Skip to content

inbox.md

Latest commit

 

History

History
187 lines (134 loc) · 6.97 KB

inbox.md

File metadata and controls

187 lines (134 loc) · 6.97 KB
Raw

The coaching inbox (rep / team / company)

The coaching inbox is a prioritized "what to improve" digest built by aggregating many call coaching reports. It exists at three scopes:

Scope Audience What it contains
rep The individual seller or CSM Personal patterns to fix this week
team Sales manager, enablement Patterns across all reps on a team
company VP of Sales, Revenue Enablement Roll-up across all teams

The rep inbox is the daily morning read: open it, know your top two focus areas, go make calls. The team and company inboxes tell managers where to direct group coaching and what to reinforce.

How it works

The inbox is fully deterministic — no LLM is involved. It rolls up the structured signals already present in each coaching report (criterion scores and flagged improvements) using a straightforward algorithm:

  1. Group by criterion. For each criterion that either appeared as a flagged improvement or scored below 60 in any report in the period, accumulate the evidence: which calls, which reps, what better-moves were suggested, and what scores were recorded.

  2. Rank by frequency × weakness. Each focus area is ranked by evidence_count × (100 − avg_score). Criteria that appear often and score poorly rise to the top; a criterion seen once at a score of 58 stays low.

  3. Set priority. A focus area is high if it appears in 3+ calls or has an average score below 45; medium if seen in 2+ calls; low otherwise.

  4. Attach the drill. Each focus area carries the best better_move from any coaching report that contributed to it — a concrete, copy-pasteable example the rep can rehearse or the manager can use for a role-play.

  5. Surface wins. Strengths that appear across multiple calls are collected as a "keep doing" list, so positive patterns get reinforced, not just gaps highlighted.

  6. For team/company scope, each focus area lists the affected_reps who show the pattern, and a per-member summary table is appended.

Because no LLM is called, an inbox run is cheap enough to schedule daily.

CLI

Rep inbox

gtmsi inbox <folder> --scope rep --for "Jordan"

<folder> can contain raw transcripts, existing coaching-report JSONs, or both. For a rep inbox, every file in the folder is treated as belonging to that rep. The --for name is stored in generated_for in the output.

Team inbox

gtmsi inbox <team-folder> --scope team --for "AE Team"

For team and company scope, each subfolder is treated as one rep (the subfolder name becomes the rep name). Put Jordan's calls in team-folder/jordan/ and Alex's in team-folder/alex/, and the team inbox will attribute patterns correctly.

Company inbox

gtmsi inbox <company-folder> --scope company --for "company"

For both team and company scope, each immediate subdirectory of the path is treated as one rep (the same one-level code path as team). Subdirectories are not recursed into — only the first level below the path becomes a rep.

Accepts transcripts or existing reports

Pass a folder of raw transcripts (.vtt, .srt, plain text) and GTM Superintelligence will run call coaching on each before building the inbox. Pass a folder of already-computed coaching-report JSONs and it skips directly to aggregation. Mix both freely.

Example output

From examples/reports/team_inbox.md:

# Team coaching inbox — AE Team

**Period:** last 3 calls  ·  **Calls:** 3  ·  **Avg score:** 59.7/100

## Work on this

### Clear Next Steps with the Right People  (seen in 1 calls · avg 38.0)
After excellent discovery, the call ended with 'send me a summary' — an open
loop with no date and no commitment. A document is not a next step; a meeting
on the calendar is.
_Reps:_ Jordan

  **Drill / better move:** "I'll have that summary to you Thursday. Rather than
  you reading it cold, can we grab 30 minutes Friday so I can walk you and your
  VP of Finance through it?"

### Booking the Meeting  (seen in 1 calls · avg 38.0)
_Reps:_ Alex

  **Drill / better move:** "How about we grab 15 minutes Wednesday at 10 so I
  can show you, rather than an email that gets buried?"

## Keep doing

- Current State & Why It Isn't Working  (in 1 calls)
- Compelling Event & Timing  (in 1 calls)

## By member

| Name   | Calls | Avg  | Top focus                        |
|--------|------:|-----:|----------------------------------|
| Riley  |     1 | 54.0 | Interactivity & Comprehension    |
| Alex   |     1 | 55.0 | Booking the Meeting              |
| Jordan |     1 | 70.0 | Decision Process & Stakeholders  |

Running it every morning

Because the inbox is deterministic and cheap, it is practical to run it on a schedule and deliver each rep their personal digest automatically.

A few patterns that work:

  • Cron / CI job — schedule gtmsi inbox rep-folder/ --scope rep --for "Jordan" for each rep, then pipe the output to a DM, email, or Slack message.
  • Claude Code /loop — use the /loop skill to re-run the inbox on an interval and surface the result in your current session.
  • Scheduled agent — use the /schedule skill to create a daily remote agent that runs the inbox and posts results to a channel.

Keep delivery tool-agnostic: the inbox command outputs a JSON document conforming to schemas/inbox.schema.json (or a markdown report if you pass --format md). Pipe it wherever your team already reads their morning updates.

Output schema

The inbox output conforms to schemas/inbox.schema.json. Key fields:

Field Type Description
scope rep | team | company Who this was generated for
generated_for string Rep name, team name, or "company"
period string e.g. "last 20 calls" or "2026-05-23..2026-05-30"
stats.calls_analyzed integer How many calls fed this inbox
stats.avg_overall_score number Average overall score over the period
focus_areas[] array Prioritized improvement themes
focus_areas[].title string The criterion name
focus_areas[].priority high | medium | low Computed from frequency and score
focus_areas[].evidence_count integer How many calls showed this pattern
focus_areas[].avg_score number Average criterion score over the period
focus_areas[].drill string A concrete practice or better-move
focus_areas[].example_calls string[] Call ids/titles illustrating the pattern
focus_areas[].affected_reps string[] (team/company only) Who shows this pattern
wins[] array Positive patterns worth reinforcing
members[] array (team/company only) Per-rep summary lines

Cross-references

  • Per-call coaching reports: scorecards.md
  • Deal and account health (the other aggregations): scoring-layers.md
  • CRM auto-fill: crm.md
  • Schema: schemas/inbox.schema.json
  • Source: src/gtmsi/inbox.py