LensReading which lens this session carries.
AI GBS & Global Business Services/Continuous Improvement

Sub-function

Continuous Improvement

Owns improvement work end to end inside AI GBS & Global Business Services. Runs 3 lines across 17 stations, drains the repeatable share into the AI-OPS tower, and holds 2 human gates.

orchestrator Continuous Improvement Orchestratorhuman owner Head of Continuous Improvement routes into ai-ops

Estate rating

BBB

73 / 100

autonomy 70%cap 78%

Detection, design and measurement are agentic. Running an experiment on a live process needs an owner who accepts the operational risk.

Agents

9

Lines

3

Units / wk

3,531

Timetable vs actual

19.6h → 18.8h

Cost / unit

$9.36

was $28

Run-rate saving

$3.3M

annualized at this volume

The 9 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.

Continuous Improvement Orchestrator

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

ObserveA350 / 24h96%16% ovr$7.23/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

ObserveWatches signals and watches the agents.

Waste Detection Agent

Watches continuous improvement continuously. Detects drift, quantifies it, and routes what matters without waiting for a reporting cycle.

ObserveA3122 / 24h93%17% ovr$5.80/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

TaskExecutes stations on a line.

Root Cause Agent

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

WorkflowA33,549 / 24h95%8% ovr$0.26/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

Experiment Design Agent

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

WorkflowA31,194 / 24h90%8% ovr$0.26/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

Idea Intake Agent

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

WorkflowA31,510 / 24h91%6% ovr$0.26/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

Change Adoption Agent

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

WorkflowA3495 / 24h91%8% ovr$0.09/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

AI GBSRuns the shared transaction spine.

Benefit Measurement Agent

Runs the repeatable portion of continuous improvement inside the shared transaction spine at spine cost and spine controls.

AI GBSA33,907 / 24h90%9% ovr$0.35/task

ceiling A3 · escalates to Escalates to Head of Continuous Improvement when the unit falls outside the decision envelope or confidence drops below the floor.

PolicyAuthors and versions the binding rules.

Improvement Policy Agent

Holds the rules that bind every other agent in continuous improvement. Versions them, tests them against live decisions, and blocks work that breaches them.

PolicyA2199 / 24h97%15% ovr$7.09/task

ceiling A3 · 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.

Continuous Improvement Challenger Agent

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

StrategyA2182 / 24h93%15% ovr$2.06/task

ceiling A3 · 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 Coverage76
Evidence Quality78
Override Discipline72
Data Integrity68
Recovery Readiness73
Cost Transparency74
Judgment Readiness69

People

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

headcount on this work27.716.9 FTE

Redeployed into data stewardship and evidence quality.

inference cost per unit$2.30

Lines running in this sub-function

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

10 clear4 evidenced1 held
GB10ASignal to ImprovementAnalytical · 7 stations

Detect waste

14 q

Root cause

1 live

Design experiment

22 q

Approve

8 q

Run

1 live

Measure

46 q

Standardize

30 q

GB10BIdea to BacklogTransactional · 5 stations

Intake

1 live

Score

1 live

Route

2 live

Track

3 live

Close loop

2 live

GB10CBenefit VerificationAnalytical · 5 stations

Pull baseline

13 q

Measure actual

16 q

Reconcile

1 live

Draft verdict

2 live

Ratify

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.

GB10-POS-100

Continuous Improvement Operating Position

Ratifiedv4

What operating shape carries twice the current improvement volume without adding headcount, and what breaks first if we are wrong?

Hold the ceiling at A3 for continuous improvement. Detection, design and measurement are agentic. Running an experiment on a live process needs an owner who accepts the operational risk. Raise it only after two consecutive quarters where the override rate stays under five percent and every override has a written cause.

confidence

80%

The challenge — Continuous Improvement Challenger Agent

Continuous Improvement Challenger Agent argues the recommendation leans on 3 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 6 percent for two consecutive months
  • ·Cost per improvement 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 $30k are sampled, not fully measured
Alternative consideredCostRiskVerdict
Hold the current position$243k run rateKnown and priced. Cedes ground if comparators move faster.not selected
Raise the ceiling one level now$1754k to build controlsOverride rate is 9 percent; raising the ceiling before that settles imports the error into production.rejected on evidence
Move the exception tail to the spine$994k transitionLoses local context. Rework risk on the cases that are hardest to recover.selected
authored by Continuous Improvement Orchestratorratifier Head of Continuous Improvement3 stated assumptions
outcome did not hold

Ratified 6 months ago. Measured since: cost per improvement rose 6 percent against a forecast that it would fall 21 percent.

Policy register

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

RefRuleScopeOwner agentApproved byVersionStatus
GB10-POL-200

Continuous Improvement Decision Envelope

Agents in this lane may act without a human when the improvement 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 5 agents · LATAM · US

envelopeImprovement Policy AgentHead of Continuous Improvementv4.8Active
GB10-POL-201

Continuous Improvement 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 5 agents · EU · UK · US

evidenceImprovement Policy AgentHead of Continuous Improvementv1.2Active
GB10-POL-202

Continuous Improvement 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. Head of Continuous Improvement owns the response clock.

binds 5 agents · Global

escalationImprovement Policy AgentHead of Continuous Improvementv4.2Active
GB10-POL-203

Continuous Improvement Autonomy Ceiling

The ceiling for this lane is 78 on the platform scale. Detection, design and measurement are agentic. Running an experiment on a live process needs an owner who accepts the operational risk. The ceiling is a governance decision, not a technical limit, and only Head of Continuous Improvement can propose moving it.

binds 5 agents · EU · UK

ceilingImprovement Policy AgentHead of Continuous Improvementv1.2Active