EM Developer AI Usage Dashboard — Narrative Wireframe (Step 3)

Primary audience: Engineering Manager — the line manager responsible for a team's AI adoption, productivity, and token spend. Evidence: the CEO's 6/22 split created this surface specifically for the EM who needs AI-depth detail that neither the executive AI Adoption Report nor the general EM Workflow Dashboard provides. OBJ-1Give the engineering manager a dedicated surface for understanding how their team uses AI — separate from general workflow.

Tag Legend

Sources

OBJ-n Objectives (01)

DR-n Data-relevance recommendation (02)

Decisions & status

TR-n Design decision (Trent / agent)

G-n Data gap (blocks or degrades)

Review annotations

✦ Takeaway strip (section-level)

⚠ Weak takeaway (needs fix)

[red brackets] Unknown / unconfirmed data

the Dev Lead's Team — Backend Services Work Area · 8 engineers OBJ-7Live within the work-area architecture. Dashboard scopes to the Work Area the manager entered from.

▾ This week (Jun 16–22) Data as of Jun 22, 11:40 PM TR-2Design decision: default to weekly view with explicit period selector and "data as of" timestamp. Satisfies a team member's "trend over snapshot" emphasis (OBJ-5) and the trust requirement (data-basis visible at all times).

1 AI Value Summary OBJ-2Show AI value output by platform and work model. The hero KPIs orient the manager before the breakdowns below. OBJ-5Trend lines over snapshots. Every KPI carries a sparkline and period-over-period delta.

Team VLOOKU

342 +12%

vs 305 prior week

DR-1Data-relevance: Total VLOOKU is Partial/estimated via spendAllocation.valueUnitsPerDollar. Team aggregate available via scope filtering.

VLOOKU per Dollar

$0.42 +8%

normalized: salary + AI cost

OBJ-2VLOOKU per dollar is the normalized productivity metric. Denominator = engineer salary + AI token cost.

AI-Code Share

72% +3pp

of retained output is AI-origin

AI Spend

$4,180 -5%

token cost this week

DR-2Per-developer AI spend available from rollup_subject_tool_period.session_cost_usd_sum. Partial/estimated (pricing heuristics).

AI Adoption Funnel DR-3Adjacent concept from 02: AI adoption funnel (adoption → daily → power user) all Available/Partial at team scope. Natural top-of-page orientation. TR-3Design decision: added adoption funnel as sub-KPI row. Not in original objectives but directly serves OBJ-1 (understanding team's AI usage) and is data-ready.

87.5%

Adopted (7/8)

→

75%

Daily active (6/8)

→

50%

Power user (4/8)

✦ "I can see at a glance whether my team is getting value from AI — the numbers are improving, and I know exactly how deep adoption goes."

2 Platform / Tool Breakdown OBJ-2Show VLOOKU broken down by AI platform (Claude Code, Codex, etc.) and by work model. Platforms should be color-coded consistently across all charts.

AI Spend by Platform — 6 Week Trend TR-4Design decision: chart shows cost by platform (available data) rather than VLOOKU by platform (G1 — not served). Cost is the honest lens here; outcome attribution is aggregate.

Claude Code GitHub Copilot Codex CLI Codex CLI RS

This Week by Platform

Claude Code

$2,170 52%

Codex CLI

$1,050 25%

Copilot

$750 18%

Codex RS

$210 5%

Sessions by Platform

Claude Code892

Copilot614

Codex CLI410

Codex RS102

Total: 2,018 sessions

Tool Efficiency Comparison Aggregate / estimated G1Per-tool VLOOKU requires fact_contribution_session joined to tool dimension — no GraphQL aggregation yet. toolEfficiency uses aggregate outcome attribution. Chart shows efficiency proxy, not true per-tool retained value.

Outcome efficiency by platform — outcome attribution is aggregate; per-tool value pending [G1]

Claude Code

0.48

eff. score

+6%

Copilot

0.39

eff. score

— flat

Codex CLI

0.44

eff. score

+3%

Codex RS

0.36

eff. score

+11%

Per-tool VLOOKU breakdown will replace this panel when G1 is resolved

✦ "I know which AI tools my team relies on, what each costs, and which ones are delivering the most efficient results."

3 Work-Model Breakdown OBJ-2VLOOKU broken down by work model (human only / human w/ AI / agentic). The three categories are the nomenclature decided in the 6/22 session.

Contribution Mix — 6 Week Trend G53-way split (Human Only / Human w/ AI / Agentic) designed for target state. Current data serves 2-way only (human vs all-AI). Agentic sub-detail requires fact_contribution_session data population. TR-5Design decision: show the 3-way model as the target design. V1 falls back to 2-way (Human Only vs "AI-Assisted" covering both copilot and agentic). The 3-way renders once G5 data populates — no redesign needed.

% of retained output by work model [3-way requires G5; V1 shows 2-way]

42%

40%

18%

May 12

45%

39%

16%

May 19

48%

38%

14%

May 26

50%

37%

13%

Jun 2

54%

36%

10%

Jun 9

57%

35%

Jun 16

Agentic Human w/ AI Human Only

This Week

Agentic57%

Human w/ AI35%

Human Only8%

IDE sessions614

Agent sessions1,404

DR-4Adjacent concept: IDE vs agent session split (128/1,833 at company scope; both Available). Maps directly to Human w/ AI vs Agentic framing.

✦ "My team is steadily shifting toward agentic work — 57% this week, up from 42% six weeks ago. The Human Only share is shrinking, which is what I want to see."

4 AI Efficiency Signals OBJ-3Surface prompt quality and rework patterns. the Head of Product: "if developers are doing agentic work but having to redo it frequently, they may need help with prompting." TR-6Design decision: framed as "AI efficiency signals" not "prompt quality." Per data relevance, no true prompt-quality or rework metric exists. Available signals: abandoned-AI share (G2 approval needed), prompt volume, prompts-per-session as iteration proxy.

AI Code Survival Rate G2Mart components (abandoned_ai_sum, retained_ai_sum) exist per-subject per-commit. The rate is intentionally not served via GraphQL — "native rework units are not defined." Needs product approval for rate calculation.

% of AI-generated code retained (higher is better) [rate pending G2 approval]

88%

90%

86%

84%

85%

78%

72%

Team avg: 82% · Names link to developer profile

OBJ-6Support drill-down to the developer profile. Dashboard links to it, not replicates it.

Prompt Volume & Iteration Depth G6No true prompt-quality metric. Prompt "specificity" is a count proxy (prompt_count × 10 / engineers). Prompt text is never inspected (privacy-bounded).

Prompt count trended; high volume + flat VLOOKU may signal struggle

Dots sized by session count. Marcus: high prompts, low survival — investigate. David: low prompts, low survival — different issue.

Waste Cost (Abandoned AI) Estimated DR-5Adjacent concept: spendAllocation.wasteCost ($123K company scope, Partial/estimated). "Cost of abandoned AI work" ties to token-cost narrative and the Head of Product's rework concern.

$680 -8% vs prior 16.3% of total AI spend

Cost per Outcome (retained VLOOKU) DR-6Adjacent concept: spendAllocation.costPerOutcome ($2.75 company scope). Inverse of VLOOKU/$ — "what does each unit of retained value cost?"

$2.45 -11%

lower is better

✦ "I can spot developers who might be struggling with AI — Marcus is generating a lot of prompts but his code survival is low. I should check in on his workflow."

⚠ Why weak: Prompt volume and survival rate are proxies, not true rework or prompt-quality measures. The scatter plot suggests correlation but no causal signal exists in the data. Fix → Keep the visualization but badge it "efficiency proxy" and do not label it "prompt quality." The honest framing is "signals worth investigating" not "diagnosis." G6No path to true prompt-quality measurement with current capture (privacy-bounded: no raw content leaves the workstation).

5 Spend & Cohort View OBJ-4Show budget cohort productivity. Break developers into cohorts by AI budget bracket; show VLOOKU and AI-code % per bracket. TR-7Design decision: budget-tier framing is fully data-blocked (G3+G4). Using spend-based dynamic quartiles instead — ranks developers by actual AI spend and asks "do high-spend developers produce proportionally more?" Same question, available data.

AI Spend by Developer — This Week G3No platform tier/plan/SKU dimension exists. Cannot map developers to standard/premium/max budget brackets. G4Budget/quota facts per developer are Source missing. Cost policy is tenant-global only.

Ranked by total token spend. Budget tier labels [require G3 + G4]

$820

$710

$680

$590

$520

$480

$280

$100

Stacked by platform · hover for per-tool breakdown · names link to developer profile

Spend-Based Cohorts Dynamic quartiles TR-8Design decision: since budget tiers don't exist in the data (G3+G4), cohorts use dynamic spend quartiles. This answers the related question: "do high-AI-spend developers produce proportionally more?"

Developers grouped by AI spend quartile. VLOOKU per cohort. [Budget tier labels deferred — G3]

Top quartile

avg VLOOKU

the Dev Lead, Sarah

$765/wk avg spend

2nd quartile

avg VLOOKU

Marcus, Priya

$635/wk avg spend

3rd quartile

avg VLOOKU

Elena, James

$500/wk avg spend

Bottom quartile

avg VLOOKU

David, Alex

$190/wk avg spend

✦ "I can see that higher AI spend does correlate with more output on my team — top-quartile developers produce 3.3x the VLOOKU of the bottom. That's ammunition for broader AI allocation."

6 Developer-Level Signals OBJ-6Support drill-down to the developer profile. The dashboard carries team-level views; individual detail lives on the existing developer profile surface. TR-9Design decision: this section surfaces the "who needs attention" table. Outlier logic: flag developers who are > 1 SD below team mean on AI survival rate, or who haven't adopted AI, or whose spend/output ratio is anomalous.

Team Overview Sort · Filter

Developer ↓	Adoption	VLOOKU	AI Spend	Survival [G2]	Signal
the Dev Lead	Power User	62 +15%	$820	88%	—
Sarah Chen	Power User	58 +10%	$710	86%	—
Priya Patel	Power User	52 +8%	$590	90%	—
Marcus Johnson	Daily Active	48 +4%	$680	78%	Low survival
James Wilson	Daily Active	44 —	$480	84%	—
Elena Rodriguez	Power User	42 +6%	$520	85%	—
David Kim	Adopted	22 -3%	$280	72%	Low output + low survival
Alex Thompson	Not adopted	14 -5%	$100	—	No AI adoption

✦ "I know exactly which developers to talk to: David needs help with his AI workflow (low survival suggests prompt issues), and Alex hasn't adopted AI tools at all — time for a 1:1."

Not in scope / Deferred: Individual PR-level drill-down (lives on developer profile) · True prompt-quality analysis (privacy-bounded; no path with current capture — G6) · Budget-tier cohorts (requires platform tier mapping G3 + budget facts G4) · Cross-session rework detection (no mechanism — G7) · Model-level breakdown within tools (model identity inconsistently captured) · Executive-level AI adoption metrics (belongs on AI Adoption Report) · Activity patterns / work volume (belongs on EM Workflow Dashboard)

Coverage Cross-Check

Objective	Where answered	Data status
OBJ-1 Dedicated AI usage surface for EM	Entire dashboard — separate surface from Workflow Dashboard	Architecture decision
OBJ-2 VLOOKU by platform	§2 Platform/Tool Breakdown (cost lens); §2 Tool Efficiency (proxy)	Cost available; per-tool VLOOKU blocked (G1)
OBJ-2 VLOOKU by work model	§3 Work-Model Breakdown	2-way available; 3-way pending G5
OBJ-3 Prompt quality / rework	§4 AI Efficiency Signals	Proxies only — G2, G6, G7
OBJ-4 Budget cohort productivity	§5 Spend & Cohort View (spend-based alternative)	Budget tiers blocked — G3, G4; spend quartiles available
OBJ-5 Trend lines over snapshots	Every KPI carries sparkline + delta; §2–3 show 6-week trends	Complete — period rollups available
OBJ-6 Drill-down to developer profile	§4, §5, §6 — every developer name links to profile	Navigation supported; surface exists
OBJ-7 Work-area architecture	Dashboard chrome — scoped to EM's Work Area, period selector, "data as of"	Work Area machinery natively supports

All 7 objectives addressed. 3 partially served (G1, G5 degrade but don't block). 2 require alternative framing (G2–G4, G6–G7).

Open Questions

G1 — Per-tool VLOOKU aggregation. Can engineering prioritize a GraphQL aggregation over fact_contribution_session × tool? Without it, the platform breakdown is cost-only. Blocks: §2 tool efficiency → true per-platform value
G2 — AI retention/abandonment rate approval. The mart components exist. Will product approve serving abandoned_ai_sum / (retained_ai_sum + abandoned_ai_sum) as a rate? the Head of Product's ask depends on it. Blocks: §4 survival rate chart
G3 + G4 — Budget tier mapping & budget facts. Is there a plan to capture platform tier/SKU or per-developer budget/quota data? If not, spend-quartile cohorts are the permanent alternative. Blocks: §5 budget-tier labels
G5 — Session-aware contribution population. When will work_branch_session_scores populate so the 3-way work-model split activates? The mart and GraphQL are structurally ready. Blocks: §3 Agentic sub-split
Outlier thresholds. What defines "low survival" or "high spend / low output" for the §6 signal column? Proposal: > 1 SD below team mean on survival, or bottom quartile VLOOKU + top quartile spend. Needs EM validation. Affects: §6 developer signals
Companion surface boundaries. Confirm: this dashboard does NOT show activity patterns (hours, commit cadence, PR cycle time) — those live on the EM Workflow Dashboard. And it does NOT show org-wide adoption funnels or acceleration factor — those live on the AI Adoption Report. Any overlap to resolve? Affects: scope

EM Developer AI Usage Dashboard TR-1Design decision: dedicated surface split from EM Workflow Dashboard. the CEO explicitly separated AI-depth charts from activity/workflow patterns (6/22 session). This dashboard answers "how well is my team leveraging AI."

Tag Legend

Sources

Decisions & status

Review annotations

Team VLOOKU

VLOOKU per Dollar

AI-Code Share

AI Spend

AI Spend by Platform — 6 Week Trend TR-4Design decision: chart shows cost by platform (available data) rather than VLOOKU by platform (G1 — not served). Cost is the honest lens here; outcome attribution is aggregate.

This Week by Platform

Sessions by Platform

Tool Efficiency Comparison Aggregate / estimated G1Per-tool VLOOKU requires fact_contribution_session joined to tool dimension — no GraphQL aggregation yet. toolEfficiency uses aggregate outcome attribution. Chart shows efficiency proxy, not true per-tool retained value.

This Week

AI Code Survival Rate G2Mart components (abandoned_ai_sum, retained_ai_sum) exist per-subject per-commit. The rate is intentionally not served via GraphQL — "native rework units are not defined." Needs product approval for rate calculation.

Prompt Volume & Iteration Depth G6No true prompt-quality metric. Prompt "specificity" is a count proxy (prompt_count × 10 / engineers). Prompt text is never inspected (privacy-bounded).

Waste Cost (Abandoned AI) Estimated DR-5Adjacent concept: spendAllocation.wasteCost ($123K company scope, Partial/estimated). "Cost of abandoned AI work" ties to token-cost narrative and the Head of Product's rework concern.

AI Spend by Developer — This Week G3No platform tier/plan/SKU dimension exists. Cannot map developers to standard/premium/max budget brackets. G4Budget/quota facts per developer are Source missing. Cost policy is tenant-global only.

Spend-Based Cohorts Dynamic quartiles TR-8Design decision: since budget tiers don't exist in the data (G3+G4), cohorts use dynamic spend quartiles. This answers the related question: "do high-AI-spend developers produce proportionally more?"

Team Overview Sort · Filter

Coverage Cross-Check

Open Questions