Job Intake Form for Recruiters: Fill It From the Call With AI

Use a fixed job intake form, capture the client's call with an agreed recording or structured notes, and ask AI to fill only fields supported by the call. Mark gaps and contradictions instead of guessing. Before sourcing, confirm salary, genuine must-haves, working pattern, interview stages and who can approve changes directly with the client.

The draft is an evidence map, not a signed-off vacancy. A fluent summary can hide an expensive misunderstanding: “up to $110,000” may describe an exceptional ceiling, while “hybrid” may mean two fixed office days. Keep the source phrase beside every consequential field, then make the hiring manager resolve uncertainty in writing.

Follow me on Instagram@sagnikteaches

Fix the form before the client speaks

A job intake form for recruiters works best as a repeatable agreement about one vacancy. Create the form before choosing a transcription or AI tool. If every recruiter uses a different set of headings, an extraction prompt cannot tell which omissions matter. Keep fields short enough to complete during an ordinary briefing call, but separate facts that lead to different sourcing decisions. A single “requirements” box is too vague to show whether a qualification is essential, merely preferred or still under discussion.

Connect on LinkedInSagnik Bhattacharya

Start with vacancy identity: internal reference, job title, department, reporting manager, reason for hiring, headcount and target start date. Then capture the work: three outcomes expected in the first months, the routine duties, tools the person will actually use, team size and the working arrangement. Add compensation as distinct fields for approved base range, any variable pay, approval ceiling, benefits the recruiter may mention and any figures that must stay internal. The final section should cover screening criteria, interview stages, named decision makers, feedback turnaround and authority to change the brief.

Subscribe on YouTube@sagnikteaches

The employer intake guide from Indeed treats role requirements, salary, hiring timeline and screening steps as intake topics. That is a useful starting list, but a working agency form needs an extra column: how the recruiter knows each answer. Use “client stated on call”, “approved document”, “client confirmed by email” or “unconfirmed”. These labels stop an attractive AI paraphrase from being mistaken for an instruction.

Give the form a version and owner. “Vacancy 27, version 1, awaiting pay approval” is more useful than a document named “final brief”. It also lets a second consultant see whether a later change came from the client or from an AI rewrite. If the client has an existing job description, place it beside the call notes as a separate source. The call may supersede it; the extractor should identify a conflict, not quietly choose one. For wording a candidate will see, the job advert tutorial deals with the next stage after the approved intake.

One vacancy from call to approved brief

Consider an illustrative recruitment agency with three consultants filling a data operations role for a client. The client briefing lasts 24 minutes. The agency's form has 18 fields: six identity and process fields, five work and location fields, four requirements fields and three compensation fields. A consultant records the call with the client's agreement and stores the transcript in an approved workspace. This is an example workflow, not a claim about how long every intake takes or how much automation saves.

On the call, the manager says the hire will “own the weekly reporting pack”, work with spreadsheet exports and a database query tool, and support a team of four. The stated base range is $82,000 to $96,000. The manager then adds, “We might stretch to $105,000 for somebody exceptional, but I need finance to say yes.” They describe office attendance as “a couple of days most weeks”, call previous sector experience “helpful”, and say the person “must have five years of experience”. They mention two interviews, then later refer to “a short practical exercise”.

A first AI pass fills 11 fields cleanly, marks four as ambiguous and leaves three blank. It should put $82,000 to $96,000 in stated base range, $105,000 in conditional ceiling with “finance approval outstanding”, and no number in approved advertising range. It should not convert “a couple of days” into exactly two fixed days, or turn “helpful” sector experience into a knockout criterion. The five-year statement needs a human discussion about the capability required and how to assess it; an automatic screen based on years alone may reject people who can do the work.

The consultant spends an illustrative 12 minutes checking the transcript against those 18 fields and sends a five-question clarification email. The client confirms an advertising range of $82,000 to $96,000, says the $105,000 ceiling is internal only, specifies two on-site days agreed with the team, names database querying as essential, and treats sector experience as desirable. The practical exercise is part of the second interview rather than a third stage. The form now has 16 confirmed fields and two openly pending fields: start date and feedback turnaround. The agency can discuss the role internally, but it waits for those two answers before promising candidates a timeline.

The numbers here are illustrative checkpoints, not a benchmark. In this example the AI's valuable work was sorting 18 fields and surfacing seven items that needed attention: four ambiguous, three missing. The consultant's value was deciding which questions mattered, checking the transcript and obtaining approval. If a second recruiter receives only the AI summary, they may quote the internal ceiling, misdescribe the working pattern and invent an interview stage. A filled form with source lines and approval status prevents that drift.

Capture the call with a defined boundary

Tell the client what you propose to capture, which tool or service will process it, who at the agency can access it and when it will be deleted under your policy. Follow the client agreement and applicable recording and data-protection rules; ask a qualified adviser when the position is unclear. Do not treat permission to discuss a vacancy as permission to record or upload every conversation. If the client declines recording, use notes. The practical goal is a source that both parties understand, not a perfect transcript at any cost.

At the start, read back the role title and the meaning of “must-have”. During the conversation, ask for examples: “What would this person do in the first month that proves they can query the database?” Follow a vague salary statement with “Is that the range we may quote, or an internal ceiling awaiting approval?” Close by reading back the criteria, pay, working arrangement and interview plan. Those four read-backs make the later AI pass easier because the client's corrections appear in the same source record.

Record only the briefing you need. Candidate names, current employees' performance details and sensitive personal stories can be removed from notes before submission to an AI service if they are irrelevant to the vacancy. Store the recording and draft form in the agency's approved system, with access limited to the team working the role. The recruiter note-taking guide goes deeper on tool choice and permission. For this form workflow, the important rule is simple: if you cannot safely use the recording, a careful live note is a better input than an unauthorised upload.

For example, an illustrative client says, “The last person struggled after a medical absence, so we need someone resilient.” The medical detail does not belong in the vacancy form or an AI prompt. The consultant can ask for the work demand instead: “Which duties need cover during busy weeks?” The form can then say “maintains the weekly reporting pack during month-end peaks” if the client confirms it. That turns an irrelevant personal explanation into a job-related outcome and leaves out someone else's health information.

Another client may refuse recording but permit the recruiter to keep notes. Use a four-column note page: field, exact client phrase, follow-up question and status. A note such as “two days? manager said ‘usually’” is more honest than a tidy sentence claiming a fixed two-day policy. At the end of the call, read back the disputed items and log the client's correction. The AI pass can organise these notes, but must label them as notes rather than a verbatim transcript.

Ask AI for evidence, gaps and conflicts

Paste the fixed field list and the authorised source text into the agency's approved AI workspace. If the tool cannot be used with client data under your agreement, use a redacted extract or do the task manually. Keep the job description, call transcript and your notes in separate labelled blocks. Ask the tool to return one row per field with a proposed value, source quote or line reference, status and question to resolve. “Summarise the call” is too open: it produces readable prose but rarely preserves the uncertainty a recruiter needs.

Task: Populate our vacancy intake form from the sources below.
For each field, return: proposed value | source label and short exact phrase |
status (confirmed on call / ambiguous / conflicting / absent) |
question for the hiring manager.
Do not infer pay, location, qualifications, interview stages or approval.
Separate an approved advertising range from a conditional internal ceiling.
If the call and job description disagree, show both versions.
Fields: title; reporting manager; vacancy reason; headcount; outcomes;
day-to-day duties; tools; team; working pattern; travel; base range;
variable pay; internal ceiling; must-haves; desirables; interview stages;
decision maker; target start date.
Source A: transcript of client call, dated [date].
Source B: client job description, version [version].

The instruction “do not infer” matters because AI systems are trained to complete patterns. A model may make a vacancy sound polished by filling a blank benefit, merging two pay numbers or smoothing a contradiction. Require a short supporting phrase for each value. If a value cannot be traced to an authorised source, delete it. A transcript with line numbers or timestamps makes this check faster, but line numbers are a convenience rather than proof; someone still has to read the underlying passage.

Here is a deliberately flawed illustrative output for the example vacancy: “Salary $82,000 to $105,000; hybrid two days; five years' sector experience required; three interview stages.” All four clauses overstate the call. The corrected row set is “stated base $82,000 to $96,000; internal possible ceiling $105,000 pending finance approval; on-site arrangement unclear; five years mentioned but job-related test unresolved; two interviews, practical exercise placement unclear.” The corrected version is less elegant, but it tells the consultant exactly what to ask.

If the client sends an old job description after the call, tell the extractor to show conflicts explicitly. Suppose the document says “fully remote” while the manager says “a couple of office days”. The output should contain two source-labelled values and a question: “Which arrangement can candidates rely on?” It must not choose the newer statement automatically. The manager may have misspoken, the document may be stale, or the position may allow different arrangements for different hires. Only the client can settle that commercial promise.

Keep the extraction stage distinct from recruiter prompts for outreach and shortlists. An intake prompt should produce an internal evidence record, with blanks and warnings. An outreach prompt is for a candidate-facing message once the brief is approved. Mixing both jobs in one request encourages the AI to turn provisional remarks into confident selling points.

Audit the draft where an error changes the search

Review every proposed field against the source, then prioritise the fields that change who gets contacted or what the agency promises. Salary and working arrangement affect whether a candidate will consider the role. Must-haves determine who may be excluded. Interview stages and feedback dates shape candidate expectations. A typo in an internal reference is inconvenient; a wrong salary or qualification can waste a search and damage trust. Use a red, amber and green review mark: red means do not advertise or screen, amber means a named question remains, green means the client approved the wording.

Check for four kinds of AI error. First, a hallucination: the output names a benefit, qualification or stage absent from every source. Second, a merge: two separate amounts become one salary range. Third, a polarity flip: “not required” becomes “required”. Fourth, a lost condition: “if finance approves” disappears. Search the transcript for the proposed value, not only for the field heading. AI may cite the right sentence while misunderstanding its qualifier.

A realistic illustrative polarity error occurs when the client says, “They do not need a formal degree if they can show the reporting work.” A draft intake row that reads “degree required” is a search-changing failure. Correct it to “formal degree not required; assess evidence of reporting work” and ask the hiring manager what portfolio or task would demonstrate the skill. This is also a prompt for the team to check whether any screening filter or advert draft accidentally reinstates a degree requirement.

A second error is a preference promoted into a barrier. The manager says, “It would be nice if they had used our system.” The AI puts that system under must-haves because it appears repeatedly in the call. Ask the manager whether an otherwise qualified applicant with a comparable system should be considered. If yes, put the named system in desirables and write the essential capability separately, such as “can reconcile weekly exports and explain discrepancies”. That wording gives sourcers a better search target and gives interviewers something assessable.

Do not let AI decide whether a criterion is fair or legally suitable. It can flag vague or inconsistent wording, but a qualified adviser should assess a criterion when it raises employment or data-protection concerns. The recruitment bias check routine is relevant once the criteria are written; this intake step supplies the precise, approved criteria that later checks can test. Keep a record of changes so an assessor can see why a requirement was added or removed.

Turn unresolved items into one client email

AI can draft the clarification email from amber and red fields, but the recruiter must choose the questions and review the message. Send the smallest set that will unblock the search. Group related questions so the client can answer in one reply. Avoid asking “Can you confirm everything?”; a broad request often earns a vague “looks fine” that does not settle the key distinctions. Quote the proposed wording and ask for a choice or correction.

Subject: Confirming the vacancy brief before we start sourcing

Hi [first name], thanks for the call. I have the base range as $82,000 to $96,000. May we quote that range to candidates? I have kept the possible $105,000 ceiling internal because finance approval is pending. Please correct that if needed.

Could you also confirm whether on-site attendance means two agreed days each week, whether database querying is essential and prior sector experience is preferred, and whether the practical exercise sits within the second interview? I will hold those points as unconfirmed until you reply.

Best, [your name]

This illustrative email is intentionally explicit about the difference between an advertising range and a ceiling. It does not assume the candidate can be offered the ceiling. It names the practical exercise because a candidate may plan time around it. When the client replies, attach the answer to the corresponding form fields and change their status. Do not simply replace the earlier form; the history explains why a consultant who saw version 1 may remember a different number.

Suppose a client replies, “Use $90,000 to $105,000; finance has approved it.” The recruiter should update both the approved range and approval source, not just edit the top line of a summary. Check whether the advert, sourcing message and any live candidate conversations already used $82,000 to $96,000. If they did, tell those candidates the corrected range. A changed brief is an operational update across all touchpoints, not a quiet document edit.

In another illustrative case, the client answers only “yes” to an email containing four questions. That reply does not reveal which proposed wording they accepted if the email included alternatives. Ask one follow-up with numbered options: “For the working pattern, is it A: two agreed on-site days each week, B: flexible attendance as work requires, or C: another arrangement?” Use the same discipline for compensation. “Yes” is evidence only when the proposition was unambiguous.

Move the approved fields into the recruitment system

Once the named decision maker approves the brief, transfer fields into the agency's vacancy record. Map each form field to a system field before any automation: title to title, approved pay range to the outward-facing compensation field, internal ceiling to a restricted note if the system supports one, must-haves to a human-reviewed screening checklist, and interview stages to the process record. Do not paste the whole transcript into a general notes box just because the structured fields are inconvenient. A precise intake should make later retrieval and change control easier.

If the system has an AI summary field, treat it as a convenience view. The approved form and client email remain the authority for disputed facts. Restrict access to internal commercial information and remove unnecessary personal information from material used for candidate messaging. Draft the advert only from approved public fields; the job specification and outreach workflow covers turning those fields into separate pieces of copy. For candidate summaries later in the process, use a distinct evidence standard described in the candidate summary tutorial.

Test the transfer with one vacancy before automating it for every role. Choose a record containing an approved range, a conditional internal ceiling, one desirable skill and one missing field. After import, have a second consultant open the vacancy as if they were about to source. Can they tell which salary to quote? Can they see that the ceiling is private? Is the desirable skill clearly distinct from a must-have? Does the missing item still show an owner? If the answer to any is no, fix the field mapping rather than editing the AI prompt again.

For an illustrative two-role client, the manager may discuss a senior and a junior vacancy on one call. Do not ask AI for a single blended form. Create two vacancy IDs and require each extracted statement to be attributed to senior, junior or shared. A sentence such as “the senior role can approve reports; the junior role prepares them” changes both responsibilities and screening criteria. If the transcript never specifies which salary belongs to which role, mark both pay fields unresolved. A neat combined range would be worse than two honest blanks.

Check whether the new intake process works

For the first few vacancies, record how many form fields were filled from a traceable source, how many needed clarification, and how many changed after client approval. Those are process counts, not claims that AI saved time. Track corrections that reached an advert, outreach message or screening decision, because those are the errors the form is meant to prevent. Read several forms yourself: a high completion percentage can hide invented values if the source audit is weak.

A small illustrative review might inspect six briefs with 18 fields each, or 108 field decisions. If 94 fields have clear source and approval, eight are explicitly pending and six were corrected after the extraction pass, the useful question is whether any of the six corrections escaped into candidate-facing work. If two did, the release gate needs changing. A team might require a second reviewer for salary and must-haves until those errors stop. The 94, eight and six figures are an example of how to inspect the workflow, not a target for agencies.

Ask consultants whether the client clarification email is arriving before sourcing starts. If not, the bottleneck may be the client's approval route rather than the AI tool. Name a deputy approver or agree a shorter decision window; do not mark uncertain fields as confirmed just to keep a vacancy moving. When an intake change arrives mid-search, compare the old and new criteria, review candidates screened out under the old brief, and send a concise update to everyone working the role. The form should remain useful after the initial call.

The simplest standard is that another recruiter can open the approved record and explain the salary they may quote, the skills they may screen for, the practical assessment candidates will face and the questions still awaiting a client answer. If they cannot, the intake is unfinished regardless of how polished the AI summary looks.

Questions about recruiter intake records

Can I use AI if the client will not allow a call recording?

Yes. Take structured notes during the call, read back the sensitive fields, and run the same extraction process on your notes afterwards. Label the source as notes rather than a transcript. AI cannot recover words you did not record, so leave any missing field blank and ask the client instead of letting the tool guess.

Should the completed intake form become the job advert?

No. The intake form is an internal agreement about scope, pay, screening and decision makers. A job advert is candidate-facing copy. First confirm the form with the client, then draft the advert from the approved, shareable fields. Keep confidential reasons for a vacancy, internal budget ceilings and interview notes out of public copy.

Who approves a change to a must-have requirement after sourcing starts?

The named hiring decision maker should approve it in writing. Record the old wording, new wording, reason and date against the vacancy. Tell the recruiters already working the role, then review any candidates screened out under the old requirement. An AI-generated update may help draft the change log, but the approval and candidate review remain human decisions.

Further reads

• ICO AI Recruitment Report: A Checklist for UK Recruitment Agencies — Check safeguards when recruitment AI touches personal information.
• Recruitment CRM AI: Candidate Records, Duplicates and Follow-Up — Keep vacancy and candidate records clean after intake.
• AI CV Screening for Recruitment Agencies: Setup and Safeguards — Set up screening only after role criteria are confirmed.
• Questions to Ask a Recruitment Agency About Its AI Screening — See what clients may ask about later screening decisions.
• Recruitment Agency KPIs: A Weekly Scorecard From Your CRM Export — Track whether better briefs improve the search process.
• AI Talent Recruiting Software: Manatal, Loxo or Bullhorn Amplify? — Manatal from $15 a user, Loxo from $149, Bullhorn from $99 on its small-agency plans: what each costs a five-person agency and what the AI does.
• Free AI Tool for Recruitment: What a Small Agency Gets for Nothing — What free AI plans and notetakers cover for a small recruitment agency, where free limits and candidate-data risks bite, and when paying makes sense.
• AI Agents for Recruitment Agencies: Sourcing, Screening and Review Points — What an AI agent can take off a recruiter's desk, the four points where a person must check its work, and what agent credits cost per role.
• AI Boolean Search Generator for Recruiters: Prompts and Checks — How to brief an AI for a Boolean search string, what a hiring manager should review, and how to test the string on three sites and log false positives.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Indeed Employer Content Team, How to Conduct an Intake Meeting (last updated June 2025, viewed 4 November 2026). The vacancy, salary figures, timings and field counts are illustrative.