1. Purpose of this policy

This policy explains how Midshift uses artificial intelligence in a responsible, safe and transparent way. It is designed to help clients and stakeholders understand where AI is used, where human judgement remains essential, and what controls are in place to reduce risk.

Midshift’s approach is informed by New Zealand responsible AI guidance for the public service and Australia’s AI Ethics Principles, including safe, transparent and accountable AI use, privacy protection, fairness, human oversight, security, reliability, contestability and accountability.

2. How Midshift uses AI

Midshift uses AI to support process improvement engagements. AI may assist with structured stakeholder interviews, transcript analysis, theme identification, process analysis, opportunity identification, report drafting and plain-English summarisation.

• AI helps process and organise information collected during an engagement.
• AI helps identify patterns, risks, pain points and improvement opportunities.
• AI helps generate draft analysis for human review.
• AI does not make final decisions for clients.
• AI does not replace professional judgement, business analysis review or client accountability.

3. Human oversight and accountability

All completed Midshift reports are reviewed by a senior business analyst before delivery. This review checks whether the analysis is reasonable, relevant to the engagement context, grounded in the information provided, and presented in a way that clients can understand and challenge.

• A human reviewer remains accountable for the final delivered report.
• AI output is treated as draft support, not as final professional advice.
• Client decisions remain the responsibility of the client organisation.
• Recommendations should be assessed against operational reality, legal obligations, employment obligations, privacy obligations and organisational context before implementation.

4. Privacy and personal information

Midshift is designed to collect only the information needed to deliver the engagement, as explained in our Data Handling and Privacy Policy pages. Where personal information is involved, we aim to handle it consistently with New Zealand Privacy Act 2020 principles and Australian privacy expectations, including transparency, reasonable collection, secure handling, limited access and deletion on request where appropriate.

• We avoid collecting unnecessary personal information.
• We ask clients not to upload sensitive information unless it is genuinely required for the engagement.
• We use engagement data to deliver the service, not for advertising.
• We do not sell client or stakeholder data.
• We explain AI processing separately in our Data Handling page.

5. Transparency

Clients and stakeholders should know when AI is part of the service. Midshift is transparent that AI is used to support stakeholder interviews, transcript analysis and report generation.

• We disclose that AI is used in the Midshift process.
• We explain the role AI plays in analysis and report drafting.
• We identify the AI provider used for processing in our Data Handling page.
• We make clear that final outputs are reviewed before delivery.

6. Fairness and bias awareness

AI systems can reflect bias, incomplete information, unclear assumptions or overconfident patterns. Midshift treats AI-generated analysis as something to be checked, not blindly accepted.

• We review outputs for unsupported claims, unfair framing and overstatement.
• We avoid using AI to rank, discipline, profile or make employment decisions about individual workers.
• We focus on process-level analysis rather than blaming individuals.
• We encourage clients to test findings with operational context before acting on recommendations.

7. Safety, reliability and quality control

Midshift uses AI in a controlled way to support bounded process improvement tasks, consistent with practical responsible AI practices such as risk management, human oversight and clear governance. We do not use AI as an unrestricted decision engine or autonomous management system.

• We use structured prompts and defined engagement workflows.
• We review outputs before they are delivered to clients.
• We avoid presenting AI-generated content as certain where the evidence is incomplete.
• We distinguish between stakeholder input, analysis, assumptions and recommendations where practical.
• We may ask for clarification where the process scope or evidence is unclear.

8. Contestability and correction

Clients should be able to question, correct or challenge Midshift outputs. If a report includes an inaccurate process detail, misunderstood context or unsupported conclusion, clients can contact us for review.

• Clients can request clarification about findings or recommendations.
• Clients can identify factual errors or missing context.
• Where appropriate, we may revise a report to correct material inaccuracies.
• Stakeholder feedback should be considered before major operational changes are implemented.

9. Acceptable AI use

Midshift should be used for constructive process improvement, not surveillance, punishment or covert monitoring. Clients must not use Midshift to process information they do not have the right to provide.

• Do not upload confidential, personal or sensitive information unless you have authority to do so.
• Do not use Midshift to covertly monitor staff or stakeholders.
• Do not use Midshift outputs as the sole basis for employment, disciplinary, legal or financial decisions.
• Do not submit false, misleading or deliberately incomplete information.

10. New Zealand and Australia context

Midshift is built for organisations operating in New Zealand and Australia. Responsible AI use in these markets requires more than technical performance. It requires privacy awareness, transparency, human oversight, fairness, contestability, security, cultural awareness, and practical accountability.

For New Zealand clients, responsible use should consider local privacy expectations, employment context, Māori data considerations where relevant, public trust, and the need for clear human accountability. For Australian clients, responsible use should consider privacy obligations, Australia’s AI Ethics Principles, emerging government assurance practices, and sector-specific obligations where relevant.

11. Security and supplier control

AI use creates supplier and data-flow risks. Midshift limits AI processing to the AI provider described in our Data Handling page and uses supporting service providers only where required to deliver the service.

• We aim to limit data access to what is required for the engagement.
• We do not use client data for advertising.
• We do not intentionally train our own AI models on identifiable client engagement data.
• We explain storage, AI processing and deletion in our Data Handling page.

12. Review and improvement

Responsible AI practice will continue to evolve. Midshift may update this policy as laws, standards, government guidance, AI tools, privacy expectations and client requirements change.

Where material changes affect how client data is processed or how AI is used in the service, we will update the relevant privacy, data handling or terms policy pages and, where appropriate, notify affected clients.

13. Questions about responsible AI use

For questions about Midshift’s AI use, human review process, data handling, privacy, procurement review or responsible AI controls, use the Midshift contact page or contact steve@changeable.co.nz.