Book a Demo Try Speak AI Free Login
Case study · Recruiting & HR · A recruiting firm

How a Recruiting Firm Writes Candidate Reports Faster With Interview Transcripts

A recruiting firm ran its candidate interviews through Speak AI, turned each transcript into the same set of report sections with saved prompts, and compared candidates across a folder.

Book a DemoTry Speak AI Free
Measured from Speak transcript records, June 2023 to June 2025
  • 150+
    Hours of interviews analysed
  • 230+
    Interviews processed
  • 37+ minutes
    Average interview length
  • Small team
    Using it

Summary

A recruiting firm interviewed candidates at length and then wrote each one up by hand for the client. The writing was slow, it varied from recruiter to recruiter, and the best signals were buried in long recordings.

Recruiters uploaded the interview recordings and got a transcript with speaker labels and timestamps. Saved prompts mapped each transcript to the same report sections, and the report was exported as a document or a spreadsheet. A folder per search let them ask questions across all candidates for a role.

Reports followed one structure whoever wrote them, and shortlisting moved from rereading to asking. The record of the interviews themselves, searchable by phrase, stayed available for the next search.

A woman recruiter in a blazer leans in during an interview in a quiet room, a phone resting on the table.

Three lines to take away

Before

Every candidate report was drafted by hand from notes and memory, and no two looked alike.

With Speak

Transcripts fed saved prompts that filled the same report sections for every interview.

After

Recruiters reviewed and edited a draft instead of writing from scratch, and compared candidates in one folder.

Reports written from scratch, one at a time

The interview was only half the job. The write-up that followed took the longer share of a recruiter's day.

An older man sits alone at a desk in the evening with a notepad and laptop, rubbing his eyes.

Drafting from notes

After an interview, the recruiter turned notes and recollection into a client-ready report, then formatted it. The task repeated for every candidate on every search.

Uneven report quality

Each recruiter had their own way of writing a report. Clients saw different depth and different headings depending on who ran the interview.

Signals buried in audio

The evidence for a good fit, such as examples of how someone had handled a situation, sat deep in a long conversation and was easy to miss on a busy week.

The cost: Hours of drafting and formatting around each interview, which grew heavy at peak hiring periods when many interviews landed together.

The turning point

Same prompts, every interview

The firm tried the workflow on a handful of interviews first. What made it click was seeing a transcript come back already arranged under the headings a client expected: experience, skills, behaviour in the interview and practical points such as pay expectations and relocation.

Two concerns had to be settled before the team trusted it. Recruiters wanted to be sure a report stayed theirs, so the draft was something they reviewed and edited before it went to a client. And the firm wanted every recruiter to get the same structure, which is what saved prompts delivered. Once one prompt set was agreed, it was reused across the team and across searches.

What stayed with the team was how the output changed the conversation with clients. A report built from the same sections each time made comparison between candidates quicker, because a reader knew where to look for each point. Recruiters spent their time on judgement, such as how a candidate's examples fit the role, and less on typing and formatting. Saved prompts also gave newer recruiters a starting structure, so the quality of a report no longer depended on who had the most practice.

How they use Speak AI

A four-step path from recording to client-ready report.

  1. Step 1

    Upload

    Recruiters uploaded each interview recording to a shared workspace as soon as the call ended.

  2. Step 2

    Transcribe

    Speak returned a transcript with speaker labels and timestamps, searchable by phrase.

  3. Step 3

    Draft report

    Saved prompts mapped the transcript to the firm's report sections. The recruiter checked the draft and added their own judgement.

  4. Step 4

    Compare

    Candidates for one role sat in one folder, so chat across the folder could surface patterns and help build the shortlist. Reports exported as documents or spreadsheets.

What took work: The prompts took iteration before the sections read the way clients expected, and recruiters still read each draft against the transcript.

A woman places a phone on a small stand on a table between two chairs in a bright interview room.

Book a Demo

Results

A transcript record of the firm's interviews

Speak transcript records show about 150 hours of recorded interviews across more than 230 interviews.

150+
Hours of interviews analysed
230+
Interviews processed

What the team found, in short

Paraphrased from our own conversations with the team. Not quotes.

Report quality becomes consistent when the structure lives in a saved prompt instead of in each recruiter's habits.

Paraphrased from our calls with the team

Recruiters trust a draft more when they can check every claim against a searchable transcript with timestamps.

Paraphrased from our calls with the team

Comparing candidates in one folder turns shortlisting from rereading into asking the same question of every interview.

Paraphrased from our calls with the team

What is next

A man and a woman sit side by side at a desk looking at one laptop and nodding.
  • Standard reports for every searchTeams like this keep one prompt set per role type, so each new search starts from a known report structure.
  • Ask across a searchOnce all candidates for a role are in one folder, a recruiter can ask which candidates gave evidence for a given skill.
  • Share with clientsExported reports and spreadsheets give clients a consistent summary beside the recruiter's recommendation.

More customer stories

Customer name withheld at their request. Figures are measured from Speak transcript records, June 2023 to June 2025.

Write candidate reports from the interview itself

Book a demo and we will build a report structure on your own interview, from transcript to export.

Book a DemoTry Speak AI Free