Book a Demo Try Speak AI Free Login
Case study · Professional services · A coaching and consulting firm

How a Coaching and Consulting Firm Scores Every Client Session Against Its Own Method With AI

A coaching and consulting firm for small practices stopped watching, transcribing and grading every client video by hand, and now has a trainer-persona prompt score each session against its own scorecard, with a manager correcting the scores.

Book a DemoTry Speak AI Free
As of October 2026Measured from Speak usage records
  • 20,000+
    Recordings processed
  • About 2,700
    Hours of audio and video
  • 170+
    Branded recorders configured
  • 50+
    Fields set up for scoring and session details

Summary

Every client video was watched, transcribed and graded by a person against the firm's own scorecard. Each session took hours, so the number of clients was tied to the number of reviewers.

The firm captures sessions through recorders, a meeting assistant and an automated upload feed. A prompt written in the voice of its trainer scores each one against the scorecard, and a manager corrects the result.

Scoring now runs in volume across hundreds of recordings a quarter, and each client gets its own white-label library of sessions and scores.

A woman coach in a blazer gestures confidently while delivering a session in front of a phone on a tripod.

Three lines to take away

Before

Each client video was watched, transcribed and graded by hand against the firm's scorecard, taking hours per session.

With Speak

A prompt in the trainer's voice scores each recording against the same scorecard, and a manager corrects what it gets wrong.

After

Scoring runs across a large library with a consistent first pass, and clients see their own sessions in a white-label space.

Every session was graded by hand

The firm's value sits in its method, and applying that method to each recording was slow, manual work.

A man works late at a desk, watching a video on a laptop while handwriting notes beside a stack of notebooks.

Hours per session

A reviewer watched the video, produced a transcript, then graded it against the scorecard. The time per session grew with the number of clients.

Scoring varied by reviewer

Different people read the same scorecard slightly differently, so two clients could be graded on different standards for the same behaviour.

Delivery lagged

Feedback reached clients after the session had faded from memory, and backlogs formed whenever new clients joined.

The cost: Reviewer hours, not demand, set the ceiling on how many client sessions the firm could grade and how quickly feedback came back.

The turning point

A reviewer's judgment, applied the same way every time

The question for the firm was not whether software could transcribe a session. It was whether a score could reflect the firm's own method rather than a generic rubric. The team wrote a prompt in the voice of its trainer, set out the scorecard in it, and ran it on recordings that had already been graded by hand.

What persuaded the firm was the manager's role. The score is a first pass, and the manager corrects it, so the firm's judgment stays in charge and each correction shows where the prompt needs tightening. After a few rounds of that, the firm trusted the prompt for the bulk of sessions and kept human review for the rest. Client privacy was the other condition: each client had to see only its own recordings, in a library carrying the firm's branding.

How they use Speak AI

Sessions arrive by whichever route suits the client, and the scorecard prompt runs on every one.

  1. Step 1

    Capture

    Sessions come in through branded recorders, a meeting assistant that joins calls, uploads and an automated feed, with most volume arriving by the automated route.

  2. Step 2

    Transcribe

    Each recording is transcribed with speakers labelled, so the prompt and the manager can both work from the text.

  3. Step 3

    Score the session

    A trainer-persona prompt scores the recording against the firm's own scorecard and writes the result into fields that each session carries.

  4. Step 4

    Correct and share

    A manager reviews and corrects scores, then each client sees its own sessions and results in a white-label library.

Honest friction: Getting the scorecard into a prompt and a set of fields took several rounds of testing against hand-graded sessions, and building the fields and automations needed guided help. Volume is lower than the spring peak, and managers still correct scores rather than leaving them untouched.

Two colleagues sit side by side at a desk reviewing a session on a laptop, one taking notes.

Book a Demo

Results

Scoring at the scale of a whole client base

As of October 2026, the firm has more than 20,000 recordings in Speak and about 2,700 hours of audio and video, with scoring and chat running every week.

50+
Fields carrying scorecard results and session details
700+
Automation runs in the last 90 days
500+
Chat sessions in the last 90 days
How we calculated this

An estimate, not a measurement

The usage numbers on this page are measured from Speak records. This block is a separate estimate of the manual effort the same workload would take. It uses published time benchmarks at the low end of their ranges and official wage statistics at the 25th percentile. Most teams already have transcripts, so Speak’s value here is the analysis layer.

Alternative baseline

Starting from: already using a transcription service

Counts only the analysis, grading and notes effort. The conservative baseline.

At least 1,300 hours
about US$46,000 at the low end

Up to about 4,000 hours and US$190,000 at the high end of the published ranges.

Headline for this study

Starting from: no transcription today

Baseline (a) plus manual transcription. The upper bound.

At least 11,000 hours
about US$280,000 at the low end

Up to about 22,000 hours and US$690,000 at the high end of the published ranges.

Which baseline we lead with: The firm's own account in our notes is that each client video was watched, transcribed and graded by hand, so the no-transcription baseline (b) leads.

Task Manual time (published range) With Speak (our assumption)
Watching and grading each recording against the scorecard 1 to 2 hours per hour of recordings
Survey Results (scoring time vs AHT)		 0.5 hours per hour of recordings
Our assumption: 0.5 hours per hour of recording for a manager to check and correct an AI score.
Manual transcription (baseline b only) 4 to 7 hours per hour of recordings
Realities Toolkit #08: Transcribing your own qualitative data		 0.25 hours per hour of recordings
Our assumption: 0.25 hours to check an AI transcript.

Workload (measured): 2,700 hours of recordings. Speak media records for the account (not deleted, each recording capped at 12 hours so two unusually long files are excluded, zero-length rows ignored), pulled 2026-10-11.

Hourly cost: US$24.09 (25th percentile) to US$33.31 (median) for Training and development specialists; transcription is costed at US$16.42 to US$19.43. Both are multiplied by 1.43 to include benefits and employer costs. US wages are used as a neutral reference.

Hours saved = workload x (manual hours – hours with Speak). Value = hours saved x hourly wage x 1.43.

About three quarters of the recordings are short automated uploads, so the hours are mostly session video and calls. Not every recording would have been graded by hand. Equivalent manual effort, not a measured saving.

As of October 2026. This estimate is not customer-reported, is not audited, and does not include software costs.

What the team found, in short

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

A scorecard prompt is most credible when it is tested against sessions the firm has already graded by hand and corrected by the person who owns the method.

Paraphrased from our calls with the team

Most of the volume comes through automated feeds, which suggests firms like this score when sessions arrive, not when someone remembers to upload.

Paraphrased from our calls with the team

Giving each client a private white-label library turns the scoring work into something the client can see, which makes the service easier to renew.

Paraphrased from our calls with the team

What is next

A manager and a client go over a laptop together in a sunlit meeting nook, smiling.
  • Trend viewsTeams like this often chart scores over time per client, so progress is visible without opening individual sessions.
  • Prompt tuningEach manager correction is a signal for tightening the prompt, which gradually reduces how much needs correcting.
  • Client dashboardsA shared dashboard gives each client a summary of its own scores and themes next to the recordings.

More customer stories

Customer name withheld at their request. Figures are from Speak usage records, as of October 2026.

See how your own scorecard could run on every session

Talk with us about turning your method into a prompt, scoring recordings in volume and sharing results with clients.

Book a DemoTry Speak AI Free