LensReading which lens this session carries.
Engineering/Code Review & Quality

Sub-function

Code Review & Quality

Owns pull request work end to end inside Engineering. Runs 3 lines across 14 stations, drains the repeatable share into no shared tower, and holds 5 human gates.

orchestrator Code Quality Orchestratorhuman owner Director, Engineering Qualityno shared tower — judgment density too high

Estate rating

BBB

75 / 100

autonomy 82%cap 88%

Analysis and comment generation are fully agentic and never get tired. Trivial changes auto-approve; anything touching security or data paths requires a human reviewer.

Agents

8

Lines

3

Units / wk

9,574

Timetable vs actual

14.6h → 15.4h

Cost / unit

$14

was $47

Run-rate saving

$16.2M

annualized at this volume

The 8 agents running this sub-function

Grouped by what they are for, not by where they sit. The agent org is process-shaped.

OrchestratorOwns the queue and arbitrates between desks.

Code Quality Orchestrator

Owns the code review & quality lane end to end. Sequences the other agents, holds the timetable, and decides what surfaces to a human.

ObserveA3146 / 24h90%9% ovr$10/task

ceiling A4 · escalates to Escalates to Director, Engineering Quality when the unit falls outside the decision envelope or confidence drops below the floor.

TaskExecutes stations on a line.

Static Analysis Agent

Executes the code review & quality step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA41,325 / 24h92%5% ovr$0.29/task

ceiling A4 · escalates to Escalates to Director, Engineering Quality when the unit falls outside the decision envelope or confidence drops below the floor.

Review Comment Agent

Executes the code review & quality step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA41,502 / 24h93%3% ovr$0.20/task

ceiling A4 · escalates to Escalates to Director, Engineering Quality when the unit falls outside the decision envelope or confidence drops below the floor.

Coverage Agent

Executes the code review & quality step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA42,162 / 24h93%7% ovr$0.19/task

ceiling A4 · escalates to Escalates to Director, Engineering Quality when the unit falls outside the decision envelope or confidence drops below the floor.

Convention Enforcement Agent

Executes the code review & quality step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA4863 / 24h97%2% ovr$0.58/task

ceiling A4 · escalates to Escalates to Director, Engineering Quality when the unit falls outside the decision envelope or confidence drops below the floor.

ObserveWatches signals and watches the agents.

Complexity Monitor

Watches code review & quality continuously. Detects drift, quantifies it, and routes what matters without waiting for a reporting cycle.

ObserveA494 / 24h91%12% ovr$4.17/task

ceiling A4 · escalates to Escalates to Director, Engineering Quality when the unit falls outside the decision envelope or confidence drops below the floor.

PolicyAuthors and versions the binding rules.

Quality Policy Agent

Holds the rules that bind every other agent in code review & quality. Versions them, tests them against live decisions, and blocks work that breaches them.

PolicyA269 / 24h95%17% ovr$5.33/task

ceiling A4 · escalates to Escalates on any decision that would breach a live policy version, with the clause cited.

ChallengerPaid to disagree before a human has to.

Code Review & Quality Challenger Agent

Adversarial reviewer for code review & quality. Argues the opposite case on every position and flags where the evidence does not carry the claim.

StrategyA289 / 24h95%10% ovr$8.13/task

ceiling A4 · escalates to Escalates when a ratified position is still running against a break condition it flagged.

Estate rating breakdown

Seven control dimensions. Judgment readiness runs against autonomy on purpose.

Control Coverage79
Evidence Quality77
Override Discipline76
Data Integrity69
Recovery Readiness90
Cost Transparency79
Judgment Readiness53

People

What the humans stopped doing, and what they do now.

headcount on this work57.740.8 FTE

Redeployed into agent supervision, calibration and coaching.

inference cost per unit$1.78

Lines running in this sub-function

Each line is a workflow. Click a line to walk it station by station.

12 clear4 evidenced4 held
EN4APull Request ReviewTransactional · 6 stations

Static analysis

1 live

Generate review comments

1 live

Score risk

2 live

Auto-approve trivial

1 live

Human review

1 live

Merge

4 q

EN4BQuality Debt SweepTransactional · 4 stations

Scan

4 live

Rank by risk

38 q

File issues

1 live

Owner triage

7 q

EN4CConvention ChangeTransactional · 4 stations

Propose

2 live

Team review

4 live

Ratify

4 q

Enforce

3 live

Strategy positions this sub-function holds

An agent that only executes is a robot. These are the calls it made, the argument against each one, and what would prove it wrong.

EN4-POS-100

Code Review & Quality Operating Position

Ratifiedv2

Should pull request work stay in the shared transaction spine, or come back inside the function where the context lives?

Hold the ceiling at A4 for code review & quality. Analysis and comment generation are fully agentic and never get tired. Trivial changes auto-approve; anything touching security or data paths requires a human reviewer. Raise it only after two consecutive quarters where the override rate stays under five percent and every override has a written cause.

confidence

59%

The challenge — Code Review & Quality Challenger Agent

Code Review & Quality Challenger Agent argues the recommendation leans on 2 quarters of data from a period with no volume shock. The confidence band overlaps the alternative, and the position does not say what it would take to be wrong. Recorded as a dissent, not a block.

Break conditions

  • ·Override rate rises above 11 percent for two consecutive months
  • ·Cost per pull request stops falling while volume keeps rising
  • ·A control failure in this lane reaches a customer or a regulator

Evidence gaps

  • ·No independent test of the control evidence behind the touchless rate
Alternative consideredCostRiskVerdict
Hold the current position$597k run rateKnown and priced. Cedes ground if comparators move faster.not selected
Raise the ceiling one level now$1366k to build controlsOverride rate is 5 percent; raising the ceiling before that settles imports the error into production.rejected on evidence
Move the exception tail to the spine$581k transitionLoses local context. Rework risk on the cases that are hardest to recover.selected
authored by Code Quality Orchestratorratifier Director, Engineering Quality3 stated assumptions
outcome did not hold

Ratified 7 months ago. Measured since: cost per pull request rose 2 percent against a forecast that it would fall 34 percent.

Policy register

The rules the agents above are bound by. Version, owner, approver, and the agents each rule constrains.

RefRuleScopeOwner agentApproved byVersionStatus
EN4-POL-200

Code Review & Quality Decision Envelope

Agents in this lane may act without a human when the pull request sits inside the stated value, risk and confidence envelope. Outside it, the unit holds at a gate with a named approver and a running clock.

binds 4 agents · US · Canada

envelopeQuality Policy AgentDirector, Engineering Qualityv3.7under review
EN4-POL-201

Code Review & Quality Evidence Standard

Every autonomous decision writes inputs, the rule version applied, the model and prompt version, the output and a reversal path. An action with no evidence record is treated as a control failure, not a fast decision.

binds 4 agents · Global · EU

evidenceQuality Policy AgentDirector, Engineering Qualityv4.1Active
EN4-POL-202

Code Review & Quality Escalation Rule

Escalation is mandatory when confidence falls below the floor, when two options sit inside the confidence band, or when a break condition on a ratified position fires. Director, Engineering Quality owns the response clock.

binds 4 agents · LATAM · US

escalationQuality Policy AgentDirector, Engineering Qualityv2.1Active