Edge8 · Plan · Roadmap item 03

The coaching mirror: every one-on-one remembered, a fair review every quarter.

An AI-drafted quarterly review for every person, built from a quarter of one-on-ones and the work itself. Coaches edit it, and each person sees theirs in the room.

Prepared 30 October 2026 · Owner: Dave Hajdu · Coaches: Dave, Khoa, Quan

1

Define the problem

Who feels it, what it costs, why now.

After a quarter of one-on-ones the coaching database holds more than any coach can remember. There's no full picture of each person's progress and no fair basis for a quarterly review. Meanwhile AI is changing the business fast and people who feel stuck in their AI growth leave or slow down, so we take on fewer projects while the price we can charge keeps falling.

6Dave's direct reports, a one-on-one every two weeks, about 3 a week
5people left in 2026
3of those were departures we regret
~2 moof productivity lost per departure
Why now

The new coaching database has its first full quarter of data, and price pressure means the team has to grow its capacity without growing headcount.

2

Discover the data

Almost everything already exists in edge8-web. The work is joining it.

SignalWhere it livesStatus
One-on-ones: prep, summary, shared summary, who led it, coach / mentor / direct mixcoaching_one_on_onesExists
Transcripts and recordings (Lark Minutes)meetingsLinks, not text
Commitments and whether they were keptcoaching_commitments, coaching_commitment_historyExists
Goals, priorities, things noticedcoaching_priorities, coaching_noticed, goalsExists
Trend reportscoaching_trendsExtend
Reviews, ratings and self-assessmentperformance_reviewsExtend
Feedbacktraining_learner_feedback, coaching_goal_commentsExists
Workboard cards and time spent blockedtasks, task_stage_logExists
Human and AI tokenshtt.token_entriesExists
Time offtime_offExists
AIO Labz progress and certificationAIOLabz Database: enrollment, activity_completion, micro_session_completion, submissionNew link
New data needed

A link from AIO Labz users to Edge8 people, probably by email, and a review status on performance_reviews: draft, approved, discussed. The transcripts need checking: if only links are stored, the draft reads them from Lark Minutes.

3

Design the workflow

Fully transparent. The review is opened together, never sent ahead. Three screens below.

Team member

Fills in a self-assessment, then sees their own review for the first time in the one-on-one.

Coach (Dave, Khoa, Quan)

Sees their own people and gets a private mirror on their own coaching style.

Dave, as the boss

Sees everyone. Edits every draft for now and runs calibration with each manager.

Every quarter
  1. Self-assessmentEach member completes the self-assessment tool the week before.
  2. AI drafts the reviewA narrative plus ratings per person, from one-on-ones, commitments kept, Workboard, token ratio, time off, AIO Labz, self-assessment and feedback. Saved as a draft.
  3. Coach mirrorEach coach gets a brief on their own style: coach / mentor / direct mix, follow-through, what to change.
  4. Dave edits and approvesEvery draft, until the managers are ready to take it over.
  5. CalibrationDave and each manager go through that manager's people, once a quarter.
  6. Discussed liveThe coach opens the review with the member in the one-on-one, so it's a conversation rather than something to defend against.
  7. Improves each runCommitments agreed in the review become the first thing next quarter's draft checks.
Wireframe 1 of 3 · Team overview (Dave sees every coach's people)
Quarterly reviews · Q3 2026
18 people · 3 coaches
All coachesDaveKhoaQuan
18 self-assessed18 drafted6 approved2 discussed
PersonCoachCommitments keptAIO LabzToken ratioStatus
Team member ADave83% ↑6 of 81 : 40 ↑Discussed
Team member BKhoa50% ↓2 of 81 : 12 →Draft
Team member CQuan70% →8 of 8 ✓1 : 65 ↑Approved
Team member DDave—4 of 81 : 25 ↑Self-assessment missing

Sort or filter by any column. Clicking a row opens that person's review. Calibration with Khoa or Quan uses the same view filtered to their people.

Wireframe 2 of 3 · One person's review (opened together in the one-on-one)
Team member BQ3 2026 · Coach: Khoa · Draft
EditApproveOpen in one-on-one
The quarter in a paragraph · AI draft, edited by coach

Every sentence links to its source: a one-on-one, a commitment or a card.

Ratings · self vs. coach
AreaSelfCoach Building with AI34 Planning and delivery43 Growth and learning33 Working with the team44

Where self and coach differ by 1 or more, it's highlighted as a talking point.

Commitments for next quarter · agreed in the room
+ Add a commitment
Signals this quarter
One-on-ones held6 of 6 Commitments kept50% ↓ Human-to-AI token ratio1 : 12 → AIO Labz modules2 of 8 Workboard cards done14 Longest blocked card9 days Days off3
Themes from one-on-ones
Client deliveryLearning Claude CodeWorkload
Raised 3+ times
Their self-assessment
Wireframe 3 of 3 · Coach mirror (private to the coach and Dave)
Your coaching · Q3 2026
Khoa · 6 people · 34 one-on-ones
Compare with Q2
How you spent the time
Coach 35%Mentor 25%Direct 40%

Target mix agreed with Dave: mostly coaching.

2–3 things to change next quarter
Follow-through
Your commitments kept75% Team's commitments kept62% One-on-ones held on time94%
Last quarter you said you'd…
✓ Ask before advising
✗ Follow up on blockers within a week

Layout only, with placeholder people and numbers. The final screens use edge8-web's existing coaching components.

Phase 0 · November

Pilot on Q3

  • Draft Q3 reviews from existing data
  • Back-test on past one-on-ones
  • Dave edits; discussed in one-on-ones
Phase 1 · December

Build it in edge8-web

  • Quarterly review view in the coaching area
  • AIO Labz link and review status
  • Coach mirror brief
  • Visibility rules for member, coach and boss
Phase 2 · Q1 2027

Managers own it

  • First full cycle in early January
  • Khoa and Quan edit and hold the meeting
  • Dave keeps full visibility and runs calibration
4

Determine the ROI

The FAST goal, how we measure it and what it returns.

FAST goal

Zero regretted departures from November 2026 to April 2027. 100% of the team AI Officer certified and building fast with AI. $130K a month in revenue with no new hires.

If we hit it: a 200% monthly-salary bonus for the product team and a 25% average raise.

Success measureTypeBaselineTarget
Regretted departuresLagging3 in 20260 in 6 months
Monthly revenue, with headcount flatLaggingCurrent run rate$130K a month
AI Officer certifiedLaggingFrom the AIO database100%
Human-to-AI token ratioLeadingMeasured in the Q3 pilotRising every quarter
Commitments kept by the next one-on-oneLeadingMeasured in the Q3 pilot80%
Reviews done and discussed on timeLeading0100% every quarter
Retention

Three regretted departures at about two months' productivity each is roughly 6 person-months a year. That's capacity the $130K goal needs without hiring.

Coach time

About 10 hours per person per quarter today, down to about 2: edit the draft, then have the conversation. For Dave's 6 people alone, about 60 hours becomes about 12.

Open question: the company key result on record in edge8-web is an average 20% raise and an average $2,000 bonus, and there's no $130K revenue goal. This plan uses Dave's numbers. Update the key result if they're the new targets.

5

Definition of done

What counts as finished, where it lives and who needs to understand it.

  • Q3 pilot done. A draft review for every coached person, edited by Dave and discussed in a one-on-one during November.
  • Back-test done. Drafts run on the one-on-ones of two recent departures we regret, plus one we didn't, to check whether the warning signs were visible early.
  • Shipped in edge8-web. The quarterly review view and coach mirror are live in the coaching area, deployed through git push and CI, with visibility for member, coach and boss.
  • Managers trained. Khoa and Quan have a short guide on running the review conversation, and the first calibration sessions are held.
  • Team trained. A session for everyone on what data feeds the review, who sees what and how the conversation works.
  • First full cycle. Q4 reviews are drafted, approved and discussed in early January 2027.