AI Training in Recruitment: We're Teaching the Wrong Use Cases

85% of employees trained on artificial intelligence say they cannot use it in their daily work. That's the headline figure from the Docebo 2026 report, "The AI Readiness Gap," based on 2,000 corporate respondents across six countries including France.

The prevailing conclusion in the HR press: it's a change management problem. Companies invest in training but fail to prepare the ground. The diagnosis is convenient. It's also incomplete. What if the real problem isn't how we train, but what we train people on?

The gap nobody is measuring correctly

The numbers keep piling up and they all paint the same picture. According to the SHRM (State of AI in HR 2026), 82% of executives say they offer some form of AI training to their teams. At the same time, 59% report an AI skills gap in their organisation. The investment exists. The results don't.

On the ROI side, it's even starker. According to data compiled by HR Executive from Gartner and Phenom studies, 88% of HR tech leaders see no significant return on their AI initiatives. Only 21% of leaders report measurable positive ROI.

In France, the situation stands out. SD Worx reports that only 37% of French companies actively use AI features in their HR software. That's the lowest rate among the countries studied. And 54% of French HR professionals say they have received no AI training at all. This isn't a change management problem. It's a relevance problem.

The scoring demo trap

Take recruitment. When a vendor showcases the AI capabilities of their ATS, what do they show? CV scoring. Semantic matching. Automated shortlisting. These are the features that impress in demos. The ones that make people say "that's magic" in a meeting room.

The problem is that these features don't match the daily reality of most recruiters. The Docebo report highlights a telling figure: 56% of employees are so overwhelmed by "pre-AI" manual tasks that they don't have time to learn the new tools. Applied to recruitment, that number has a concrete face: the recruiter sitting on 200 unreviewed CVs, 15 candidates ghosted for two weeks, 3 job ads that were due yesterday, and hiring managers asking for updates on their open roles.

Training that person on AI scoring is like giving a sat-nav to someone whose engine won't start. The technology is sound. The placement is wrong.

What AI actually does well in recruitment

We explored this question in depth during the webinar "Recruitment and AI: Can We Optimise Without Dehumanising?", hosted with Culture RH. The conclusion was clear: AI in recruitment doesn't need to be spectacular to be useful. It needs to show up in the right places.

The use cases where AI delivers measurable, immediate value for a recruiter:

Writing better job postings. Not generating generic ads. Helping rephrase a technical role in candidate-friendly language, checking text for inclusivity, adapting tone to the target audience. A recruiter who spends 45 minutes on a job ad can cut that to 15 with well-calibrated assistance.

Preparing interviews from a CV. AI reads the career path, identifies areas to dig into, and suggests contextualised questions. The recruiter walks into the interview with a prepared framework instead of improvising.

Following up with unresponsive candidates. A personalised follow-up message based on the candidate's history in the process. Not a generic "your application is still under review" email. A message that shows someone is actually tracking the case.

Flagging stalled processes. A role open for 6 weeks with no interviews? A candidate stuck at the "manager feedback" stage for 10 days? AI can detect these signals and raise the alarm before the hire goes off the rails.

These aren't the features that dazzle in a demo. They're the ones that change the day-to-day.

Scoring has its place, but not at the front of the queue

Let's be clear: AI CV scoring isn't useless. We published a detailed analysis of the topic showing what it delivers and what it doesn't. Scoring is a prioritisation tool. It becomes relevant when the recruiter already has the bandwidth to act on its results.

The problem arises when you start there. It's the inverse of what works. Imagine a restaurant investing in a predictive reservation system while the kitchen can't get plates out on time. The tool is sophisticated. The diagnosis is off.

The right sequence is: automate the daily work first. Free up time on repetitive tasks. Then introduce analytics and scoring once the team has the cognitive and temporal capacity to use them.

IDC/Deel data adds another layer of concern: 66% of companies plan to reduce junior hiring because of AI. In other words, headcount is being cut based on tools that 88% of companies see no ROI from. The maths doesn't hold.

What this reveals about AI vision in recruitment

At Intuition Software, we've been building JobAffinity since 2006. We've seen the waves come and go: "magic" parsing, "intelligent" matching, and now "predictive" scoring. Every time, the same pattern: a strong technical promise, weak adoption, and a return to the Excel spreadsheet for daily tasks.

Our conviction is simple: AI in recruitment must be a daily assistant before it becomes an analyst. Recruiters' time should be invested where it adds the most value. Talking to a candidate, reading a CV, debriefing a manager: these are tasks where human judgement is irreplaceable, and where delegating to AI means losing what matters most.

AI training doesn't fail because employees don't understand AI. It fails because they're being shown the wrong film. The right subject isn't "how scoring works." It's "how AI can absorb the things that stop you from recruiting properly."

Sources

• Docebo, The AI Readiness Gap: The 2026 Enterprise Learning Wake Up Call
• HRDive, Why AI readiness training fails (April 2026)
• SHRM, The State of AI in HR 2026
• HR Executive / Gartner, AI's ROI problem
• SD Worx, HR Challenges 2026
• IDC / Deel, AI impact on junior hiring
• Webinar Intuition Software / Culture RH, Recruitment and AI