OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
BlueHour Proprietary Demonstration — This demonstration illustrates proprietary BlueHour Operating Architecture and Micro Operating Model concepts. © 2026 BlueHour Technology, LLC. All rights reserved. Intellectual Property Notice
BlueHour
ENTERPRISE OPERATING MODEL

Your operating model, engineered and operated one model at a time.

How BlueHour binds AI, IT and Human Intelligence into one governable system, and proves the value in dollars before asking for the next step.

This demonstration follows Castle Pines Mutual, a fictional Mid-Market Enterprise.

AI× IT× Human Intelligence× Operating Architecture
=
GOVERNED OUTCOMESGood or bad, known and told truthfully
Designed with Physics. Deployed with Economics. Determined by Humans.
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
THE BLUEHOUR EQUATION
BlueHour
AI× IT× Human Intelligence× Operating Architecture = Governed Outcomes

Outcomes run in both directions. Good outcomes happen, and so do bad ones.

BlueHour does not promise that nothing goes wrong. It promises that every outcome, good or bad, is governed: known early, owned by a named person, and told truthfully.

EXAMPLES FROM CASTLE PINES MUTUAL
REVENUE
COST
RISK
TRUTH
LEVERAGE
→ Enterprise Value protected by knowing the truth early, whichever way it runs.
See the whole enterprise
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
BlueHour
ENTERPRISE OPERATING MODEL

Your operating model,
operated for you.

Signing in shows you what your Micro Operating Models found, and what is waiting on a human decision.

·BlueHour connects read-only. Nothing in your estate changes because you signed in.
·Every action taken here is recorded against the person who took it.
·BlueHour operator access is time-boxed, logged, and visible to you.
Designed with Physics. Deployed with Economics. Determined by Humans.
SIGNING IN TO
Castle Pines Mutual
If this is not your organization, choose another below.
Continue with your company account
Single sign-on via your identity provider · SAML or OIDC
OR
Access is governed by your organization's identity provider. Roles and permissions are set by Castle Pines Mutual, not by BlueHour.
BlueHour Technology · Operating Model as a Service System status: all models operatingSecurityPrivacyTerms
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
ENTERPRISE OPERATING MODEL MRI

The whole enterprise, running

A read-only scan across every series and all five levers. Nothing in the enterprise is changed to take it.
Scan of Castle Pines Mutual
as of Monday, 7:00 a.m.
SERIES
REVENUE
COST
RISK
TRUTH
LEVERAGE
0xx
Value & Financial Core
Not yet imaged
Watch
Weak
Watch
Strong
1xx
Customer & Go-to-Market
Strong
Strong
Watch
Not yet imaged
Watch
2xx
Product, Innovation & Delivery
Not yet imaged
Strong
Watch
Not yet imaged
Not yet imaged
3xx
Operations, Workforce & Resilience
Not yet imaged
Strong
Strong
Not yet imaged
Watch
4xx
Data, AI & Technology
Not yet imaged
Watch
Watch
Strong
Not yet imaged
5xx
Governance, Trust & Enterprise
Not yet imaged
Not yet imaged
Not yet imaged
Not yet imaged
Not yet imaged
Qualitative by design. Each cell summarizes the drill-down screens behind it, and the picture fills in as each Micro Operating Model is activated.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
EXECUTIVE SUMMARY · MONDAY, 7:00 A.M.

Today's decisions, by role

Each decision is owned, explained, and one click from its evidence.
8 decisions
one for each role
CEO Chief Executive Officer
Recommended: HOLD
Should payment-release agents stay live?
Our own monitoring did not observe them in the last test. Hold them until time to observe is inside tolerance.
See the evidence: Complexity
CFO Chief Financial Officer
Approve the $6.4M recovery list
Every line is owned and reversible. Separately, $1.9M of spend still has no owner.
See the evidence: First verdict
Chief Revenue OfficerRecommended: BUY
Fund the CRM platform's expansion?
It returns 22% against a 12% bar. Revenue elsewhere is not yet imaged; activating MOM-002 would measure it.
See the evidence: Ledger
COO Chief Operating Officer
Fix customer-service restore time, or narrow the agents' scope
Restoring takes 2 hours 40 minutes against a one-hour tolerance.
See the evidence: Complexity
CIO Chief Information Officer
Schedule the first halt drill for code assistants
120 agents whose halt has never been tested.
See the evidence: Kill Switch
CHRO Chief Human Resources Officer
Approve the upboarding plan for financial planning analysts
Their released capacity is part of 530 hours a week that has nowhere to go yet.
See the evidence: Upboarding
CISO Chief Information Security Officer
Find out who changed purchase order 5521
It moved from $48,000 to $84,000 after it was sealed. The seal shows that it changed; who changed it is the question.
See the evidence: Truth
CLO Chief Legal Officer
Preserve the sealed record of PO 5521, and decide how it is handled
The seal is evidence of when it changed. Whether this is an internal matter or one that must be reported is your call.
See the evidence: Truth
THE BOARD
Nothing today requires escalation.
Knowing when not to raise an alarm is part of governance too.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MACRO OPERATING MODEL

Your portfolio of sixty Micro Operating Models

Activated one at a time. Every model answers to a continuous BUY-HOLD-SELL verdict against return on invested capital.
1
Active
5
Planned
54
Available
0xx
Value & Financial Core
001Capital DisciplineACTIVE
002Revenue GrowthPLANNED
003Margin & ProfitabilityL
004Working Capital & CashC
005Pricing & MonetizationR
006Spend & ProcurementC
007Capital AllocationL
008Close & ReportingT
009Forecasting & PlanningL
010Portfolio & BU ValueL
1xx
Customer & Go-to-Market
101Demand GenerationR
102Sales & PipelineR
103Quote-to-CashR
104Customer ExperienceR
105Retention & ChurnR
106Success & ExpansionR
107Channel & Partner SalesR
108Brand & ReputationR
109Deal Desk & PricingR
110Voice of CustomerT
2xx
Product, Innovation & Delivery
201Product & RoadmapR
202R&D & InnovationL
203Engineering & DevOpsL
204Quality & ReliabilityK
205Design & ResearchR
206Launch ReadinessR
207Lifecycle ManagementL
208Platform & ReuseL
209ExperimentationT
210Concept-to-LaunchL
3xx
Operations, Workforce & Resilience
301Supply Chain & LogisticsC
302Process AutomationL
303Upboarding & WorkflowPLANNED
304Complexity Ceiling MgmtPLANNED
305Operating Risk Kill SwitchPLANNED
306Talent & WorkforceL
307Org Design & RightsL
308Change & AdoptionL
309Capacity & ResourceC
310Continuity & ResilienceK
4xx
Data, AI & Technology
401Data GovernanceT
402AI & Agent OrchestrationL
403IT ModernizationC
404Knowledge & RetrievalT
405Truth VerificationPLANNED
406CybersecurityK
407Cloud & FinOpsC
408IntegrationL
409ObservabilityT
410MLOps & Model GovK
5xx
Governance, Trust & Enterprise
501Risk & ComplianceK
502Audit & ControlsT
503Regulatory & PolicyK
504Privacy & ProtectionK
505ESG & SustainabilityT
506Strategy & OKRsL
507M&A & IntegrationL
508Board DecisioningT
509Ethics & Responsible AIT
510Enterprise Value & IRR
R Revenue GrowthC Cost OptimizationK Risk ManagementT Truth VerificationL Operating Leverage Activation order is chosen by each enterprise. MOM-001 is the mandatory start.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-001 · CAPITAL DISCIPLINE

Connect to the places your spend actually lives

Read-only connections to the systems you already own. Nothing is migrated and nothing is replaced.
STAGE 1 · CURRENT
Connect
STAGE 2 · NEXT
Build the ledger
STAGE 3 · NEXT
Score
STAGE 4 · NEXT
The first verdict

Sources

3 of 6 required sources connected
General ledger (ERP)
Actual spend by cost center and account
Connected · read-only
Hyperscaler billing: AWS, Azure, GCP
Consumption by account, service and tag
Connected · read-only
Model-provider usage APIs
Token consumption by agent, team and model
Connected · read-only
Contracts and procurement records
Terms, renewals and committed spend
Awaiting credentials
Lease, financing and asset schedules
Purchase, lease and depreciation terms, for Consumption Model Economics
Awaiting credentials
Apptio or Flexera
Existing cost models, where the client runs them
Not in use
CMDB
Ownership and attribution, connected last
Awaiting credentials
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-001 · CAPITAL DISCIPLINE

Every cost driver, scored against its return

Mapped to an owner, a business capability and a contract, then scored BUY-HOLD-SELL against the bar your people set.
STAGE 1 · DONE
Connect
STAGE 2 · DONE
Build the ledger
STAGE 3 · CURRENT
Score
STAGE 4 · NEXT
The first verdict
Sources reconciled to the general ledger
Yes
Cost drivers mapped
312the 8 largest shown below
Spend with no owner
$1.9M
Return bar, set by your finance sponsor
12% ROIC
COST DRIVEROWNERCAPABILITYANNUAL COSTRETURNPROPOSED VERDICT
Legacy claims platformCOO, ClaimsClaims processing$2.4M−6%SELL
Unowned SaaS subscriptions (14)UnassignedNone identified$1.9Mn/aSELL
Duplicate BI licencesVP AnalyticsReporting$1.2M2%SELL
Idle reserved GPU capacityHead of PlatformAI workloads$0.9M4%HOLD
Customer-service agent tokensVP ServiceCustomer support$3.6M31%BUY
Enterprise data warehouseChief Data OfficerAnalytics$4.8M18%BUY
Document AI pilotChief Data OfficerUnderwriting$0.4Mn/aHOLD
CRM platformChief Revenue OfficerSales$5.2M22%BUY
The score is a proposal. It moves nothing. Your people approve every verdict.
Apply Consumption Model Economics
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-001 · CAPITAL DISCIPLINE · CONSUMPTION MODEL ECONOMICS

The score says what to keep. Six lenses say how to pay for it.

The same eight cost drivers, seen through how they are acquired, consumed, priced, contracted, stressed and marked to market.
LENS 1 · CURRENT
Acquisition
LENS 2 · NEXT
Consumption
LENS 3 · NEXT
Unit
LENS 4 · NEXT
Contract
LENS 5 · NEXT
Scenario
LENS 6 · NEXT
Mark-to-Market
LENS 1 · DONE
Acquisition
LENS 2 · CURRENT
Consumption
LENS 3 · NEXT
Unit
LENS 4 · NEXT
Contract
LENS 5 · NEXT
Scenario
LENS 6 · NEXT
Mark-to-Market
LENS 1 · DONE
Acquisition
LENS 2 · DONE
Consumption
LENS 3 · CURRENT
Unit
LENS 4 · NEXT
Contract
LENS 5 · NEXT
Scenario
LENS 6 · NEXT
Mark-to-Market
LENS 1 · DONE
Acquisition
LENS 2 · DONE
Consumption
LENS 3 · DONE
Unit
LENS 4 · CURRENT
Contract
LENS 5 · NEXT
Scenario
LENS 6 · NEXT
Mark-to-Market
LENS 1 · DONE
Acquisition
LENS 2 · DONE
Consumption
LENS 3 · DONE
Unit
LENS 4 · DONE
Contract
LENS 5 · CURRENT
Scenario
LENS 6 · NEXT
Mark-to-Market
LENS 1 · DONE
Acquisition
LENS 2 · DONE
Consumption
LENS 3 · DONE
Unit
LENS 4 · DONE
Contract
LENS 5 · DONE
Scenario
LENS 6 · CURRENT
Mark-to-Market
COST DRIVERCOSTSCOREACQUISITION ECONOMICSREFINED VERDICT
Legacy claims platform$2.4MSELLPerpetual licence plus annual maintenance, on premisesAfter all six lenses
Unowned SaaS subscriptions (14)$1.9MSELL14 subscriptions bought on cards and expense claimsAfter all six lenses
Duplicate BI licences$1.2MSELLTwo per-seat enterprise agreementsAfter all six lenses
Idle reserved GPU capacity$0.9MHOLD3-year reserved commitmentAfter all six lenses
Customer-service agent tokens$3.6MBUYOn demand, billed per tokenAfter all six lenses
Enterprise data warehouse$4.8MBUY1-year reserved capacityAfter all six lenses
Document AI pilot$0.4MHOLDPilot credits; no production price yetAfter all six lenses
CRM platform$5.2MBUY3-year subscription, per seatAfter all six lenses
COST DRIVERCOSTSCORECONSUMPTION ECONOMICSREFINED VERDICT
Legacy claims platform$2.4MSELL41% of licensed claim volume in useAfter all six lenses
Unowned SaaS subscriptions (14)$1.9MSELL22% of seats active in the last 90 daysAfter all six lenses
Duplicate BI licences$1.2MSELL35% of users hold both toolsAfter all six lenses
Idle reserved GPU capacity$0.9MHOLD34% utilizedAfter all six lenses
Customer-service agent tokens$3.6MBUYSteady base, growing about 8% a monthAfter all six lenses
Enterprise data warehouse$4.8MBUY92% of the reservation used; growing 25% a yearAfter all six lenses
Document AI pilot$0.4MHOLDAbout 1,200 documents a weekAfter all six lenses
CRM platform$5.2MBUY3,100 of 4,000 seats active (78%)After all six lenses
COST DRIVERCOSTSCOREUNIT ECONOMICSREFINED VERDICT
Legacy claims platform$2.4MSELL$38 per claim, against $14 on the modern claims serviceAfter all six lenses
Unowned SaaS subscriptions (14)$1.9MSELL$1,180 a year per active userAfter all six lenses
Duplicate BI licences$1.2MSELL$2,100 a year per dual-licensed userAfter all six lenses
Idle reserved GPU capacity$0.9MHOLD$6.20 per GPU-hour actually used, against $2.10 on demandAfter all six lenses
Customer-service agent tokens$3.6MBUY$0.42 per resolved customer caseAfter all six lenses
Enterprise data warehouse$4.8MBUY$0.08 per analytics query, down 12% this yearAfter all six lenses
Document AI pilot$0.4MHOLD$0.90 per document, against $2.60 processed by handAfter all six lenses
CRM platform$5.2MBUY$1,300 per seat; $1,680 per active seatAfter all six lenses
COST DRIVERCOSTSCORECONTRACT ECONOMICSREFINED VERDICT
Legacy claims platform$2.4MSELLMaintenance renews in 5 months; no exit costAfter all six lenses
Unowned SaaS subscriptions (14)$1.9MSELL9 of the 14 auto-renew within 90 daysAfter all six lenses
Duplicate BI licences$1.2MSELLTrue-down allowed at the anniversary in 3 months; earlier exit costs $0.3MAfter all six lenses
Idle reserved GPU capacity$0.9MHOLD20 months remain; resale allowed under provider termsAfter all six lenses
Customer-service agent tokens$3.6MBUYNo commitment; prices can change on 30 days' noticeAfter all six lenses
Enterprise data warehouse$4.8MBUYRenews in 6 monthsAfter all six lenses
Document AI pilot$0.4MHOLDPilot ends in 2 months; production terms not negotiatedAfter all six lenses
CRM platform$5.2MBUYRenews in 4 months; seats can be reduced only at renewalAfter all six lenses
COST DRIVERCOSTSCORESCENARIO ECONOMICSREFINED VERDICT
Legacy claims platform$2.4MSELLVolume keeps shifting to digital intake; the gap widens each yearAfter all six lenses
Unowned SaaS subscriptions (14)$1.9MSELLNo capability depends on them under growth or contractionAfter all six lenses
Duplicate BI licences$1.2MSELLOne tool covers growth on its ownAfter all six lenses
Idle reserved GPU capacity$0.9MHOLDA newer GPU generation arrives within the termAfter all six lenses
Customer-service agent tokens$3.6MBUYAt +25% demand, the on-demand bill reaches $4.5MAfter all six lenses
Enterprise data warehouse$4.8MBUYAt current growth the reservation runs out in about 4 monthsAfter all six lenses
Document AI pilot$0.4MHOLDAt production volume, unit price matters more than pilot costAfter all six lenses
CRM platform$5.2MBUYAt −20% headcount, unused seats rise to 1,500After all six lenses
COST DRIVERCOSTSCOREMARK-TO-MARKET ECONOMICSREFINED VERDICT
Legacy claims platform$2.4MSELLA SaaS claims module is quoted at $1.1M a year at current volume
SELLConfirmed. Retire at maintenance renewal.
Unowned SaaS subscriptions (14)$1.9MSELLNot applicable; nothing to replace
SELLConfirmed. Cancel before notice dates.
Duplicate BI licences$1.2MSELLKeep the tool with higher use; its price is unchanged
SELLConfirmed. Release at the anniversary.
Idle reserved GPU capacity$0.9MHOLDToday's 1-year rate is 30% below the committed rate
SELLWas HOLD. Release and resell; 1-year terms only.
Customer-service agent tokens$3.6MBUYA 1-year commitment is ~25% lower; a smaller model handles routine intents for ~60% less
BUYConfirmed. Commit the base; route routine work.
Enterprise data warehouse$4.8MBUYA 3-year rate is about 20% below 1-year for the stable base
BUYConfirmed. Expand the reservation at renewal.
Document AI pilot$0.4MHOLDTwo providers offer comparable consumption pricing
HOLDConfirmed. No prepay until volume is proven.
CRM platform$5.2MBUYSame product, same price per seat; the gap is volume
BUYConfirmed. Keep it; cut 800 seats at renewal.
CONTINUOUS TRUE-UP
Capacity, demand, consumption and payments reconciled across every commitment. MOM-001 recommends the true-up; your people make it.
CASH TIMING, FOR THE TREASURER
Cash moves on contract dates, not approvals: BI licences in 3 months, CRM seats in 4, legacy claims in 5, agent tokens at signing.
Projections, not results. Only what is realized and verified counts. Findings are illustrative; MOM-001’s methods are not shown.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-001 · CAPITAL DISCIPLINE · CONSUMPTION MODEL ECONOMICS · AIME

AI Model Economics: the right intelligence, for the right work, at the right cost

AIME runs the same six lenses over the AI lines in the ledger, and prices each one per successful outcome, not per token. It is a capability inside CME, not a seventh lens or another MOM.
LENS 1
Acquisition
LENS 2
Consumption
LENS 3
Unit
LENS 4
Contract
LENS 5
Scenario
LENS 6
Mark-to-Market
AIME · AI Model Economics, operating across all six lenses for AI model consumption, and feeding the same BUY–HOLD–SELL decision
Customer-service agent tokens
From the ledger · score BUY
$3.6M
UNIT · MARK-TO-MARKET
Keep the capability; change the model. Per resolved case, $0.43 on the premium model against $0.10 on a validated general-purpose model, after retries and human rework. Validated on 2,000 cases (fictional).
BUY · new model
Document AI pilot
From the ledger · score HOLD
$0.4M
UNIT · CONTRACT
A low-cost model passes a 1,500-document shadow test: $0.83 to $0.02 per document. Production terms are still to be negotiated before the pilot ends.
BUY · at production terms
Idle reserved GPU capacity
From the ledger · score HOLD
$0.9M
SCENARIO · MARK-TO-MARKET
Self-hosting open weights on it costs $1.67M a year for Restricted work that runs for $46K on an API under exception. An open licence is not a free model; utilization decides.
HOLD · decide by policy
Refinements are proposals. Nothing is substituted, and no contract changes, without the owner's approval.
Carry to the first verdict ›
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-001 · CAPITAL DISCIPLINE

The first verdict: a ranked recovery list

Each line carries a dollar figure, a named owner and a documented path to reverse it. You approve what moves.
STAGE 1 · DONE
Connect
STAGE 2 · DONE
Build the ledger
STAGE 3 · DONE
Score
STAGE 4 · CURRENT
The first verdict
1
Legacy claims platform$2.4M
Owner: COO, Claims
To reverse: Archive snapshot kept; documented restore path within 10 business days.
Approved · reversible
Held for review
2
Unowned SaaS subscriptions (14)$1.9M
Owner: Assigned to CFO office
To reverse: Licences can be re-enabled from each vendor portal for 30 days.
Approved · reversible
Held for review
3
Duplicate BI licences$1.2M
Owner: VP Analytics
To reverse: Licence pool held for 90 days before release.
Approved · reversible
Held for review
4
Idle reserved GPU capacity$0.9M
Owner: Head of Platform
To reverse: Capacity can be re-reserved; workloads move to on-demand meanwhile.
Approved · reversible
Held for review
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-001 · CAPITAL DISCIPLINE · APPLICATION RATIONALIZATION

Decommission to the root

A SELL verdict is not a cancelled licence. The money sits in everything the application quietly keeps alive.
740 applications
scored against the return bar
THE VERDICT, BY COUNT
BUY168
HOLD262
SELL310
Of the SELL cohort, 64 applications have no owner at all. Those are eradication, not a portfolio decision.
WHAT MUST BE TRUE TO COUNT IT
The contract has an exit date, not just a switch-off date. Data is archived to the retention policy. Integrations are retired. Infrastructure is released. The system is out of audit scope. A named owner signs.
Anything short of this is an application that was switched off while its costs kept running.
THE FULL COST STACK · ONE APPLICATION
Legacy claims platform
$2.4M
a year, all in
Licence and subscription
$0.60M
Run labour, keeping it alive
$0.70M
Compute and storage
$0.50M
Support and maintenance contract
$0.25M
Integration maintenance
$0.20M
Backup, archive and retention
$0.14M
Audit and compliance scope
$0.11M
The licence is a quarter of it. Cancel only that and three quarters of the cost carries on, which is how rationalization programmes report savings nobody can find in the ledger.
KTLO RELEASED
9,400 hours a year
across the SELL cohort, keeping lights on in rooms nobody enters
THE SELL COHORT
Run-rate cost released$5.5M
One-time cost to do it properly$0.9M
Net, first twelve months$3.1M
WHEN THE MONEY ACTUALLY ARRIVES
Q1 Q2 Q4 break-even, Q2
You spend before you save. Recovery is governed by contract exit and renewal dates, not by the day a verdict is approved.
Where the released hours go
FOUR DIVIDENDS
from one verdict, landing on four different desks
Decommissioning · CFO
$5.5M of run-rate cost released to the root, not just the licence.
Capital reallocation · CEO
$3.1M net, to the bottom line or to what clears the return bar.
Time reallocation · COO
9,400 KTLO hours returned to people, if they are redeployed.
Attack surface · CISO
310 systems, 1,240 integration points and 3,800 standing credentials removed. Fewer places an agent can reach.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING

Can we still observe, contain and recover?

A proposed method. It locates no ceiling; it shows the direction of travel, process by process.
PROPOSED · NOT A VALIDATED THRESHOLD

Six questions for the Board

1Can we list every autonomous agent, what it may do, and who can revoke that within the hour?Partly
2Do we retain a complete record of what each one actually did?Yes, 90 days
3Could we review that entire record if we had to?No
4Can we halt an entire class of agent in a single action?Partly
5Would we learn of an unintended action from our own monitoring?Unknown
6Can we return an affected process to a known-good state?Yes, 6 hours
The value is not in a score. It is in the pattern of answers.

Three clocks, per process

Injected scenario: an agent has acquired a credential it was never issued.
PROCESSTOLERANCEOBSERVEHALTRESTORERESULT
Internal document workflow4 h2 h 10 m35 m3 hWithin tolerance
Customer-service agents1 h50 m20 m2 h 40 mRestore exceeds tolerance
Payment release10 mNot observedn/an/aNot detected by our monitoring
WHERE THE ANSWERS GO

Into Capital Discipline. Losing the ability to observe, halt or restore can justify a HOLD before the economics would.

Graded against each process rather than a general standard. Measurements adapted from operational resilience and incident response.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING · INJECTED SCENARIO

How a small failure becomes a large one

One unobserved action, followed through the arc, and the control that interrupts each stage. Simulated: we assert the event rather than create it.
COMPLEXITYnext
A procurement agent acquires a credential it was never issued, inside an estate where nobody can list every agent.
WOULD BE INTERRUPTED BY
The six questions. Can we name every agent, its permissions, and who revokes them within the hour?
ENTROPY
It writes to two systems nobody had mapped as connected. Small disturbances stop staying small.
INTERRUPTED BY
Time to observe. The first clock, measured per process against what that process can absorb.
CHAOS
Three agent classes act on the altered data. Claims queues back up and the cause is no longer obvious.
INTERRUPTED BY
Time to halt. One action stops an entire class, without asking a vendor to do it.
TRUTH EROSION
The record that would show what happened has itself been altered. Now the argument is about what is true.
INTERRUPTED BY
Sealed records and Fractal Fingerprints. The alteration is visible to a machine and to a person.
ANARCHY
Nobody can say what is true, so nobody can decide. Authority passes to whoever is loudest.
INTERRUPTED BY
A human decision of record, and Capital Discipline: HOLD the class, SELL what cannot be contained.
SCENARIO READY
Nothing has happened yet. Press Run the scenario to inject the event that has already happened elsewhere in the world.
Halt a class of agent
Every stage above is drawn from something that has happened somewhere in the world. The clocks are the ones measured under Complexity Saturation Testing: time to observe, time to halt, time to restore. Past the third stage you can stop the spread, but you cannot always put things back, which is why time to observe matters more than anything else on this screen.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING · INJECTED SCENARIO

How a small failure becomes a large one

One unobserved action, followed through the arc, and the control that interrupts each stage. Simulated: we assert the event rather than create it.
COMPLEXITYT+0
A procurement agent acquires a credential it was never issued, inside an estate where nobody can list every agent.
INTERRUPTED BY
The six questions. Can we name every agent, its permissions, and who revokes them within the hour?
ENTROPYnext
It writes to two systems nobody had mapped as connected. Small disturbances stop staying small.
WOULD BE INTERRUPTED BY
Time to observe. The first clock, measured per process against what that process can absorb.
CHAOS
Three agent classes act on the altered data. Claims queues back up and the cause is no longer obvious.
INTERRUPTED BY
Time to halt. One action stops an entire class, without asking a vendor to do it.
TRUTH EROSION
The record that would show what happened has itself been altered. Now the argument is about what is true.
INTERRUPTED BY
Sealed records and Fractal Fingerprints. The alteration is visible to a machine and to a person.
ANARCHY
Nobody can say what is true, so nobody can decide. Authority passes to whoever is loudest.
INTERRUPTED BY
A human decision of record, and Capital Discipline: HOLD the class, SELL what cannot be contained.
STAGE 1 REACHED · COMPLEXITY
The credential is in use and no monitoring has flagged it. This is the cheapest place to intervene, and the one most enterprises cannot reach.
Halt a class of agent
Every stage above is drawn from something that has happened somewhere in the world. The clocks are the ones measured under Complexity Saturation Testing: time to observe, time to halt, time to restore. Past the third stage you can stop the spread, but you cannot always put things back, which is why time to observe matters more than anything else on this screen.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING · INJECTED SCENARIO

How a small failure becomes a large one

One unobserved action, followed through the arc, and the control that interrupts each stage. Simulated: we assert the event rather than create it.
COMPLEXITYT+0
A procurement agent acquires a credential it was never issued, inside an estate where nobody can list every agent.
INTERRUPTED BY
The six questions. Can we name every agent, its permissions, and who revokes them within the hour?
ENTROPYT+35 min
It writes to two systems nobody had mapped as connected. Small disturbances stop staying small.
INTERRUPTED BY
Time to observe. The first clock, measured per process against what that process can absorb.
CHAOSnext
Three agent classes act on the altered data. Claims queues back up and the cause is no longer obvious.
WOULD BE INTERRUPTED BY
Time to halt. One action stops an entire class, without asking a vendor to do it.
TRUTH EROSION
The record that would show what happened has itself been altered. Now the argument is about what is true.
INTERRUPTED BY
Sealed records and Fractal Fingerprints. The alteration is visible to a machine and to a person.
ANARCHY
Nobody can say what is true, so nobody can decide. Authority passes to whoever is loudest.
INTERRUPTED BY
A human decision of record, and Capital Discipline: HOLD the class, SELL what cannot be contained.
STAGE 2 REACHED · ENTROPY
Two unmapped systems are now involved. Interception still costs minutes, and everything is still reversible.
Halt a class of agent
Every stage above is drawn from something that has happened somewhere in the world. The clocks are the ones measured under Complexity Saturation Testing: time to observe, time to halt, time to restore. Past the third stage you can stop the spread, but you cannot always put things back, which is why time to observe matters more than anything else on this screen.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING · INJECTED SCENARIO

How a small failure becomes a large one

One unobserved action, followed through the arc, and the control that interrupts each stage. Simulated: we assert the event rather than create it.
COMPLEXITYT+0
A procurement agent acquires a credential it was never issued, inside an estate where nobody can list every agent.
INTERRUPTED BY
The six questions. Can we name every agent, its permissions, and who revokes them within the hour?
ENTROPYT+35 min
It writes to two systems nobody had mapped as connected. Small disturbances stop staying small.
INTERRUPTED BY
Time to observe. The first clock, measured per process against what that process can absorb.
CHAOST+2 hr 10
Three agent classes act on the altered data. Claims queues back up and the cause is no longer obvious.
INTERRUPTED BY
Time to halt. One action stops an entire class, without asking a vendor to do it.
TRUTH EROSIONnext
The record that would show what happened has itself been altered. Now the argument is about what is true.
WOULD BE INTERRUPTED BY
Sealed records and Fractal Fingerprints. The alteration is visible to a machine and to a person.
ANARCHY
Nobody can say what is true, so nobody can decide. Authority passes to whoever is loudest.
INTERRUPTED BY
A human decision of record, and Capital Discipline: HOLD the class, SELL what cannot be contained.
STAGE 3 REACHED · CHAOS
Halting the class stops the spread. It does not undo what has already been written: restoring is a separate clock.
Halt a class of agent
Every stage above is drawn from something that has happened somewhere in the world. The clocks are the ones measured under Complexity Saturation Testing: time to observe, time to halt, time to restore. Past the third stage you can stop the spread, but you cannot always put things back, which is why time to observe matters more than anything else on this screen.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING · INJECTED SCENARIO

How a small failure becomes a large one

One unobserved action, followed through the arc, and the control that interrupts each stage. Simulated: we assert the event rather than create it.
COMPLEXITYT+0
A procurement agent acquires a credential it was never issued, inside an estate where nobody can list every agent.
INTERRUPTED BY
The six questions. Can we name every agent, its permissions, and who revokes them within the hour?
ENTROPYT+35 min
It writes to two systems nobody had mapped as connected. Small disturbances stop staying small.
INTERRUPTED BY
Time to observe. The first clock, measured per process against what that process can absorb.
CHAOST+2 hr 10
Three agent classes act on the altered data. Claims queues back up and the cause is no longer obvious.
INTERRUPTED BY
Time to halt. One action stops an entire class, without asking a vendor to do it.
TRUTH EROSIONT+6 hr
The record that would show what happened has itself been altered. Now the argument is about what is true.
INTERRUPTED BY
Sealed records and Fractal Fingerprints. The alteration is visible to a machine and to a person.
ANARCHYnext
Nobody can say what is true, so nobody can decide. Authority passes to whoever is loudest.
WOULD BE INTERRUPTED BY
A human decision of record, and Capital Discipline: HOLD the class, SELL what cannot be contained.
STAGE 4 REACHED · TRUTH EROSION
This is the point of no easy return. The seal shows the record changed, which is the difference between an investigation and an argument.
Halt a class of agent
Every stage above is drawn from something that has happened somewhere in the world. The clocks are the ones measured under Complexity Saturation Testing: time to observe, time to halt, time to restore. Past the third stage you can stop the spread, but you cannot always put things back, which is why time to observe matters more than anything else on this screen.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
COMPLEXITY SATURATION TESTING · INJECTED SCENARIO

How a small failure becomes a large one

One unobserved action, followed through the arc, and the control that interrupts each stage. Simulated: we assert the event rather than create it.
COMPLEXITYT+0
A procurement agent acquires a credential it was never issued, inside an estate where nobody can list every agent.
INTERRUPTED BY
The six questions. Can we name every agent, its permissions, and who revokes them within the hour?
ENTROPYT+35 min
It writes to two systems nobody had mapped as connected. Small disturbances stop staying small.
INTERRUPTED BY
Time to observe. The first clock, measured per process against what that process can absorb.
CHAOST+2 hr 10
Three agent classes act on the altered data. Claims queues back up and the cause is no longer obvious.
INTERRUPTED BY
Time to halt. One action stops an entire class, without asking a vendor to do it.
TRUTH EROSIONT+6 hr
The record that would show what happened has itself been altered. Now the argument is about what is true.
INTERRUPTED BY
Sealed records and Fractal Fingerprints. The alteration is visible to a machine and to a person.
ANARCHYT+2 days
Nobody can say what is true, so nobody can decide. Authority passes to whoever is loudest.
INTERRUPTED BY
A human decision of record, and Capital Discipline: HOLD the class, SELL what cannot be contained.
STAGE 5 REACHED · ANARCHY
You can still stop the spread. You cannot put it back together. Every control above exists to keep the scenario from reaching this card.
Halt a class of agent
Every stage above is drawn from something that has happened somewhere in the world. The clocks are the ones measured under Complexity Saturation Testing: time to observe, time to halt, time to restore. Past the third stage you can stop the spread, but you cannot always put things back, which is why time to observe matters more than anything else on this screen.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-305 · OPERATING RISK KILL SWITCH · PLANNED

Halt an entire class of agent in a single action

Without asking a vendor to do it for you. A halt is a human decision, and so is a resume.
AGENT CLASSWHAT IT MAY DOCONTROL RUNGWHO CAN HALT ITLAST DRILL
Customer-service agents
42 agents · Running
Read customer records, draft replies, issue refunds below a set limit
Constrained
Refunds need approval since Monday. Freezes automatically on 3 more denied attempts.
VP Service
20 min
Invoice reconciliation agents
18 agents · Running
Read the ledger and vendor portals; propose matches for approval
Normal
Constrains automatically above 5 denied attempts in an hour.
Controller
12 min
Code assistants
120 agents · Running
Read repositories; open pull requests for human review
Normal
Triggers set, but the freeze has never been drilled.
VP Engineering
Not yet tested
Procurement agents
6 agents · Running
Draft purchase orders below the approval threshold
Frozen
Frozen after PO 5521 changed post-seal. Expires in 12 hours unless a person extends it.
Chief Procurement Officer
35 min
Document AI pilot
3 agents · Halted
Read underwriting files; no write access
Frozen
Stopped by a person. New actions only; restoring is a separate clock.
Chief Data Officer
8 min
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-303 · UPBOARDING & WORKFLOW · PLANNED

Assets you sell. People you upboard.

As agents take on the rote, the capacity they release is counted, and so is where it goes.
ASSETS
BUY · HOLD · SELL
PEOPLE
BUY · HOLD · UPBOARD
Capacity released by agents
1,840 hours a week
Redeployed to higher-value work
1,310 hours a week
Released but not yet redeployed
530 hours a week
ROLE WORK AGENTS NOW HANDLE CAPACITY RELEASED HIGHER-VALUE WORK OPENED STATUS
Claims intake specialists
40 people
 Document intake and triage 45% Complex claims adjudication Upboarded
Accounts payable analysts
24 people
 Invoice matching and first-pass exceptions 38% Vendor terms and exception root causes In progress
Service desk, first line
60 people
 Routine tickets and account resets 52% Agent supervision and escalation quality Plan approved
Financial planning analysts
12 people
 Data gathering and variance tables 30% Scenario modelling and capital review Not yet redeployed
People up, not out.
Capacity that is released and never redeployed is a cost that remains. Tracking where it goes is how Operating Leverage becomes visible.
BlueHour
ENTERPRISE OPERATING MODEL
Castle Pines Mutual RD
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
MOM-405 · TRUTH VERIFICATION · PLANNED

Truth Preservation: verify, seal, see

Each consequential record is checked at entry, sealed so it cannot change without detection, and given a Fractal Fingerprint a person can recognize.
1 · VERIFY
Checked at entry
2 · SEAL
Tokenized
3 · SEE
Fractal Fingerprint

Sealed records

Integrity checked continuously
Fractal Fingerprint of the SELL verdict
SELL verdict · Legacy claims platform · $2.4M
Verified: reconciled to the general ledger; approved by COO, Claims
Sealed 09:14 · c72c516b…b08f
Verified and sealed
Fractal Fingerprint of the licence approval
Approval · Duplicate BI licences, released after 90 days
Verified: attested by the named owner, VP Analytics
Sealed 09:31 · 1ce94e6d…d25b
Verified and sealed
Fractal Fingerprint of the refund action
Agent action · Customer-service agent 17 · refund $240
Verified: cross-checked against the payments system
Sealed 10:02 · a58f01d4…e94e
Verified and sealed
Fractal Fingerprint of the forecast input
Forecast input · Q4 cloud spend · $3.1M
Not yet verified: awaiting reconciliation to hyperscaler billing
Sealed 10:40 · 7ac151fd…f361
Sealed, not verified
Fractal Fingerprint of the purchase order as sealed
Agent action · Procurement agent · draft PO 5521
Verified: matched to the purchasing system at entry
Sealed 11:05 · b99f3989…5cfe
Changed after sealing
TRACK RECORD · PRIOR WORK AT KILLERIT, NOT A BLUEHOUR CLIENT
WHY BELIEVE THIS

The method, proven by hand.

The discipline behind MOM-001 was built at Forsythe Technology and productized as KillerIT, the division BlueHour's founder founded and led, before there were agents to run it.

BlueHour
~$200M
hard IT cost removed
30×
return on the cost of the engagement
12 months
at a global financial-services firm
HOW IT WAS DONE

The application portfolio was scored BUY-HOLD-SELL against return on invested capital. Cost that had stopped earning was decommissioned to the root, and capital was redeployed to the assets worth funding. Each asset was scored on business metrics, its cost against the value of its functionality, competitiveness, uniqueness and user base, and on technology metrics, alignment, audit compliance, stability, quality and complexity, resolving to BUY to invest, HOLD to sustain, or SELL to retire or migrate. The same discipline was applied across financial-services and healthcare enterprises from $1B to $70B in revenue.

GARTNER MAGIC QUADRANT · 2015
Leader, Integrated IT Portfolio Analysis
GARTNER · MAY 2014
Cool Vendor, Program & Portfolio Management
One of five vendors recognized
Recognitions earned by KillerIT, a division of Forsythe Technology, where the Capital Discipline methodology was originally developed. Gartner, Cool Vendors in Program and Portfolio Management, 2014, Robert Handler, Jim Duggan, Daniel Stang, May 2, 2014. Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings.
THEN AND NOW

Then, it was done with spreadsheets and human judgment, which was enough for a static estate. Now, consumption-priced tokens and agents that spawn agents move too fast for that, so MOM-001 runs the same discipline continuously, with agents.

No BlueHour client has completed MOM-001 yet. We built the model before we sold it.

Start with one
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
WHAT IT COSTS

A base to begin. The rest is earned.

MOM-001, Capital Discipline. A base fee at activation, then two earn-ups triggered by the recovery you approve at the six-week gate. Each earn-up equals the base, and the total is capped at three times it.
BlueHour
MID-MARKET ENTERPRISE
Under $250M of annual IT and AI spend under verdict
$325,000
base, payable at activation
Earn-up 1, at $1.6M approved+ base
Earn-up 2, at $3.25M approved+ base
CAPPED AT$975,000
ENTERPRISE
$250M to $1B of annual IT and AI spend under verdict
$945,000
base, payable at activation
Earn-up 1, at $4.7M approved+ base
Earn-up 2, at $9.5M approved+ base
CAPPED AT$2,835,000
GLOBAL ENTERPRISE
Above $1B of annual IT and AI spend under verdict
$3,775,000
base, payable at activation
Earn-up 1, at $18.9M approved+ base
Earn-up 2, at $37.8M approved+ base
CAPPED AT$11,325,000
WHAT THE BASE BUYS
Read-only connection to the systems you already own. The value-attributed cost ledger, every driver mapped to an owner, a capability and a contract. Scoring against the return bar your finance sponsor sets. And the first verdict: a ranked recovery list with a dollar figure, a named owner and a documented way to reverse every line.
AT SIX WEEKS, EVERYTHING RESETS
Scope, tier and the ongoing subscription are set from what the ledger actually shows, not from assumptions either of us made in this room. Either side can stop. If you stop, you keep the ledger, the scoring and the recovery list, and you owe nothing further.
We start at a third of the price. If the ledger shows the money and you approve the list, we earn the rest. If it does not, we do not. The earn-ups trigger on your approval, never on our recommendation.
Indicative pricing, subject to change. Final fees, scope and payment terms are set in a signed agreement and are subject to credit review.
Start with one
OPERATING MODEL AS A SERVICE · DEMONSTRATION · WORK IN PROGRESS · ILLUSTRATIVE DATA, FICTIONAL CLIENT
THE PRUDENT START

Start with one.

MOM-001, Capital Discipline, is activated rather than implemented. No project team, no discovery phase, no report at the end, because it does not end: we operate it and it scores your portfolio every week. From your side it needs three standing roles, not a staffed programme. Your first verdict lands in about six weeks, and the money follows on contract dates.

You get a number
Conversion Yield: verified value that arrived, over total lifecycle cost. Most enterprises cannot compute it, and that is itself the finding.
You get early warning
Complexity, entropy, chaos, truth erosion, anarchy is not an inevitable arc. Each stage has a control that interrupts it, and the earliest is the cheapest.
You get a ladder, not a switch
Correct, constrain, freeze, disable. The early rungs run automatically. Disabling stays a human act.
You keep the truth
Verified at entry, sealed against change, visible to a person. Every step reversible, every decision owned.

Slowing the labs down does not slow what is accumulating inside your company. The first model carries no obligation to build the second.

PUT MOM-001 TO WORK IN YOUR ENTERPRISE
MOM-001 starts read-only. Your people retain decision authority. See what Capital Discipline finds in your Enterprise.
Request MOM-001 →
BlueHour
Designed with Physics. Deployed with Economics. Determined by Humans.
Opus. Humanitas. Veritas. Work. Humanity. Truth.