Intelligence Into Action
From understanding to action

Human accountability by design

Responsible AI in Practice

Principles matter when they shape the everyday design of tools, workflows, environments, learning, review, escalation and accountability.

Atkinson Film-Arts visual for this page.

In brief

Responsible AI in Practice

Who this is for

Executives, technology, governance, information, workforce and transformation teams.

What this page helps you do

See Atkinson's approach to human purpose, accountability, information protection, proportional risk, accessibility, evaluation and transparent AI maturity.

Eight operating principles

Responsibility is expressed through design decisions

01
1

Purpose and necessity

Use AI where it meaningfully supports an approved outcome, not because it is available.

02
2

Human accountability

Name the people who decide, review, approve, intervene and remain responsible.

03
3

Information protection

Classify data, research, IP and confidential material before choosing a tool or environment.

04
4

Proportional risk

Match controls, evidence and approval to autonomy, consequence and affected people.

05
5

Transparency and maturity

Make sources, limitations, synthetic content and concept/pilot/production status visible.

06
6

Evaluation and monitoring

Test quality, failure, misuse, accessibility and operational fit before and after deployment.

07
7

Inclusion and accessibility

Design for diverse users, language, ability, context and meaningful human alternatives.

08
8

Lifecycle responsibility

Own changes, incidents, support, versions, retirement and ongoing improvement.

Human-in-the-loop is not a slogan

Define what the person is actually responsible for

A human checkpoint is useful only when the reviewer has authority, time, information and a clear quality standard. Atkinson identifies the decision boundary, evidence required, escalation path and record rather than using human oversight as a generic reassurance.

01

Questions to answer

  • What may the system see and do?
  • What must a person verify?
  • Who can approve or stop the workflow?
  • How is uncertainty exposed?
  • What happens when the system changes?

Turn principles into an operating framework

Explore Governance Workshops