How Do SMEs Save Time in Recruiting, Hiring, Onboarding and People Leadership With AI in 2026?
Talk about AI in HR and the conversation usually starts with recruiting. Sorting applications automatically, pre-qualifying candidates with a chatbot, job ads at the push of a button. That is where the pressure shows first.
My view is different. In a service-oriented SME, most of the hours AI can actually win back in 2026 sit in onboarding and in preparing and following up on conversations. And automated pre-screening of applications, of all things, is the part where I would currently push hardest on the brakes. The costly departure connects onboarding to capturing knowledge before a role change. A practical starting point combines knowledge preservation of team decisions with knowledge transfer during onboarding, reducing repeated explanations without delegating candidate selection to a filter.
People who disagree have a fair point. Recruiting time is easy to measure: days to fill a role, hours per application. Onboarding time is measured far less often. So AI budgets end up where the effect is quicker to show.
Recruiting: yes to drafting, caution with selection
AI clearly saves time on writing. A job ad in German and English, tuned to two audiences, is quick to draft with a general-purpose assistant. The real work starts afterwards, when the hiring team has to agree on what the role actually needs, and no assistant does that part for you. The same goes for acknowledgements, scheduling emails and rejections, which many applicant-tracking systems now suggest. A person approves them, so the risk is low.
It changes once a system filters applications or rates people. Annex III, point 4 of the EU AI Act classifies AI systems used for recruiting or selecting people as high-risk, explicitly including analysing and filtering applications and evaluating candidates. Under the Digital Omnibus (Regulation (EU) 2026/1744), the obligations for these systems apply from 2 December 2027 instead of August 2026. The classification stays. A screening tool you introduce today will need risk management, human oversight and logging from that date. From then on, employers must also inform workers' representatives and the affected employees before putting such a system into use at the workplace (Art. 26(7)).
So my buying rule is simple. If the tool writes text that someone checks and sends, it is a productivity tool. If it sorts, ranks or scores people, treat it as a high-risk system from day one, even while the deadline is still running.
One ban already applies. Since 2 February 2025, Art. 5(1)(f) prohibits AI systems that infer people's emotions in the workplace, with exceptions only for medical or safety reasons. If someone offers you sentiment analysis for employee reviews, ask them to show you exactly what is being analysed. Whether a job interview is covered as well is a question for your lawyer. For the timeline, see EU AI Act for SMEs. This section is orientation, not legal advice.
Hiring: comparable interviews, not faster ones
AI saves little time in the interview itself. The time is won before it and, often overlooked, months after it.
- A question set per role, drafted by AI from the job description and revised once by the team.
- A scoring grid that is the same for every candidate.
- A short note after the interview on why someone was hired and where there were doubts.
The note looks like paperwork. It pays off when the new colleague starts and her manager can look up which expectations were voiced in the interview. With RECO, these notes go into the same knowledge base that onboarding later draws on.
Onboarding: the block that rarely gets measured
In their first months, new hires mostly get knowledge by asking. Every question costs two people time, and the answer is gone once the conversation ends. A searchable knowledge base changes that maths, because from the second time onwards people look it up instead of asking.
You can see this at AAZB (Akademie für angewandte Zukunftsbildung), which has trained coaches, consultants and trainers for more than ten years. Meetings, jour fixe sessions and stand-ups are captured and analysed with RECO, 92 percent of conversation content is now searchable, and new staff settle into their roles faster. The use case on hiring interviews and onboarding shows how it works in practice.
If you want to estimate the effect in your own company first, a simple count is enough. Ask new hires to write down every question they put to colleagues during their first weeks, then mark which of those had already been answered for someone else. That ratio shows fairly directly how much time a knowledge base can win back.
Leadership: a calculation with your own values
Blanket figures for the time AI saves in leadership do the rounds, such as "a few hours a week". It tells you nothing about your company. Work with four values of your own instead.
- One-to-ones per week: direct reports divided by how many weeks pass between meetings with each person.
- Minutes per conversation for preparation (finding old notes, collecting open points) and follow-up (recording agreements, setting next dates). Log them for a few conversations. Don't guess.
- The share of that time a tool actually takes off your hands. You only know it after a test with real conversations; a vendor's promise is no substitute.
- Your working weeks per year.
Conversations per week × minutes per conversation × share × working weeks, divided by 60, gives you the hours per year for one manager. Multiply by the number of managers and you have a figure you can put next to the license price.
How large the share turns out depends entirely on how you document today, how much you write down after each conversation and where those notes end up when you need them again. Someone who barely takes notes today saves little time and mainly gains traceability.
Prompts during a meeting are useful as long as they refer to content, such as an open agreement from last time. Once they start interpreting the person, you are back at the ban from the recruiting section.
A sequence that avoids tool sprawl
- Pick the phase with the largest measurable effort. In my view, for most SMEs that is onboarding or conversation documentation, and rarely recruiting.
- Measure before you start: questions from new hires, minutes per conversation, days until the first task completed independently.
- Introduce as few tools per phase as you can, and involve the works council before any conversations are recorded. Our post on documenting employee reviews covers what needs to be agreed.
- Set default rules, for instance that every internal employee review is prepared and documented.
- After a period you fix before the start, measure the same values again, and only then decide on the next phase.
Where I draw the line
Hiring decisions, conflict resolution and development conversations stay with people. For me, the line runs where a system starts judging people instead of looking things up for them. On this side of it, AI saves real hours in the people process. On the other side, what you mostly buy in 2026 is a set of obligations for December 2027.
Sources
Related articles
In a 30 minute software demo we show you how RECO secures and transfers knowledge in your company. Request a demo now and we'll get back to you with a proposed time.
