Start a project →
Start a project →

03 / Runtime03 / Runtime

Keep the system working after launch.Keep the system working after launch.

Runtime is Xendai's ongoing operating layer: we monitor, evaluate, maintain and improve deployed AI systems once they are in production.Runtime is Xendai's ongoing operating layer: we monitor, evaluate, maintain and improve deployed AI systems once they are in production.

A deployed system is still a changing system.A deployed system is still a changing system.

Start a project →

02Production change02 / Production change

Putting a system into production does not mean it stops changing.Putting a system into production does not mean it stops changing.

Models, APIs and applications change. New cases, exceptions and business rules appear. Runtime defines who observes those changes, how they are detected, what is evaluated and what responsibility Xendai owns.Models, APIs and applications change. New cases, exceptions and business rules appear. Runtime defines who observes those changes, how they are detected, what is evaluated and what responsibility Xendai owns.

  1. Change register — classes of change, not incidentsSix types
  2. 01Model/

    changed

  3. 02API/

    updated

  4. 03Data/

    drifted

  5. 04Rule/

    modified

  6. 05Volume/

    increased

  7. 06Exception/

    new

  8. Coral marks the class that may need attention — no exception is asserted.
  1. 01Model

    changed

  2. 02API

    updated

  3. 03Data

    drifted

  4. 04Rule

    modified

  5. 05Volume

    increased

  6. 06Exception

    new

  7. Coral marks the class that may need attention — no exception is asserted.

03What Runtime covers03 / What Runtime covers

We do not only maintain infrastructure. We maintain system behaviour.We do not only maintain infrastructure. We maintain system behaviour.

  1. Operating coverage — six domainsInfrastructure health ≠ correct business behaviour
  2. 01HealthInfrastructure

    Availability, executions, errors, logs and dependencies.

  3. 02QualityBehaviour

    Periodic evaluation of results.

  4. 03IntegrationsInfrastructure

    Compatibility with APIs, models and tools.

  5. 04ExceptionsBehaviour

    Cases that fail to complete correctly.

  6. 05CostInfrastructure

    Model, infrastructure and operating consumption.

  7. 06ImprovementBehaviour

    Adjustments inside the agreed operating perimeter.

  1. 01HealthInfra

    Availability, executions, errors, logs and dependencies.

  2. 02QualityBehaviour

    Periodic evaluation of results.

  3. 03IntegrationsInfra

    Compatibility with APIs, models and tools.

  4. 04ExceptionsBehaviour

    Cases that fail to complete correctly.

  5. 05CostInfra

    Model, infrastructure and operating consumption.

  6. 06ImprovementBehaviour

    Adjustments inside the agreed operating perimeter.

04Levels04 / Levels

Three levels based on operating responsibility.Three levels based on operating responsibility.

  1. Operating responsibility — scoped per system and proposalNo level implies 24/7 by itself
  2. BaseKeep the system healthy.
  3. ManagedOperate and actively improve it.
  4. CriticalProtect a business-critical function.
  5. Health · Operate and improve · ProtectResponsibility grows. Service level is defined in each proposal.
  1. BaseKeep it healthy.
  2. ManagedOperate and actively improve it.
  3. CriticalProtect a business-critical function.
  4. Example: exception → human intervention → verifiedNo level implies 24/7 by itself. Service level is defined in each proposal.

05Two dimensions05 / Two dimensions

Hard to build does not mean critical to operate.Hard to build does not mean critical to operate.

Engineering measures build complexity. Runtime measures the responsibility required once the system is live. They are independent decisions.Engineering measures build complexity. Runtime measures the responsibility required once the system is live. They are independent decisions.

Axis A — EngineeringWhat must be built?Build complexity
Axis B — RuntimeHow critical is it to keep running?Operating criticality

Xendai Managed = a deployment environment. Runtime Managed = an operating responsibility. Different dimensions. A simple system on a main line can be critical. An elaborate internal system may tolerate downtime.

Axis A — EngineeringWhat must be built?Build complexity
Axis B — RuntimeHow critical is it to keep running?Operating criticality

A simple system on a main line can be critical. An elaborate internal system may tolerate downtime.

06The loop06 / The loop

Runtime is a loop, not a passive maintenance contract.Runtime is a loop, not a passive maintenance contract.

Observe → Evaluate → Identify → Correct → Verify → Improve ↺Continued observation is a legitimate outcome
  1. 01 Observe
  2. 02 Evaluate
  3. 03 Identify
  4. 04 Correct
  5. 05 Verify
Nothing actionable → keep observingConditional, controlled response
06 Improve ↺

Learning returns to observation. No auto-healing, no constant staffing implied.

Lime = a correction verified.

01 Observe → 02 Evaluate → 03 Identify → 04 Correct → 05 Verify

06 Improve ↺ — feeds back to 01 Observe

Nothing actionable → keep observing. Conditional, controlled response between Correct and Verify. Learning returns to observation. No auto-healing, no constant staffing implied.

07Boundary07 / Boundary

Responsibility must be as clearly defined as the system.Responsibility must be as clearly defined as the system.

Service schedule — terms to define in each proposal · Nine items · no values prefilled
01Operating hours.to be defined
02Incident severities.to be defined
03Response targets.to be defined
04Restoration targets when applicable.to be defined
05Monitoring perimeter.to be defined
06Evaluation cadence.to be defined
07Change allowance.to be defined
08Dependencies.to be defined
09Escalation.to be defined

Critical does not automatically mean 24/7. Each proposal defines the actual service level.

Service schedule — nine items, no values prefilled
01Operating hoursto be defined
02Incident severitiesto be defined
03Response targetsto be defined
04Restoration targetsto be defined
05Monitoring perimeterto be defined
06Evaluation cadenceto be defined
07Change allowanceto be defined
08Dependenciesto be defined
09Escalationto be defined

Critical does not automatically mean 24/7. Each proposal defines the actual service level.

08Runtime + Usage08 / Runtime + Usage

Operating the system and using it are different costs.Operating the system and using it are different costs.

Category ARuntime

Xendai's agreed operating responsibility.

Fixed scope · agreed perimeter
Category BUsage

Models · workflows · telephony · infrastructure · selective outcomes (only when agreed).

Variable · external · not unlimited by default

Two cost categories. Inclusion and exclusion belong to the agreed proposal.

Category ARuntime

Xendai's agreed operating responsibility.

Fixed scope · agreed perimeter
Category BUsage

Models · workflows · telephony · infrastructure · selective outcomes (only when agreed).

Variable · external · not unlimited by default

Two cost categories. Inclusion and exclusion belong to the agreed proposal.

09 / Start a project09 / Start a project

A deployed system is still a changing system.A deployed system is still a changing system.

Start a project →