Why and within what context?
Goals · Constraints · Context · Success criteria · Human intent
The KMS ITC Architecture Method
Intent to evidence.
Architecture is the controlled transformation of human intent into evidence-backed outcomes.
We do not architect around technology. We architect the path through which technology is selected, authorised, operated, verified and improved.
The paradigm shift
From component topology to accountable action.
Traditional focus
Components
Interfaces
Data
Deployment
I2E focus
Intent
Authority
Action
Evidence
Outcome
The minimum unit of architecture is an authorised, observable and recoverable action.
The eight-layer chain
Every outcome must remain traceable to intent.
- 01
Intent
Define the human objective, constraints, prohibitions and success criteria.
- 02
Context
Understand the operating environment and decide what capability is actually needed.
- 03
Authority
Make explicit who may delegate which actions, to whom, for how long and within what limits.
- 04
Decision
Select the appropriate mechanism: human judgement, rules, software, optimisation, AI or agents.
- 05
Action
Change real state through bounded, observable, idempotent and recoverable operations.
- 06
Evidence
Preserve sources, decisions, permissions, tool calls, verification and state transitions.
- 07
Outcome
Measure the operational result—not merely whether the workflow produced an answer.
- 08
Learning
Use evidence to improve policy, context, tools, evaluation and future authority.
Three operating planes
One continuous accountability chain.
Authority crosses all three planes. Learning carries evidence back into the next expression of intent.
Who or what may act?
People · Workflow · Services · Agents · Tools · Infrastructure
What proves the outcome?
Policy · Verification · Audit · Provenance · Outcome · Learning
IntentEvidence → Learning
Technology selection ladder
Earn complexity one step at a time.
The goal is not maximum automation. It is the minimum sufficient mechanism for a reliable outcome.
- 01
Remove the step
If it creates no meaningful outcome, eliminate it.
- 02
Clarify the process
Fix ownership, policy and information before automating confusion.
- 03
Use deterministic software
Prefer rules, queries, forms and conventional code when behaviour can be specified.
- 04
Add an AI capability
Use models for interpretation, generation or classification inside a controlled workflow.
- 05
Delegate to one agent
Allow bounded planning and tool use where paths cannot be fully predetermined.
- 06
Orchestrate multiple agents
Distribute work only when specialisation or independent perspectives justify the coordination cost.
Five invariants
Rules that survive changes in platforms, models and frameworks.
- 01
No action without intent
Every consequential action must trace back to an explicit human or organisational objective.
- 02
No autonomy without authority
Capability never grants permission. Delegated authority must be bounded and revocable.
- 03
No decision without context
A confident answer without sufficient, trustworthy context is not a defensible decision.
- 04
No trust without evidence
Trust is earned through traceable sources, verification and observable outcomes.
- 05
No learning without accountability
Systems may adapt, but people and organisations retain responsibility for what changes.
Apply the method
Start with the consequential decision—not the technology shortlist.
Bring a workflow, architecture decision or agent initiative. We will trace the intent, authority, action and evidence needed to make it defensible.
Book a discovery session