02 / Engineering02 / Engineering

We build AI systems that perform real work inside your business.We build AI systems that perform real work inside your business.

We design, integrate, evaluate and deploy AI systems with a defined responsibility, clear boundaries and measurable operating criteria.We design, integrate, evaluate and deploy AI systems with a defined responsibility, clear boundaries and measurable operating criteria.

Models, data, tools, people and controls working as a system.Models, data, tools, people and controls working as a system.

Start a project →

02Capability02 / Capability

We do not build “AI”. We build operating capabilities.We do not build "AI". We build operating capabilities.

A model alone does not solve a process. Useful systems require context, tools, permissions, exception handling and human intervention. Engineering connects those pieces into a deployable and testable capability with defined ownership.A model alone does not solve a process. Useful systems require context, tools, permissions, exception handling and human intervention. Engineering connects those pieces into a deployable and testable capability with defined ownership.

CompositionThe model is one element

ModelContextToolsDataRulesHuman controlAI system

Composition — the model is one element

ModelContextToolsDataRulesHuman controlAI system

03Engineering baseline03 / Engineering baseline

Before choosing technology, we define what the system is responsible for.Before choosing technology, we define what the system is responsible for.

  1. Blueprint — six dimensionsBoundary → Behaviour → Control → Quality → Deployment → Runtime ready
  2. 01Business boundary

    Trigger, input, output, process owner and scope.

  3. 02System behaviour

    Decisions, actions, context and exceptions.

  4. 03Control

    Permissions, approvals, boundaries, traceability and recovery.

  5. 04Quality

    Representative tests, critical cases and acceptance.

  6. 05Deployment

    Environment, accounts, access, dependencies and handover.

  7. 06Runtime readiness

    Monitoring, escalation, runbook and post-launch responsibilities.

  1. 01Business boundary

    Trigger, input, output, process owner and scope.

  2. 02System behaviour

    Decisions, actions, context and exceptions.

  3. 03Control

    Permissions, approvals, boundaries, traceability and recovery.

  4. 04Quality

    Representative tests, critical cases and acceptance.

  5. 05Deployment

    Environment, accounts, access, dependencies and handover.

  6. 06Runtime readiness

    Monitoring, escalation, runbook and post-launch responsibilities.

04AI Systems04 / AI Systems

Four families for four types of operating capability.Four families for four types of operating capability.

01

Xflow

Connect processes.

End-to-end processes across systems, decisions, actions and people.

Connect / Orchestrate / OperateXflow →
02

Xagents

Delegate work with responsibility.

Agents with responsibilities, tools, boundaries and completion criteria.

Task / Operator / CrewXagents →
03

Xmind

Turn knowledge into operational context.

Governed information and data for people and agents.

Search / Context / IntelligenceXmind →
04

Xvoice

Turn voice into an operational interface.

Conversations that query systems, take action and escalate with context.

Answer / Act / OperateXvoice →
01

Xflow

Connect processes.

End-to-end processes across systems, decisions, actions and people.

Xflow →
02

Xagents

Delegate work with responsibility.

Agents with responsibilities, tools, boundaries and completion criteria.

Xagents →
03

Xmind

Turn knowledge into operational context.

Governed information and data for people and agents.

Xmind →
04

Xvoice

Turn voice into an operational interface.

Conversations that query systems, take action and escalate with context.

Xvoice →

05Composition05 / Composition

A real problem rarely belongs to one category.A real problem rarely belongs to one category.

Families can combine. An agent can consume Xmind context, trigger Xflow and receive a request through Xvoice while each capability keeps its own responsibility and criteria.Families can combine. An agent can consume Xmind context, trigger Xflow and receive a request through Xvoice while each capability keeps its own responsibility and criteria.

One possible configuration — illustrativeEach family keeps its own responsibility and criteria

Xvoice

Request enters

Bounded work
Permissions · Completion criteria

Xagents

Acts within its responsibilityHuman check on exception

Xflow

Triggered · updates business systems

Xmind

Context consumed by the agent
ERP / CRM / Email
Database / API
Business systems

One possible configuration — illustrative

Each family keeps its own responsibility and criteriaXvoice
Request entersBounded workPermissions · Completion criteria
XagentsActs within its responsibility
Human check on exceptionXmind
One possible configuration — illustrativeXvoice

Request enters

Xagents

Acts within its responsibility — bounded work, permissions, completion criteria
Human check on exception

Xmind

Context consumed by the agentXflow

Triggered · updates business systems · ERP / CRM / Email / Database / API


One possible configuration — illustrative

XvoiceRequest enters
XagentsActs within its responsibility — bounded work, permissions, completion criteriaHuman check on exception
XmindContext consumed by the agent
XflowTriggered · updates business systems · ERP / CRM / Email / Database / API

06Scope06 / Scope

Complexity depends on the work the system must own.Complexity depends on the work the system must own.

Each system family follows its own progression, based on autonomy, integrations, context, decisions, exceptions and operational responsibility.Each system family follows its own progression, based on autonomy, integrations, context, decisions, exceptions and operational responsibility.

FamilyLevel 1Level 2Level 3
XflowConnectOrchestrateOperate
XagentsTaskOperatorCrew
XmindSearchContextIntelligence
XvoiceAnswerActOperate
XflowConnect→ Orchestrate→ Operate
XagentsTask→ Operator→ Crew
XmindSearch→ Context→ Intelligence
XvoiceAnswer→ Act→ Operate

07Deployment07 / Deployment

Architecture adapts to the client environment.Architecture adapts to the client environment.

Four configurations of ownership and executionAn architectural option — not an additional service
01

Customer environment

Inside client accounts and systems.

02

Xendai managed

Xendai manages the required execution or orchestration layer.

03

Hybrid

Responsibility split between client and Xendai.

04

Private

Dedicated or specifically controlled infrastructure when justified.

Client-ownedXendai-operated layer

Deployment ≠ Roadmap / Engineering / Runtime

01

Customer environment

Inside client accounts and systems.

02

Xendai managed

Xendai manages the required execution or orchestration layer.

03

Hybrid

Responsibility split between client and Xendai.

04

Private

Dedicated or specifically controlled infrastructure when justified.

Deployment ≠ Roadmap / Engineering / Runtime

08Handover08 / Handover

Engineering ends when the system can operate against clear criteria.Engineering ends when the system can operate against clear criteria.

Before production we define acceptance, critical exceptions, recovery, responsibilities and operating conditions. Engineering complexity and Runtime criticality are different dimensions.Before production we define acceptance, critical exceptions, recovery, responsibilities and operating conditions. Engineering complexity and Runtime criticality are different dimensions.

Engineering → Acceptance → Runtime as potential operating responsibility
Engineering

Complexity — the work the system owns

Acceptance

Acceptance · Critical exceptions · Recovery · Responsibilities · Operating conditions

Runtime

Criticality — a different dimension

Engineering — complexity
Acceptance
Runtime — criticality
Explore Runtime →

09 / Start a project09 / Start a project

If the work can be defined, it can be engineered.If the work can be defined, it can be engineered.

Start a project →