AI handles volume well in recruitment and judgement badly. It is genuinely useful for sourcing, scheduling and first-pass screening against defined criteria. It does not remove bias, and any vendor claiming otherwise is describing something that has never been demonstrated.
It is also now regulated. New York City requires bias audits for automated hiring tools, Illinois requires consent for AI video analysis, Colorado legislates for high-risk AI including employment, and the EU AI Act classifies recruitment AI as high-risk. Last updated September 2026.
| AI does this well | AI does this badly |
|---|---|
| Sourcing and matching at volume | Assessing judgement or potential |
| Scheduling and coordination | Evaluating unusual career paths |
| Screening against defined criteria | Removing bias |
| Drafting and summarising | Explaining its own decisions |
| Answering candidate questions | Anything you cannot audit |
Gartner reports that 38% of hiring managers have used AI to improve recruitment efficiency. The question is no longer whether to use it, but where it helps and what obligations come with it.
Where does AI genuinely help?
Sourcing. Searching and matching across far more candidates than a person could review, surfacing people who would not have applied.
Scheduling and coordination. Unglamorous, time-consuming, and well suited to automation.
First-pass screening against defined criteria. Where requirements are genuinely objective, such as a specific qualification or right to work, automation is faster and more consistent than manual review.
Candidate communication. At volume, most applicants hear nothing. Automated status updates are a real improvement to candidate experience.
Where does it fail?
It does not remove bias. This is the claim to be most careful about. Systems trained on historical hiring data learn the patterns in that data, including who was hired and who was not. Removing names and demographics does not solve it, because proxies remain: postcodes, schools, employment gaps, phrasing.
AI can make bias more consistent, which is not the same as removing it. A human reviewer has a bad day; an algorithm applies the same skew to every candidate.
It struggles with unusual paths. Pattern-matching favours candidates resembling previous hires. Career changers, people with non-linear histories and those from underrepresented backgrounds are exactly who it screens out.
The failures are invisible. You never meet the strong candidate the filter rejected. There is no feedback loop, so a badly configured system can run for years looking successful.
It cannot explain itself. Increasingly a legal problem as well as a practical one.
What does the law now require?
The area the original guidance on this topic almost always misses.
New York City. Local Law 144 requires employers using automated employment decision tools to commission an independent bias audit, publish a summary of the results, and notify candidates before use.
Illinois. The Artificial Intelligence Video Interview Act requires notice, explanation and consent before AI is used to analyse video interviews, with reporting obligations attached.
Colorado. Its AI legislation covers high-risk systems including those making consequential employment decisions, with duties around reasonable care and disclosure.
European Union. The AI Act classifies recruitment and employment AI as high-risk, bringing obligations on risk management, data governance, human oversight, transparency and record-keeping. Requirements phase in over time.
This area is moving quickly. Confirm the current position in each jurisdiction where you recruit, because obligations generally attach to where the candidate is, not where you are.
How should you implement it?
Decide what problem you are solving. Volume, speed and consistency are good reasons. "Using AI" is not.
Keep humans on consequential decisions. Use AI to widen the pool and handle administration. Keep people deciding who progresses, which is also what several regulatory regimes expect.
Audit for bias, and keep the evidence. Required in some jurisdictions and sensible everywhere. Test outcomes across demographic groups rather than relying on vendor assurances.
Tell candidates. Required in several places, and reasonable regardless. Explain what the tool does and how decisions are made.
Ask vendors hard questions. What data was it trained on? Has it been independently audited? Can it explain an individual outcome? Who carries liability for a discriminatory result?
Measure what matters. Time-to-hire tells you the tool is fast. Ninety-day retention and quality of hire tell you whether it works. See high-volume hiring.
What does this mean for international hiring?
Two things.
Your obligations follow the candidate. Recruiting across several countries means several regimes, and the EU AI Act applies to systems affecting people in the EU regardless of where the employer sits.
And AI does not solve the constraint that actually limits international hiring. Finding the candidate is not the hard part; employing them lawfully is. An Employer of Record addresses that, and it is a separate problem from recruitment technology. We are explicit about this in what an EOR actually changes.
Country-level requirements are in CountryPedia.
Work with Global Expansion
We employ people on your behalf across 214 countries and territories, handling contracts, payroll, benefits and compliance once you have chosen who to hire.
Talk to our team about where you are recruiting.
Frequently asked questions
Does AI remove bias from recruitment?
No. Systems trained on historical hiring data learn the patterns in that data, including its biases. Removing names and demographic fields does not solve it, because proxies such as postcode, school and employment gaps remain. AI can make bias more consistent, which is different from removing it.
Is AI in recruitment regulated?
Increasingly. New York City requires independent bias audits and candidate notice for automated employment decision tools. Illinois regulates AI analysis of video interviews. Colorado legislates for high-risk AI in employment. The EU AI Act classifies recruitment AI as high-risk.
Do we have to tell candidates we use AI?
In several jurisdictions, yes, and it is good practice everywhere. Requirements vary on what must be disclosed and when, so check the rules where your candidates are rather than where you are.
What is AI best used for in hiring?
Sourcing, scheduling, candidate communication and first-pass screening against genuinely objective criteria. It is weakest at assessing judgement, potential and unusual career paths.
What should we ask an AI recruitment vendor?
What data the model was trained on, whether it has been independently audited for bias, whether it can explain an individual decision, and who carries liability if an outcome is found to be discriminatory.
Does AI help with international hiring?
With finding candidates, yes. Not with employing them. Employing someone in a country where you have no entity requires either incorporation or an Employer of Record, which is a separate problem from recruitment technology.
