Executive dashboard

For stakeholders: spend, speed and stability at a glance.

Token spend by model size

Month 6 AI spend
$13,632/ month
$341 per engineer · peak so far $14,710 in month 4
  • Large $11,606
  • Medium $1,752
  • Small $274
Month 6 · Expand · Spend per change × throughput × 6 teams × 4.33 weeks; split by token share weighted by price (1 : 3 : 12)

Velocity

Merged changes per team per week
15.9/ team / wk+14%bettervs baseline
Month 6 · Expand · trend over the year

Cycle time

Median, first commit to production
46h−12%bettervs baseline
Month 6 · Expand · trend over the year

Deploys and change failures

Deploys
3.7/ svc / wk+19%bettervs baseline
Change failure rate
12.4%+0.4 ptswithin noisevs baseline
WatchChange failure rate is drifting the wrong way, still inside noise.
Month 6 · Expand · trend over the year

Delivery health

1 of 4 pairs healthy
  • Cycle time vs change failure rateWatch
  • Deploys vs uptimeHealthy
  • Agent-authored share vs rework rateRegression
  • Throughput vs review loadWatch

DORA 2024: as AI adoption rose, delivery throughput fell an estimated 1.5% and stability 7.2%. The gates on the Board exist to beat that.

Month 6 · Expand · trend over the year

Uptime and error budget

Uptime, trailing 30 days
99.92%−0.01 ptswithin noisevs baseline
SLO
Error budget leftNearly spent
8.6 min of 43.2 left · 34.6 used
Month 6 · Expand · 99.9% SLO over 30 days = 43 min budget

AI adoption

Agent-authored share of merged code
33%+30 ptsbettervs baseline
Rework rate, agent code
9.1%+1.5 ptsworsevs baseline
Month 6 · Expand · trend over the year

Decision rules

1of 6 rules firing
  • Change failure rate above baseline two weeks runningCleared m6
  • Error budget spent on a serviceCleared m6
  • Eval pass rate below the release thresholdNot fired
  • Review latency over one day for two weeksCleared m6
  • Agent code reworked more than human codeFiring
  • A metric is used to rank individualsNot fired
Month 6 · Expand · trend over the year

Add a gadget

  • Token spend by model size
    Monthly AI spend in dollars, stacked by small, medium and large models.
  • Velocity
    Merged changes per team per week, against the baseline.
  • Cycle time
    Median hours from first commit to production.
  • Deploys and change failures
    Deploy frequency next to change failure rate, with the pair verdict.
  • Delivery health
    The four speed and quality pairs, each with its verdict, against the DORA 2024 finding.
  • Uptime and error budget
    Uptime against the 99.9% SLO and the minutes of error budget left.
  • AI adoption
    Share of merged code written by agents, and how much of it is reworked.
  • Decision rules
    Which rules are firing now, and which fired and cleared.
  • Eval pass rate
    Eval pass rate against the release threshold.
  • Cost per change
    AI spend per merged change and prompt-cache hit rate.
  • Rollout phase
    The current rollout phase and progress through the year.
  • Tickets in flight
    The tickets on the Board, with model spend and elapsed time end to end.

Gadgets follow the month picked on the dashboard.