Choosing an AI recruitment tool: what I would ask as a recruiter

1. Name the problem before comparing tools

Map how an application arrives, who reviews it, where communication happens and how the decision is recorded. Identify where candidates disappear or a recruiter has to copy information by hand. Our heyree story began with exactly that kind of fragmentation. If you already have a reliable ATS, you may need a focused AI assistant rather than replacing the entire system. Buy against the problem, not the length of the feature list.

2. Ask what the AI actually evaluates

Separate drafting and administrative assistance from scoring, filtering and candidate recommendations. Ask what inputs the system uses, whether recruiters can inspect the underlying application, how errors are reported and what a score means. heyree's public About us page names Ree, CV scoring and candidate verification. That describes the product direction; it is not a substitute for demonstrating the current feature, explaining its limitations or supplying legal documentation.

3. Check intended use and EU risk classification

Ask the vendor to document the intended use and explain its AI Act classification. Annex III, point 4(a), covers recruitment and selection, including targeted job ads, application filtering and candidate evaluation. Limited Article 6(3) exceptions depend on risk, influence on decisions and specified conditions; profiling natural persons in these listed uses remains high risk. A human making the final hiring decision does not automatically exempt the tool. Under Regulation (EU) 2026/1744, the Annex III high risk rules apply from 2 December 2027.

4. Verify data handling, not just the hosting region

heyree states on its About us page that data stays solely in the EU and is never sold to third parties. Even with EU hosting, a buyer should check the processing agreement, subprocessors, any international transfers, access controls, retention and deletion, and whether candidate data is used for model training. Establish your own GDPR legal basis and candidate notices. A data protection impact assessment may be required where processing is likely to result in high risk. Hosting location alone does not establish compliance.

5. Make the recruiter demonstrate control

Ask someone who will use the tool daily to inspect a recommendation, find the original evidence, correct an error and override the output. The high risk AI Act framework assigns duties to both providers and employers using their systems, including effective oversight, appropriate monitoring and logging. GDPR Article 22 separately restricts solely automated decisions with legal or similarly significant effects, subject to specific exceptions and safeguards. Do not accept a rubber stamp approval step as the answer to both sets of rules.

6. Test the candidate experience and connections

Use a realistic workflow to check application handling, replies, scheduling and the handover between people. heyree's candidate first principle is that applicants deserve an answer and respect. Ask which calendar, email, job board or HR connections exist in the version you are buying, and demonstrate the ones you need. I am not claiming that every integration is available in heyree. A manual workaround can erase the time saving promised by the AI.

7. Compare the full cost and the exit

Compare the quote against your actual team size, hiring volume and required features. Ask about implementation, support, AI usage limits and export of candidate records and decision history when you leave. These are buyer questions, not invented heyree pricing. A tool is easier to adopt when you understand both how to start and how to leave without losing the record of your hiring process.

8. Run a trial with an accountable owner

Choose one role, use candidate data only with a lawful basis and safeguards, and agree on success criteria before the trial. Track administrative effort, missed or mishandled applications, the evidence supporting the shortlist and the reliability of candidate replies. Include difficult examples and check errors, not just the best demo case. As a founder, I would rather have a buyer test whether the product solves their actual problem than choose it because an AI score looks persuasive.

Sources and further reading

• heyree: About us and the founders' story
• EU AI Act: Article 6, Articles 14 and 26, Annex III
• Digital Omnibus on AI: Regulation (EU) 2026/1744
• GDPR: Articles 5, 6, 13, 14, 22 and 35

The heyree story is based on its public About us page. EU rules are summarised as checked on 13 October 2026. This is general information, not legal advice or a compliance certification of heyree.