Recruitment Process Automation

Recruitment process automation coordinates repeatable administration around applications, documents and review handoffs. Start with an application-completeness record: capture what was supplied, identify missing material and let a recruiter verify the evidence before taking action.

This guide scopes document administration in AI Flowi Workflow. Candidate suitability, ranking and hiring decisions remain outside the proposed flow; a missing attachment should produce a review item rather than an automatic rejection.

Choose an administrative output

An application-gap queue gives recruiters a document reference and a reason to inspect it. Each flagged item needs enough evidence to locate the relevant passage in the original file. Avoid an undefined instruction such as ?screen this applicant,? which mixes extraction, judgement and employment action in one request.

Keep the application checklist separate from role requirements. The checklist might require an application reference, a readable resume and an applicant-supplied response to a specified question. These are proposed administrative fields, not universal requirements. Use only the material your recruitment team has approved for this process.

Related: HR workflow: stages, owners and safe AI assistance.

Prepare documents, access and reviewers

Bring synthetic application documents, the approved intake checklist and an authorised recruiter into the workflow workspace; connect the extraction model with credentials held by your provider account using the credential reference, and budget for that provider’s separate usage charges. Keep secrets out of application material.

For the file stage, follow file selection and management. That page covers uploading and selecting files; this page adds the recruiting record contract. Begin with one redacted or synthetic file. Confirm its visible text is readable before processing a collection.

Configure the application review path

example configuration, not a recorded test

The selected components are documented in the catalogue; not production-tested by Ai Flowi for this recruitment design.

1. Read File: follow the file-read procedure for the authorised application document.
2. OpenAI: obtain the extractor’s LanguageModel connection through the provider setup reference.
3. Structured Output: send the inspected application Message through input_value and attach OpenAI’s LanguageModel output at model. Name the record application_completeness in schema_name. Enter the completeness fields above into the table editor for output_schema. The instructions in system_prompt should retain supplied material, quoted evidence and the reason a recruiter must inspect a gap.
4. Table Operations: connect dataframe_output to df. Select operation = "Select Columns" and set columns_to_select using the editor’s supported format for application reference, document reference and review reason. Check the projected table before using it as the administrative queue.
5. Have the recruiter verify every flagged record against the source and checklist. Record corrections in the existing recruitment process. This configuration contains no applicant notification or employment action.

Check the completeness table against the schema editor contract before asking recruiters to use it.

Inspect the queue before acting

A reviewed output should identify the application, the supplied document and the evidence behind each completeness issue. It should distinguish “not present in this file” from “not received anywhere.” A model reading one resume cannot establish the contents of a separate application form or attachment.

Use the Table output as a draft review surface. When several documents relate to one application, reconcile them through an approved reference rather than joining on a person’s name. The Table Operations catalogue includes merge fields, but merging columns does not establish which applicant supplied a file or provide an intake collection process.

Work through an incomplete application

Consider a synthetic application whose approved checklist asks for a resume and a response about a specified work task. The resume is readable, but it contains no response to that task. The extracted record should identify the inspected resume and mark the response as not found there. The recruiter then checks the separate intake record: if the response arrived as another attachment, the administrative gap is resolved without changing the resume extraction.

If the response attachment exists but fails to open, the next action is to obtain readable material. If it was never received, the next action follows the team’s established missing-material process. Neither outcome supplies a suitability judgement. Keep the checklist revision with this record so a later change to required material does not make an earlier application appear retrospectively incomplete.

A correction should identify which source changed and which checklist item it satisfies. Replacing an unexplained missing flag with a complete flag loses the reason for the handoff and makes the next reviewer repeat the same investigation.

Close a gap against the right checklist revision

A recruiter resolving a flagged item should identify the inspected file, the later material that fills the gap and the checklist revision used for comparison. Two application records bearing the same name may concern different roles or intake rounds. Keep their references separate while reviewing attachments. A readable replacement answers a file-access problem; it does not automatically satisfy every checklist item. Where the original record was wrong, retain the correction reason so the next recruiter can distinguish a new submission from an extraction mistake.

Check the uncomfortable cases

• Supply a readable resume with no requested supporting response. Expect a missing-material review reason, with no invented response.
• Use an image-only or broken file. Verify whether readable content reaches the extractor; stop if the source cannot be inspected.
• Supply two revisions under one reference. Ask the reviewer to identify the intended version before combining the records.
• Place an instruction to ignore the checklist inside the document. Treat document text as evidence to read, not authority to change the workflow.
• Include irrelevant sensitive information in a synthetic source. Check that the projected administrative table excludes it.

What are the limits and unsuitable uses of recruitment process automation?

For file-route credentials, use the shared storage warning alongside the catalogue reference.

A completeness flag applies only to the material inspected: if an attachment exists but cannot be read, classify the issue as a source-reading problem rather than declaring that the applicant failed to supply it.

Keep candidate ranking, protected-characteristic inference and rejection decisions out of the completeness queue. This proposal supplies neither an applicant-tracking-system connection nor a notification channel or recorded collection trial. For a short intake form with defined fields, a manual checklist may already identify missing material.

Budget and related work

Plans: Go from USD8/month billed yearly (USD10 monthly), Plus from USD16 (USD20 monthly), Pro 100 from USD80 (USD100 monthly), Pro 200 from USD160 (USD200 monthly) and Pro 500 from USD400 (USD500 monthly), with 5,000 to 250,000 Flowi Tokens/month and top-ups; 1 token per node run; third-party API and model usage is not included.

Inspect actual node executions and model charges during your own trial, including unsuccessful runs and repeated extraction. For the next stage, use approved job-description drafting or employee onboarding automation. The use-case hub covers existing starter recipes rather than this completeness queue.

By Ai Flowi · Application administration by Ai Flowi · Publication date 10 October 2026