LensReading which lens this session carries.

Function cockpit

Engineering

Agents build, test and document. Engineers own architecture and release.

Autonomy0% / 80%

Phase: Phase 3 — Functional orchestration

Run the intent-to-release graph as a governed system: decomposition, code generation, testing, documentation, vulnerability detection and deployment preparation are agent work; product intent, architecture and production release remain human.

Trust score

0

Touchless

0%

Human review

0%

Override rate

0%

Capacity released

25-40% engineering capacity in selected software-factory settings (case-specific)

Accountable human

Chief Technology Officer

CoS agent: Engineering Chief of Staff

Autonomy, trust, and throughput at a glance

Health

Is this function holding

Four readings with a printed rule behind each, then the outcome measures the function is judged on. A reading without a rule is decoration.

Autonomy against target

71%

9 points short of the 80% target

Trust score

85

at or above the 85 floor for a rung promotion

Touchless rate

64%

36% of items still take a person somewhere in the run

Override rate

9%

people are reversing the agents often enough that the ceiling is doing real work

Lead time to change

-58%

2.4days

Approved intent to production-ready package

Escaped defects

-11

7/release

Defects found after release, rolling four releases

Model-generated code

+9

46%

Share of merged lines authored by agents under review

Change failure rate

-2.6

4.1%

Releases requiring rollback or hotfix

Open critical findings

-8

3items

Security findings blocking promotion

Operations

What the function is running

The roster and where it sits on the ladder, the split between what runs alone and what a person still signs, and a slice of the floor as it stands.

Where this roster sits on the autonomy ladder

A0 assisted through A4 autonomous. Moving an agent up a rung is a governance decision, not a config change.

Loading ladder…

Agents

11

Active now

10

Tasks / 24h

10,976

Mean success

92%

Work mix, as it stands

How the function's volume divides between the agents and the people.

Runs end to end without a person

64%

cleared inside the ceiling, no queue, no signature

A person reads it before it clears

27%

evidence posted, a named reviewer signs

A person reverses the agent

9%

the agent proposed, the human decided otherwise

Capacity released

25-40% engineering capacity in selected software-factory settings (case-specific)

IntentDesignBuildVerifyRelease

What runs without a person

Committed to autonomy inside a defined ceiling.

  • Code scaffolding and bounded change generation
  • Test suite creation, execution and defect reproduction
  • Design and operating documentation refresh
  • Static analysis and low-risk dependency updates
  • Sandbox deployment and approved remediation

What stays with people

Judgment, accountability, and anything a regulator would ask a human about.

  • Architecture and platform standards
  • Requirements acceptance and product intent
  • Safety-critical code and threat decisions
  • Breaking changes and public API contracts
  • Production release and irreversible data migration

On the floor right now

A slice of live work. The full board carries every item.

  • ENG-7412

    Release pack — payments service v4.2

    Release · Release manager · 54m old · 14 services

    Awaiting human
  • ENG-7413

    Regression suite regeneration — billing module

    Verify · Engineering CoS · 18m old · 1,840 cases

    Autonomous
  • ENG-7414

    Legacy translation — interest calculation engine

    Build · Principal architect · 6h 12m old · 31K LOC

    Escalated
  • ENG-7415

    Story decomposition — partner onboarding intent

    Intent · Product owner · 1h 36m old · 18 stories

    Drafted
  • ENG-7416

    Incident remediation proposal — checkout latency

    Verify · Service owner · 27m old · Sev-2

    Awaiting human
  • ENG-7417

    Dependency uplift batch — internal libraries

    Build · Engineering CoS · 3h 25m old · 23 packages

    Complete

Actions

What this function is asking a person to do

Ordered by size. Each figure is a live count from this function's own work and governance records, not a target.

Brakes available right now

What a named human can pull today to stop this function, without waiting for an engineer.

1 agent not active
  • Unexplained dependency introduced into the build
  • Secret or credential exposure detected in a diff
  • Test bypass or coverage suppression
  • Architecture-policy violation in a merged change
  • Rising incident rate after an agent-prepared release

Live observability

What has actually been decided, and where each agent stops

A dashboard that shows only outcomes hides the decisions that produced them. This is the governance record as written, and the ceiling every agent is held to.

Governance record

Most recent first. Each entry names the actor and the call.

4 entries
  • Payment service release pack awaiting authorization

    Pending

    Release Agent assembled the deployment plan, rollback rehearsal evidence and test attestation for the payments service. Two moderate findings remain open and are documented in the pack.

    approval · Release manager · materiality high

  • Secret detected in candidate branch

    Contained

    Security Agent detected a live credential in a feature branch, blocked promotion, revoked the token through the identity service and opened an incident record.

    escalation · Security Agent · materiality high

  • Dependency batch updated autonomously

    Auto-executed

    Developer Agent applied 23 low-risk dependency updates with full regression passes and provenance recorded against each change.

    notify · Developer Agent · materiality low

  • Legacy migration equivalence gap

    Pending

    Migration Agent found behavioral divergence in an interest-calculation routine during translation. Work paused and the divergence packaged for the owning architect.

    escalation · Principal architect · materiality medium

Escalation ceilings

Past the line the agent stops and hands the decision to the named human with the evidence attached.

  • A4

    Engineering Chief of Staff

    Ceiling: Coordination only — no merge or release authority

    Then: Any architectural conflict or release-evidence gap escalates to the CTO before the release window.

  • A3

    Requirements Agent

    Ceiling: Drafting only; acceptance remains with the product owner

    Then: Ambiguous or conflicting intent is returned to the product owner rather than resolved by inference.

  • A3

    Architecture Analyst

    Ceiling: Advisory only; no architecture decision authority

    Then: Any change crossing a published architecture boundary routes to the architecture review board.

  • A2

    Developer Agent

    Ceiling: Max 400 changed lines per unit of work; no schema or public API change

    Then: Security-sensitive paths, migrations and public interfaces require named senior approval before merge.

  • A3

    Test Agent

    Ceiling: Non-production environments only

    Then: Coverage regression beyond the published threshold blocks the pipeline and notifies the release manager.

Is policy and strategy coming to fruition

Does the intent above this function reach the work inside it

Counted per sub-function, where each step only counts if the step before it did. A policy that never reaches a running workflow has not landed, however well it reads.

Chain from stance to transaction

6 of 11 sub-functions carry a ratified position, an active policy and a live workflow

Ratified position

6

of 11 sub-functions — a stance the leadership signed, not a draft

...and an active policy

6

a rule in force, with a version and an approver

...reaching a live workflow

6

the rule reaches something that actually runs

...and landing in a GBS tower

4

the run is executed on the shared transaction spine

5 of the 11 sub-functions are executing at volume without the full chain behind them. They run to a general standard rather than to a rule with a version and an approver, and the gap first appears at the position step.

What this page is, and what it is not

Every reading here is computed live from this function's own records, which makes it exact and makes it narrow. Volumes, unit costs, autonomy and trust are modeled: none has been reconciled against an enterprise resource planning system, a service management tool or a payroll register. Read the outcome measures as the shape of the argument, not as an audited result.

Deloitte's Agent-Orchestrated Development Life Cycle shifts engineers toward orchestration, architecture and validation. BCG Platinion reports 3-5 times productivity in selected agentic software-factory settings, while Accenture reports 30% efficiency improvement in certain migration work and a 50% reduction in AI-application build time across more than 75 deployed use cases. These are case-specific and not general software-engineering averages.