Check the input, then compare the complete deliverables
Explore the original materials, item-by-item outcomes and full report below. Attempt the input independently before using the reference to find omissions or disagreements.
Transcripts and rubric are synthetic. No audio processing, real employee assessment, CRM write or client contact was performed.
Original synthetic material. Try the input independently, then compare the reference results.
Original synthetic input
{
"synthetic": true,
"case": "call-transcripts-review",
"data": {
"rubric": [
{
"id": "K1",
"criterion": "Customer explicitly states a need:1 point"
},
{
"id": "K2",
"criterion": "Seller explicitly asks budget and receives a response:1 point, even if not approved"
},
{
"id": "K3",
"criterion": "Next step has an explicit owner and date:1 point"
}
],
"calls": [
{
"id": "C01",
"complete": true,
"date": "2026-10-16",
"lines": [
{
"at": "01:10",
"speaker": "Customer",
"text": "We need to organize customer email."
},
{
"at": "03:00",
"speaker": "Seller",
"text": "Is the budget agreed?"
},
{
"at": "03:20",
"speaker": "Customer",
"text": "Not approved yet."
},
{
"at": "05:00",
"speaker": "Seller",
"text": "We will be in touch later."
}
]
},
{
"id": "C02",
"complete": false,
"missing": "Ending excerpt only",
"lines": [
{
"at": "12:00",
"speaker": "Seller",
"text": "shancha will send product materials on October 20."
}
]
}
]
}
}Processing rules
Score only K1–K3 from the supplied words. Missing evidence in a complete record can score 0; absent coverage in a truncated record is unknown, not 0. Asked about budget differs from approved budget. Actions need an explicit owner/date. Do not generalize to overall employee performance.
Complete reference results under this tutorial’s rules; nothing was written to a third-party service.Scroll horizontally within the table if needed.
| Call / criterion | Score | Evidence | Explanation |
|---|---|---|---|
| C01 / K1 | 1 | 01:10 customer need | Email organization explicitly stated |
| C01 / K2 | 1 | 03:00 / 03:20 budget exchange | Asked; approval still pending |
| C01 / K3 | 0 | 05:00 in touch later | Complete record lacks owner + date |
| C02 / K1 | Unknown | Ending excerpt only | Unseen is not zero |
| C02 / K2 | Unknown | Ending excerpt only | Do not infer budget was never asked |
| C02 / K3 | 1 | 12:00 shancha, October 20 | Explicit action; not completed |
Only the three exercise criteria. Budget approval is not required by K2.
K3 is 1; other criteria are unknown. Do not directly rank it against the complete call.
Call review — synthetic transcripts C01: K1=1, K2=1, K3=0, total 2/3. Need clear; budget unapproved; next commitment vague. C02: K3=1; K1/K2 unknown; no total/ranking. Explicit action: shancha sends materials October 20. [C02 12:00] Practice: replace later with an owner, date and deliverable. No real performance judgment, CRM write or external delivery.
Prepare the materials and check your tools
Start with the practice files
- Download transcripts/rubric and locate evidence.
- C01 is complete; C02 is an ending excerpt.
- This is not speech recognition or personnel evaluation.
Before using real services
- Verify permission, participant agreements and retention scope.
- Start with a small de-identified batch and retain timestamps.
- Agree rubric/purpose and require human review.
Complete one file-based exercise first. For real services, verify the actual account, available tools and operation scope, then follow the acceptance check at each stage.
Download the original input and rules ↓Execution and takeover boundaries
The actions below require the agreed inputs, accounts and scope. Configuring a Routine does not expand that authorization.
Can run automatically
Extract quotes, timestamps, actions and assessable items from supplied transcripts using the agreed rubric.
Requires confirmation
Confirm purpose and scope before processing real recordings, changing the rubric, writing CRM, ranking staff, distributing results or configuring a Routine.
Take over after a failure
Mark truncated transcripts, speaker errors and disputed scores as unknown and stop total-score comparisons. Have a human correct the source, then reassess affected items.
Check coverage and scoring purpose
An ending excerpt cannot prove that an earlier budget question never occurred. Declare coverage before scoring.
Inventory both transcripts, completeness and K1–K3. List missing sections and unscorable criteria. Process text only; do not read recordings, contact clients or add standards.A coverage/rubric agreement.
C02 is explicitly incomplete.
Extract needs and commitments with timestamps
Keep what each speaker actually said. Unapproved budget is not no budget, and a vague promise is not a dated action.
Extract needs, budget, objections, commitments and actions with call id, speaker, time and quotation. Leave unstated owner/date blank. Save evidence.csv.A timestamped evidence table.
C01 budget is unapproved; C02 action remains unfinished.
Score criteria, not gaps in the material
Another reviewer should be able to follow each score. Do not turn missing transcript quality into low performance.
Score K1–K3 only. Use 0/1 for complete C01. For uncovered C02 criteria, write unknown with explanation, not 0. Attach evidence to each score and calculate no incomplete total.Six explained criterion outcomes.
C01 is 2/3; C02 only K3=1, without a false total.
Give specific practice and explicit actions
Advice should correspond to evidence. Two examples cannot establish the week’s most common issue or a person’s overall ability.
Summarize each call’s known evidence, gaps, explicit actions and one practice behavior. For C01 suggest a named, dated next step. Do not rank employees, infer weekly trends or edit CRM.A report and one explicit action.
Every suggestion has source words; no team-wide conclusion.
Review a small live batch with a human
Transcripts may misidentify speakers or miss terminology. Verify key passages before using scores.
Apply the rubric to two authorized de-identified transcripts. Report coverage and uncertain wording. Ask a human to check key timestamps and list differences. Do not upload to other systems, write CRM or distribute client details.A human-reviewable trial.
Scores trace to evidence and uncertainty is retained.
Version the rubric before periodic reports
Changed standards affect historical comparability. Record coverage count and remove duplicate calls first.
Draft a call-review Skill with rubric version, call id, coverage, timestamp citations, unknown values, human review and deduplication. Propose a Routine only. Do not automatically assess employees, write CRM or send reports externally.A reusable review standard.
Different rubric versions are not silently compared; duplicate calls count once.
Find the evidence before choosing the next action
Why does unapproved budget score 1?
K2 checks whether budget was asked and answered, not approved.
An excerpt lacks evidence so it scored 0
Check coverage; uncovered behavior is unknown.
The next step only says later
It lacks the required owner/date.
Can two calls rank a team?
No; retain item-level observations and human review.
Check every item before finishing
These page-local checkboxes reset on refresh. Checking them does not verify a real service connection.
Checked 0 / 6 items
Download the inputs separately from the reference answers
Start with input and rules without showing the reference answers. After acceptance, save the report and complete prompt set.
source-data.json↓Rules and scopeDefinitions, exceptions and delivery requirements.workflow-rules.md↓Item-by-item reference resultsCompare every input with its outcome.expected-results.csv↓Complete reference reportResults, evidence and missing information.reference-report.md↓All stage prompts6 instructions with expected outputs and acceptance checks.practice-prompts.txt↓Check documented capabilities and exercise results separately
Content updated: · Editing a page does not reverify its sources or live workflow.
Documentation review: Unverified · no review date recorded
No case-specific documentation review is recorded. References below are provided for review and do not establish verification.
Live workflow verification: Unverified
Live accounts, external writes, scheduled execution and long-term operation are unverified. Synthetic exercises and page checkboxes do not establish live verification.
Transcripts and rubric are synthetic. No audio processing, real employee assessment, CRM write or client contact was performed.
- Grok Bot: files and results ↗
Check uploaded inputs, actual output paths and delivered files.
- Grok Bot: apps and the shared computer ↗
Check connected accounts, available tools, visibility and user takeover.
- Grok Bot: Skills and Routines ↗
Validate a manual run before saving a method or scheduling it.
Try another small batch of your own material
Retain original input and operation records. Verify that each result can be traced before adding connections or recurring runs.