so it reads like a note from a person rather than the fourth copy of a template with the name swapped in. The recruiter reads it, adjusts a line if needed, and sends.
The candidate summary and client proposal agent
For a recruitment agency, the moment after the interview is where consistency matters most. Every consultant writes up candidates in their own way, and the client receives a different kind of document depending on who ran the process.
Drop the CV and the interview notes into the chat and the agent produces the same profile format every time: match against the intake, strengths, risks, open questions, and a draft proposal text for the client. The consultant decides whether to present the candidate, corrects what the agent could not know, and sends. Two candidates from two consultants now arrive at the client in the same shape, which is also the shape the client can compare.
The policy assistant
An in-house HR team answers the same employee questions all year: how many days of leave are left after the summer, what the rules are for working from home in the first month, how to declare expenses, what happens with the company car during long-term absence.
Give the agent the employee handbook and the internal policies as its context and it answers those questions from the source, with a pointer to the relevant section, and flags when a question needs a person. Colleagues get the same answer whether they ask on a Monday morning or a Friday afternoon, and the HR manager stops being the only place that answer lives.
The onboarding process agent
A new hire’s first weeks are built from documents that exist somewhere: a checklist from the last onboarding, a welcome email someone wrote two years ago, a 30/60/90 plan for a different role. Each time, someone reassembles them by hand. Agencies placing temporary staff face the same task at a higher frequency.
Give the agent one good example of each and ask it to produce the version for the new role: the checklist adapted to the team, the welcome email with the right names and dates, the plan with the milestones that apply. HR reviews, personalises the parts that should feel personal, and the new colleague arrives to a first day that was planned rather than improvised.
The one thing an agent should not do
The European AI Act classifies AI used for recruitment and selection as high risk. That covers systems that target vacancies at specific people, filter or rank applications, and evaluate candidates. For those systems, the core obligations on human oversight, transparency and data quality are scheduled to apply from August 2026, with a possible postponement under the Digital Omnibus still being negotiated. Separately, employers must tell candidates when AI is used in the hiring process, and any candidate interacting directly with an AI must be told so.
The practical line is simple. Let AI structure, draft, summarise, score and flag. Keep the decision to advance or reject a candidate with a person, and make sure that person can see why the tool suggested what it did. Be open with candidates about where AI was involved, including when they interact with it directly.
This is also why an IttesAI agent is designed the way it is. It prepares, then waits for confirmation. The recruiter or HR manager reads what it produced and decides. That is not a limitation added for caution. It is the reason the tool is usable in hiring at all.
Want to see what an AI Agent would look like for your office? Tell us how your hiring actually runs, in-house or for clients. We will tell you where an agent would make a difference and where it wouldn’t. Thirty minutes with somebody who knows the platform and has seen how other HR teams and recruitment agencies use it.
