How to audit candidate-to-ATS administration in a recruitment agency

Map what arrives

Start with one narrow intake route, such as speculative CVs received through a shared mailbox or applicants submitted for a defined desk. Do not begin with every candidate journey in the agency. For ten to twenty recent examples, record the channel, file types, naming patterns and information actually supplied. Note whether consultants receive Word files, PDFs, email signatures, portal fields or links to external profiles. This establishes the real input rather than the process people remember.

Next, follow each item until a usable ATS record exists. Write down every opening, download, rename, copy, search and status update. Include hand-offs between consultants, resourcers and compliance staff, plus any private spreadsheets used to bridge gaps. Record the ATS edition and permissions in use without assuming an integration is available. The useful map shows where information changes form, where context is lost and where a named person already checks the work.

Separate preparation from placement decisions

Draw a firm boundary between administrative preparation and recruitment judgement. Preparation can include identifying a likely name and contact details, classifying a CV, extracting agreed employment dates, proposing consistent field formats and flagging information that is absent or contradictory. Screening for a role, judging suitability, deciding whether to contact someone and progressing a placement remain human responsibilities. A workflow should never turn an uncertain extraction into an adverse candidate outcome.

Label every proposed output as prepared, flagged or approved. For example, a prepared ATS record can sit in a review queue with links to its source files and confidence notes. A recruiter then corrects the record, selects the appropriate status and decides the next action. Right-to-work, equality, data-protection and client-specific requirements need the agency’s authorised people and policies. The audit can expose where those checks occur; it does not provide legal or compliance advice.

1. List fields that may be extracted exactly from a supplied document.
2. Mark fields that require interpretation, policy or recruiter judgement.
3. Assign a named reviewer for exceptions and candidate-facing action.
4. Prevent an unreviewed record from triggering rejection, submission or messaging.

Measure the repeated work

Measure activity, not an imagined automation saving. For a representative week, count candidate items, median handling steps, incomplete records, corrections and repeat chases. Sample the time spent on a few ordinary cases and a few awkward ones. Separate waiting time from staff handling time. This produces a baseline that can be checked later without turning a small sample into a promised result.

Look for variation by source. One job board may provide predictable fields while direct email produces inconsistent attachments. One desk may use agreed ATS values while another relies on free text. These differences matter because a controlled workflow needs explicit inputs and destinations. If the route changes every time, standardise the operating process before considering technology. If most cases follow a stable path with a manageable exception queue, document that path in enough detail to test it.

Design exception routes

Exceptions should be a first-class output, not a failure hidden inside the normal queue. Define what happens when a CV is unreadable, two candidates share similar details, dates conflict, an attachment contains unexpected sensitive information or the ATS cannot accept an update. The safe default is to stop preparation, retain the source reference and send the item to a named reviewer with a short reason. Avoid guessing a value merely to complete a field.

Design the review screen or tracker around action. Show the source, proposed value, confidence or reason for the flag, and the allowed next steps. Keep an audit trail of the reviewer’s correction and approval. Agree retention, access and deletion rules with the agency’s responsible staff before using live candidate data. Test with anonymised or representative examples where possible, and include deliberately messy samples so the exception route is exercised before operational use.

1. Define a controlled vocabulary for common missing and conflicting information.
2. Route technical failures separately from candidate-information exceptions.
3. Require human approval before any candidate or client communication is sent.
4. Record corrections so repeated extraction or mapping problems can be investigated.

Check implementation readiness

A workflow is ready for scoping when its input, preparation rules, exception route, reviewer and destination are all named. Confirm that the agency can supply representative examples, explain its existing field definitions and provide authorised access to a test environment. Check whether the ATS supports the necessary read or write method in the agency’s edition. A manual export or review-ready file may be safer than a direct update when permissions are limited.

Write a short acceptance set before building anything. Include ordinary records, missing contact details, conflicting dates, duplicate candidates, unsupported files and a failed ATS hand-off. For each sample, state what should be prepared, what must be flagged and who decides. Success means the workflow follows those agreed controls consistently; it does not mean every case is automated. If the process lacks an accountable reviewer or clear exception handling, resolve that operational gap first.