02 - The Framework
From intention to evidence
A looped method to decide whether Augmentership is justified, then design, deploy, evaluate and adapt it according to the evidence.
Integrity Loop
Five steps
- 01
NEED
Need
Determine whether the additional effort of Augmentership is justified.
- 02
DESIGN
Design
Define the target capability, holder, beneficiaries, people affected, work configuration, expected mechanisms, critical skills, agency, comparator, measures, costs, horizon and adaptation rules.
- 03
DEPLOY
Deploy
Implement the defined configuration while documenting relevant changes.
- 04
EVALUATE
Evaluate
Assess performance, capability, agency, value, costs, risks and distribution.
- 05
ADAPT
Adapt
Continue, extend, modify, reduce, stop or replace according to the evidence.
Need Assessment
Is Augmentership really needed here?
Four possible decisions
No intervention
Current performance and capability are sufficient.
Non-AI intervention
Training, simplification, documentation, organisational change or another non-AI intervention is more appropriate.
AI assistance / automation is enough
The main goal is efficiency, throughput or routine execution; additional design is not justified.
Augmentership justified
There is an explicit need to develop, preserve, redistribute or extend a capability, and existing approaches are not sufficient.
Augmentation-by-Design
What must be defined
Target Capability
The capability to augment, preserve, redistribute or extend.
Capability Holder
The person, configuration, team or organisation that should hold the capability.
Work Configuration
How work is distributed between human, AI, tools, processes and roles.
Expected Mechanism
Expected practice, feedback, adaptation, transfer, accessibility or coordination.
Critical Skills
Capabilities that must not regress beyond defined thresholds.
Agency Conditions
Who can direct, intervene, suspend, correct, challenge or refuse.
People Affected
Who participates, benefits, bears the costs or can contest.
Evaluation Plan
How performance, capability, agency, value and costs will be measured.
Adaptation Rule
What to do if the evidence contradicts the promise.
Measurement and value
The active comparator - 3 comparison conditions
Usual practice
AI + good professional integration
AI + explicit Augmentership intervention
Comparing 2 vs 1 estimates what integrated AI practice brings. Comparing 3 vs 2 estimates the incremental effect of the Augmentership programme being tested. These contrasts are not automatically causal.
ΔAI_Y = Y_good-AI-integration − Y_usual-practice
ΔAUG_Y = Y_Augmentership-intervention − Y_good-AI-integration
Teaching block: these formulas claim no automatic causality. Values from a pilot are managed separately, in Evidence & Results.
Augmentation Value Scorecard
What is assessed
| Domain | What is assessed | Examples of evidence |
|---|---|---|
| Financial | Incremental monetary benefits and costs | Budget, revenue, avoided cost, verified savings |
| Productivity | Time, throughput, cycle time | Operational metrics |
| Quality | Defects, errors, detection, consistency | Quality metrics |
| Capability | Repertoire, reliability, transfer, resilience | Capability assessments |
| Agency & working conditions | Intervention, contestability, critical skills, workload | Scenarios, surveys, observations |
| Organisational resilience | Knowledge distribution, key-person dependency | Team/organisation indicators |
| Distribution | Who benefits and who bears the costs | Outcomes by affected group |
Effective agency and agentic AI
Agency is an assessable condition of legitimate augmentation.
Permissions
Identity, authority, scope, duration, resources.
Boundaries
Technical and risk limits, including sub-delegation.
Escalation
Triggers, recipient and response expectations.
Oversight
Usable logs and sufficient visibility.
Accountability
Clear responsibility for design, deployment, management and operation.
Revocation
Effective withdrawal and propagation of rights.
Intervention
Suspend, correct, contain or roll back.
Recourse
Ability to contest and repair consequences.
For agentic AI, Augmentership does not replace agent safety or governance standards. It asks one further question: which human capabilities and agency conditions are needed for delegation to remain controllable?
Augmentership Principle
« Augmentation must be designed, evaluated and adapted - not merely declared. »