Short answer: There is no single best ATS for recruitment agencies. The best one for your agency runs several clients from one place, keeps the evidence behind each candidate, hands work to clients and your back office without retyping, and follows Dutch and EU data rules. Prove it with a test on your own live roles, not the vendor's demo script.
Key takeaways
- Seven criteria separate agency systems: clients and branches, intake at volume, evidence per requirement, checks before the offer, handovers, reporting per client, and data rules with a clean exit.
- Weigh them by how your agency earns its margin, then give every shortlisted vendor the same six tasks on your own roles.
- AI features that filter applications or evaluate candidates are listed as high risk in the EU AI Act; the obligations apply from 2 December 2027.
- Count retyping hours and the cost of leaving, not only the licence fee.
On this page: What makes an agency ATS different from an employer's ATS? · Which criteria matter most when choosing an ATS for a recruitment agency? · Why should an agency ATS hold evidence, not just CVs? · Which Dutch and EU rules should an agency ATS help you follow? · How do you test an ATS on your own roles? · What does an agency ATS really cost? · What does this mean for frontline and high volume agencies? · What can this checklist not tell you? · So what is the best ATS for your agency? · How Radius helps · Frequently asked questions
What makes an agency ATS different from an employer's ATS?
An agency ATS has to serve candidates, clients and recruiters at once. An employer's applicant tracking system assumes one organisation, one set of vacancies and one owner of the data. An agency works for many clients, often from several branches, and the same candidate may suit three clients this month. Each client should see only the candidates you put forward to it.
A system with excellent job posting can still fail an agency if it cannot keep clients apart, show a client one candidate without exposing your pool, or pass a placement to the back office that runs contracts, hours and payroll.
A second difference matters more. A client pays for a person who meets the requirements and turns up for the first shift, not for a CV. In frontline work the deciding facts are concrete: a valid forklift certificate, the right to work in the Netherlands, availability for nights from a given date. An agency ATS should hold those facts and show where each came from.
Which criteria matter most when choosing an ATS for a recruitment agency?
Seven criteria matter, weighted by how your agency earns its margin. Use the table in every demo.
| Criterion | What good looks like | Ask the vendor | Warning sign |
|---|---|---|---|
| 1. Clients, sites and branches | Each client's vacancies, requirements and submissions kept apart, with access set per team and branch | "Show me one candidate put forward to two clients, and what each client sees." | Clients separated only by a tag or a naming habit |
| 2. Intake at volume | Every channel lands on one candidate record; candidates apply and reply from a phone without an account | "How many steps does a candidate take to apply from a phone?" | A login or a long form before the first question |
| 3. Evidence per requirement | For each requirement: the question asked, the answer given, and whether a document or check backs it | "Show me how I tell what a candidate said from what someone checked." | Requirements kept as free text notes |
| 4. Checks before the offer | Identity, right to work and certificates recorded with date, source and expiry, reviewed by a named person | "What happens on the record when a certificate expires?" | Checks tracked in a spreadsheet outside the system |
| 5. Handovers | Submission to the client, then the placement passed to your back office, without retyping | "Which fields does my team type twice between the ATS and payroll?" | "You can download a file" as the only answer |
| 6. Reporting per client | Time to first contact, dropout per step and fill rate, per client, branch and source | "Show me where candidates dropped out for one client last month." | Agency totals only |
| 7. Data rules and exit | Hosted in the EU, a data processing agreement, retention set per type of data, and a full export when the contract ends | "How do we get all our data out, in what format, and how fast?" | Export only on request, at extra cost, or in a format you cannot reuse |
The weights depend on your model. An agency placing hundreds of order pickers a month weighs intake and reporting heavily; a specialist in certificated roles such as forklift or security work weighs evidence and checks. Agree the weights with recruiters, planners, finance and your privacy lead before the first demo, so the flashiest feature on the day does not set them.
For AI features, see how to choose AI recruitment software: 10 questions to ask any provider.
Explore how interviews, evidence and checks sit on one candidate record: see the Radius ATS.
Why should an agency ATS hold evidence, not just CVs?
Because clients pay for placements that work, and a CV cannot show that. A CV is what a candidate wrote about themselves, often for a different job. A 2023 review by Sackett, Zhang, Berry and Lievens found that structured interviews predicted job performance best, on average, of the selection methods it compared [1], building on the same authors' 2022 reanalysis [2]. A structured interview asks every candidate the same questions about the same requirements and records the answers.
For an ATS, each requirement in a vacancy should become a line on the candidate record with three possible states: shown by a document or a check, said by the candidate, or still missing. A recruiter can then put a candidate forward with an honest summary, and the client sees the same picture.
If the vendor says each requirement lives "in the notes", the system will not help when a client asks why a candidate was put forward. See also how to run structured interviews at high volume.
Which Dutch and EU rules should an agency ATS help you follow?
Four sets of rules touch what your ATS records. They are the GDPR, the Dutch foreign nationals employment act (Wav), the Wtta and the AI Act. Software does not make you meet them, but it makes the records easy or hard to keep.
| Rule | What it says | What to check in the ATS |
|---|---|---|
| GDPR, Article 5(1)(e) | Personal data may be kept in a form that identifies people for no longer than necessary for the purpose [3] | Retention set per type of data and per client, with deletion that actually runs |
| GDPR, Article 28(3) | A processing contract must require the vendor to delete or return all personal data at the end of the service, at your choice, unless EU or national law requires storage [3] | A data processing agreement before real candidate data goes in, and an export you have tested |
| GDPR, Article 22 | Candidates have the right not to be subject to a decision based solely on automated processing that significantly affects them, such as an automatic rejection, unless an exception applies [3] | Whether knockout questions can reject candidates without a person seeing the result |
| Wav, Article 15 | When an agency places a worker without Dutch nationality with a client, the client must receive a copy of the identity document at the start of the work, check the worker's identity against it and keep the copy for at least five years after the end of the calendar year in which the work ended. EU, EEA and Swiss nationals with free movement are exempt [4] | Whether the record shows what was passed to which client, and when |
| Wtta (Stb. 2025, 385) | The admission system for labour suppliers takes effect on 1 January 2027, with new notification and administration duties for agencies and their clients; the bans on supplying or hiring workers without admission take effect on 1 January 2028 [5] | Whether those records live in the ATS, the back office or nowhere yet |
| EU AI Act, Annex III and Article 26 | AI systems intended to analyse and filter job applications or evaluate candidates are listed as high risk. From 2 December 2027, organisations using them must, among other things, assign people with the competence, training and authority to oversee them, keep the logs under their control for at least six months unless data protection law provides otherwise, and tell candidates a high risk AI system is used [6] [7] [8] | Which features use AI, whether the vendor treats them as high risk, and whether logs are kept and exportable |
The Wtta and AI Act dates an agency ATS must be ready for
- 2 August 2026AI ActThe earlier date for the high risk duties, moved by the Digital Omnibus on AI
- 1 January 2027WttaThe admission system for labour suppliers starts, with new notification and administration duties for agencies and their clients
- 2 December 2027AI ActDuties apply to organisations using AI that filters applications or evaluates candidates
- 1 January 2028WttaBans on supplying, or hiring, workers without admission
In AI Act terms, the vendor is normally the "provider" and the agency using the tool in its own hiring is the "deployer". An agency can take on provider obligations if it puts its own name or trademark on a high risk system, substantially modifies it, or uses it for a high risk purpose the vendor did not intend [6], so ask how obligations are split before using AI features under your own brand. The European Commission's draft guidelines on high risk classification, published for consultation on 19 May 2026, are not final and not binding. They give an AI system that employment agencies use to assign candidates to vacancies, by ranking candidates for them, as a high risk example that cannot use the narrow exception [9]. Full detail: the EU AI Act and recruitment.
How do you test an ATS on your own roles?
Give every shortlisted vendor the same six tasks, built from your own roles. A demo shows tidy data; your test should show a busy Tuesday, with applicants waiting and a client on the phone.
- Pick three live roles from two clients: one high volume role, one that needs a certificate, and one with shift conditions such as nights. Write the requirements down first.
- Send the tasks a week ahead. Use test records, not real candidate data, until a data processing agreement is in place.
- Put a recruiter and a planner at the keyboard and time each task:
- Set up both clients and show what each client, branch and recruiter can see.
- Apply for the volume role from a phone, then answer the first message. Count the steps.
- Record the answer to each requirement for one candidate, and show which answers a document backs.
- Add a forklift certificate that expires next month. Show what the record does when it expires.
- Put the candidate forward to the client, then pass the placement to your back office. List every field typed twice.
- Delete one candidate's data, then export everything you would take with you if you left.
- Mark each task pass, partial or fail against the criteria you agreed.
- Ask two reference agencies of your size and model about tasks 4, 5 and 6.
Run the same test on every vendor, including us.
Illustrative example: weighing two systems after the test
An agency placing mostly order pickers and forklift drivers for six warehouse clients sets weights out of 100 and rates two shortlisted systems from 1 (fails the task) to 5 (does it well). Points = weight × rating ÷ 5.
| Criterion | Weight | System A rating | System A points | System B rating | System B points |
|---|---|---|---|---|---|
| Clients, sites and branches | 15 | 4 | 12 | 3 | 9 |
| Intake at volume | 20 | 3 | 12 | 5 | 20 |
| Evidence per requirement | 15 | 2 | 6 | 4 | 12 |
| Checks before the offer | 15 | 2 | 6 | 4 | 12 |
| Handovers | 15 | 5 | 15 | 3 | 9 |
| Reporting per client | 10 | 4 | 8 | 3 | 6 |
| Data rules and exit | 10 | 3 | 6 | 4 | 8 |
| Total | 100 | 65 | 76 |
Points per criterion, System A and System B
Illustrative agency with six warehouse clients. Total out of 100: System A 65, System B 76
System A leads on handovers, clients and reporting; System B leads on intake, evidence and checks, the work that fills this agency's roles. Before choosing B, price its handover gap (next section) and ask whether the vendor can close it. Set one floor too: a rating of 1 on any criterion rules a system out, whatever its total.
What does an agency ATS really cost?
The licence is one line; count retyping and leaving too. Vendors price per user, branch, vacancy or placement, or as an annual plan sized to hiring volume, so price each offer on your own branches, placements and client users.
Two costs are specific to agencies. The first is double entry between the ATS and the back office.
Illustrative example. An agency makes 1,500 placements a year. If each needs 8 minutes of retyping between the ATS and the back office, that is 1,500 × 8 ÷ 60 = 200 hours a year. At an assumed internal cost of €40 an hour, that is €8,000 a year, before anyone fixes a typing error. Replace every number with your own.
The second is the cost of leaving: export fees, the notice period and the work of moving candidate and client records. Ask every vendor: Does the price change when we add a branch or a client? Do client users count as paid seats? What does a full export cost, and in what format?
What does this mean for frontline and high volume agencies?
Speed and proof decide frontline placements, so weigh intake and checks heavily. A candidate who applies from a phone at 21:00 after a shift may have taken another job by the time a morning batch is processed.
- First contact: check how fast the system answers a new applicant, including outside office hours.
- Shift facts: start times, nights and location belong in the requirements. See why you should tell candidates about the 05:30 start.
- Checks: certificates and right to work belong before the offer, not at the client's gate. See which checks can happen before the offer.
What can this checklist not tell you?
It cannot fix a vague intake: if the client's requirements are unclear, no system records good evidence against them. The weights are judgement, and a vendor's demo answer is a claim until your test confirms it. The legal points above are general; how they apply depends on your setup, contracts and the features you switch on, so get advice for your situation.
So what is the best ATS for your agency?
The best ATS for your agency is the one that fits how you serve clients: it keeps clients apart, records evidence and checks before the offer, hands placements on without retyping and lets you leave with your data. Write your criteria down, weigh them by your margin model and make every vendor pass the same test on your own roles.
How Radius helps
Radius Hire is the AI hiring platform for frontline roles. Its AI Interviewer, Match and Radius Passport sit on one candidate record.
- Several clients and sites. Radius is built for staffing agencies that recruit for several clients and sites; how clients, branches and permissions are set up is agreed in the demo. See Radius for agencies.
- Evidence per requirement. Applicants take a short, structured voice interview on their phone, in the language set for the role, and are told at the start that the interviewer is automated and that a person reviews the result. For each requirement, Radius shows whether it is backed by a document or a check, rests only on what the candidate said, or is still missing.
- Checks before the offer. Where the employer uses Radius Passport, it checks identity and right to work with a trusted verification partner and records certificates with their expiry dates, so a recruiter can review the evidence before an offer. It does not replace the employer's own identity check.
- A person decides. Radius never rejects a candidate on its own. A recruiter reviews the evidence, can overrule anything Radius presents and makes every hiring decision.
- Your setup and your data. Radius is built to run next to your existing ATS, or to be used as the Radius ATS; which systems connect is agreed in the demo. Candidate data is hosted in the EU and processed under a data processing agreement, and the employer sets how long it is kept. See trust and compliance.
Radius is new, so this section describes how the product works rather than claiming results.
Frequently asked questions
Should an agency ATS include a CRM?
For most agencies, client and candidate records should at least be connected. An ATS tracks candidates through vacancies; a CRM tracks clients, contacts and sales, and every placement links both. If they are separate systems, check that they sync without retyping.
Can an ATS reject candidates automatically?
Some can, but treat automatic rejection with care. The GDPR does not ban automated screening, but where an employer relies on the contract or explicit consent exception for a solely automated decision such as a rejection, candidates must at least be able to get human intervention, give their view and contest it [3]. Safer practice: send a missing or failed answer to a person, not to a rejection.
Is an ATS with AI features high risk under the EU AI Act?
Yes, when the AI filters applications or evaluates candidates. Those uses are listed in Annex III, with a narrow exception for AI that does not materially influence hiring decisions, never available where the system profiles candidates [6]. The obligations apply from 2 December 2027, after the Digital Omnibus on AI moved the date from 2 August 2026 [7] [8].
What should the data processing agreement with an ATS vendor cover?
It should cover at least what Article 28 of the GDPR requires, including processing only on your documented instructions, confidentiality, security measures, conditions for bringing in other processors, help with candidates' rights requests, and deleting or returning all personal data at the end of the service, at your choice [3]. Add hosting location, retention per type of data and how fast a full export is delivered.
Do we need to replace our back office when we change ATS?
Usually not; the two can run side by side, with the ATS handling recruitment up to the placement and the back office handling contracts, hours and payroll. Ask each vendor to show which fields move across automatically and which are typed again.
General information, not legal advice. Position as of 22 October 2026.
Sources
- Sackett, P. R., Zhang, C., Berry, C. M. and Lievens, F. (2023). "Revisiting the design of selection systems in light of new findings regarding the validity of widely used predictors." Industrial and Organizational Psychology, 16(3), 283 to 300. https://doi.org/10.1017/iop.2023.24. Checked 22 October 2026.
- Sackett, P. R., Zhang, C., Berry, C. M. and Lievens, F. (2022). "Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range." Journal of Applied Psychology, 107(11), 2040 to 2068. https://doi.org/10.1037/apl0000994. Checked 22 October 2026.
- Regulation (EU) 2016/679 (General Data Protection Regulation), Articles 5(1)(e), 22 and 28. EUR-Lex. https://eur-lex.europa.eu/eli/reg/2016/679/oj. Checked 22 October 2026.
- Wet arbeid vreemdelingen (Wav), Article 15, version in force from 12 June 2026. Dutch government legislation portal. https://wetten.overheid.nl/BWBR0007149/. Checked 22 October 2026.
- Besluit van 24 juni 2026 over de inwerkingtreding van de Wet toelating terbeschikkingstelling van arbeidskrachten, Staatsblad 2026, 159, with explanatory note. https://zoek.officielebekendmakingen.nl/stb-2026-159.html. Checked 22 October 2026.
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 3, 6, 25 and 26 and Annex III, point 4. EUR-Lex. https://eur-lex.europa.eu/eli/reg/2024/1689/oj. Checked 22 October 2026.
- Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal L, 24 July 2026, Article 1(40). EUR-Lex. https://eur-lex.europa.eu/eli/reg/2026/1744/oj. Checked 22 October 2026.
- European Commission. "AI Act" policy page, last updated 19 October 2026. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai. Checked 22 October 2026.
- European Commission. "Draft Commission guidelines on the classification of high-risk AI systems", Annex III part, point 4(a) examples, published for consultation 19 May 2026, page updated 23 July 2026, still a draft and not binding; no final version published as of 23 October 2026. https://digital-strategy.ec.europa.eu/en/library/draft-commission-guidelines-classification-high-risk-ai-systems. Checked 23 October 2026.