LensReading which lens this session carries.

Sub-function

Build

Owns change work end to end inside Engineering. Runs 3 lines across 16 stations, drains the repeatable share into no shared tower, and holds 4 human gates.

orchestrator Build Orchestratorhuman owner Director, Platform Engineeringno shared tower — judgment density too high

Estate rating

A

75 / 100

autonomy 76%cap 86%

Code generation and dependency maintenance are near-touchless. A human still approves every merge into a protected branch — that gate is the whole control.

Agents

8

Lines

3

Units / wk

15,527

Timetable vs actual

19.7h → 22.6h

Cost / unit

$15

was $40

Run-rate saving

$20.3M

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.

Build Orchestrator

Owns the build lane end to end. Sequences the other agents, holds the timetable, and decides what surfaces to a human.

ObserveA394 / 24h93%9% ovr$1.38/task

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

TaskExecutes stations on a line.

Specification Agent

Executes the build step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA44,069 / 24h97%3% ovr$0.57/task

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

Code Generation Agent

Executes the build step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA43,363 / 24h90%3% ovr$0.46/task

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

Scaffold Agent

Executes the build step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA41,917 / 24h91%3% ovr$0.34/task

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

Refactor Agent

Executes the build step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA4611 / 24h89%8% ovr$0.27/task

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

Documentation Agent

Executes the build step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA44,195 / 24h94%6% ovr$0.59/task

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

Dependency Update Agent

Executes the build step it owns, posts its evidence, and hands the unit to the next stage.

WorkflowA43,123 / 24h92%8% ovr$0.19/task

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

ChallengerPaid to disagree before a human has to.

Build Challenger Agent

Adversarial reviewer for build. Argues the opposite case on every position and flags where the evidence does not carry the claim.

StrategyA265 / 24h88%13% ovr$5.69/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 Coverage72
Evidence Quality73
Override Discipline82
Data Integrity86
Recovery Readiness73
Cost Transparency85
Judgment Readiness61

People

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

headcount on this work80.140.7 FTE

Redeployed into customer-facing escalation ownership.

inference cost per unit$3.40

Lines running in this sub-function

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

12 clear4 evidenced3 held
EN3AFeature BuildTransactional · 6 stations

Refine specification

1 live

Generate code

26 q

Self-test

2 live

Open pull request

2 live

Human review

0 q

Merge

1 live

EN3BDependency UpdateTransactional · 6 stations

Detect update

22 q

Assess breaking change

2 live

Patch

1 q

Test

1 live

Auto-merge low risk

1 live

Escalate

2 live

EN3CDocumentation SyncTransactional · 4 stations

Detect drift

1 live

Regenerate

3 live

Owner review

2 live

Publish

1 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.

EN3-POS-100

Build Operating Position

Draftv4

Is the current cost per change defensible against external comparators once inference spend is counted in full?

Hold the ceiling at A4 for build. Code generation and dependency maintenance are near-touchless. A human still approves every merge into a protected branch — that gate is the whole control. Raise it only after two consecutive quarters where the override rate stays under five percent and every override has a written cause.

confidence

87%

The challenge — Build Challenger Agent

Build Challenger Agent argues the recommendation leans on 5 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 10 percent for two consecutive months
  • ·Cost per change stops falling while volume keeps rising
  • ·A control failure in this lane reaches a customer or a regulator

Evidence gaps

  • ·No external comparator on a like-for-like unit definition
  • ·Exception cases under $29k are sampled, not fully measured
Alternative consideredCostRiskVerdict
Hold the current position$374k run rateKnown and priced. Cedes ground if comparators move faster.live
Raise the ceiling one level now$1480k to build controlsOverride rate is 10 percent; raising the ceiling before that settles imports the error into production.rejected on evidence
Move the exception tail to the spine$378k transitionLoses local context. Rework risk on the cases that are hardest to recover.under review
authored by Build Orchestratorratifier Director, Platform Engineering3 stated assumptions

Policy register

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

RefRuleScopeOwner agentApproved byVersionStatus
EN3-POL-200

Build Decision Envelope

Agents in this lane may act without a human when the change 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 6 agents · Global

envelopeBuild OrchestratorDirector, Platform Engineeringv3.2Active
EN3-POL-201

Build 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 6 agents · Global

evidenceBuild OrchestratorDirector, Platform Engineeringv1.0Active
EN3-POL-202

Build 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, Platform Engineering owns the response clock.

binds 6 agents · Global

escalationBuild OrchestratorDirector, Platform Engineeringv2.9Active
EN3-POL-203

Build Autonomy Ceiling

The ceiling for this lane is 86 on the platform scale. Code generation and dependency maintenance are near-touchless. A human still approves every merge into a protected branch — that gate is the whole control. The ceiling is a governance decision, not a technical limit, and only Director, Platform Engineering can propose moving it.

binds 6 agents · EU · UK

ceilingBuild OrchestratorDirector, Platform Engineeringv5.0Active