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

Cost to serve per unit including inference, and where the capacity went

Unit economics

What a unit of Engineering work costs to serve

Cost per unit is stated after inference, not before it. Model spend, retrieval, evaluation and the human review time that agents still trigger are all inside the number. A cost-per-unit figure that excludes inference is a marketing number.

Units per week

116,848

Cost to serve / wk

$1.3M

was $3.9M

Run-rate saving

$133.8M

annualized at current volume

Inference spend / wk

$294K

22% of cost to serve

Capacity released

248.4 FTE

626 → 377 on the work

Cost to serve by sub-function

Before and after, per unit, with the inference component shown separately.

Sub-functionUnits / wkCost / unit beforeCost / unit nowOf which inferenceChangeWeekly delta
Portfolio & Intake4,462$42$15$1.97-64%$120K
Architecture & Design6,946$21$8.80$1.44-59%$87K
Build15,527$40$15$3.40-63%$390K
Code Review & Quality9,574$47$14$1.78-69%$312K
Test & Assurance16,000$13$4.20$0.78-68%$142K
Release & Deployment14,063$21$6.74$1.50-68%$199K
Site Reliability16,109$24$6.59$1.78-72%$273K
Technical Debt12,503$43$17$4.79-61%$323K
Developer Experience13,955$36$13$3.05-65%$321K
Security Engineering7,570$47$13$3.00-72%$258K
Engineering Analytics139$2K$441$125-71%$148K
Total116,848$3.9M$1.3M$294K-66%$2.6M

Inference is a variable cost that scales with volume, unlike the headcount it replaced. If volume doubles, the inference column doubles. That is the honest trade being made here.

Where the released capacity went

Capacity that is not redeployed is not a saving, it is a rounding error waiting to be re-hired.

Portfolio & Intake

34.222.6 FTE

11.6 FTE

Redeployed into control design and independent testing.

Architecture & Design

51.431.9 FTE

19.5 FTE

Redeployed into exception judgment and supplier relationship work.

Build

80.140.7 FTE

39.4 FTE

Redeployed into customer-facing escalation ownership.

Code Review & Quality

57.740.8 FTE

16.9 FTE

Redeployed into agent supervision, calibration and coaching.

Test & Assurance

84.554.9 FTE

29.6 FTE

Roles held flat; released hours absorbed volume growth rather than headcount reduction.

Release & Deployment

56.731.2 FTE

25.5 FTE

Redeployed into agent supervision, calibration and coaching.

Site Reliability

97.261.8 FTE

35.4 FTE

Redeployed into scenario work and strategy authoring review.

Technical Debt

51.033.6 FTE

17.4 FTE

Roles held flat; released hours absorbed volume growth rather than headcount reduction.

Developer Experience

62.930.0 FTE

32.9 FTE

Redeployed into cross-function process redesign.

Security Engineering

39.423.2 FTE

16.2 FTE

Redeployed into exception judgment and supplier relationship work.

Engineering Analytics

10.76.7 FTE

4.0 FTE

Redeployed into customer-facing escalation ownership.