Our position on AI

Pūtake Labs uses AI-supported analysis and simulation as part of its decision intelligence method. AI is a capability we apply, not a product we sell. Clients engage a decision intelligence, assurance and advisory practice, not a technology product.

AI accelerates evidence synthesis, pattern detection, scenario testing, consequence modelling and recommendation development. It does not own accountability, make final decisions or replace human judgement.

How AI is used in our work

AI tools are used to support the following activities within the Pūtake Labs method:

• Evidence synthesis and pattern analysis across large document sets
• Scenario modelling and consequence simulation
• Assumption testing and sensitivity analysis
• Stakeholder response modelling
• Operational pattern detection
• Risk trajectory testing across decision pathways
• Drafting support for communications, reports and decision material

In all cases, human review, interpretation and judgement remain responsible for final outputs. AI-supported analysis is identified within deliverables where relevant to the client’s decision.

The Pūtake Labs method

AI is applied within the Pūtake Labs method, not as a standalone decision-maker. The method includes:

• The Context Engine, which tests evidence quality, source reliability, assumptions and decision context
• The Risk Trajectory Engine, which tracks how risk may evolve, transfer, compound or drift after commitment
• The Direction Engine, which converts evidence and risk analysis into defensible recommendations and decision conditions
• The Kaupapa Methodology Module, which is activated where Māori interests, Māori data, mātauranga Māori or Te Tiriti obligations are in scope

Human accountability

Every recommendation, finding and deliverable produced by Pūtake Labs remains the responsibility of Pūtake Labs. AI assists the analysis. Humans are accountable for the judgement, interpretation, cultural context and final recommendation.

This principle is non-negotiable and applies regardless of the complexity of the engagement or the volume of AI-supported analysis involved.

Data handling

• Client data shared during engagements is treated as confidential unless otherwise agreed
• We do not submit client data for the purpose of training AI models
• Where client data is processed through AI tools, this is done within the scope of the engagement agreement
• Data is stored and managed in accordance with our Privacy Policy
• Clients may request that specific data not be processed through AI tools

Data sovereignty

Pūtake Labs recognises the importance of data sovereignty, particularly for Māori and Indigenous communities. Where engagements involve culturally significant, Māori, Indigenous, community-held or iwi data, we work with clients to establish appropriate governance, access and storage arrangements before AI tools are applied.

Where Māori interests, Māori data, mātauranga Māori or Te Tiriti obligations are in scope, the Kaupapa Methodology Module is activated as part of the engagement method. We take account of Māori data sovereignty principles, including guidance from Te Mana Raraunga, alongside the specific governance requirements of the engagement.

Transparency and traceability

Every recommendation produced through the Pūtake Labs method is traceable back to evidence, assumptions, reasoning and decision conditions. Where AI has contributed to the analysis, this is documented so decision-makers understand the basis of the findings.

The Context Engine governs evidence quality entering the Lab system. The Risk Trajectory Engine tracks how risk may evolve. The Direction Engine governs the quality and traceability of recommendations leaving it.

Bias, fairness and accountability

AI tools can reflect biases present in their training data, source material or user prompts. Pūtake Labs mitigates this through:

• Human review of all AI-supported analysis before it informs recommendations
• Explicit identification of assumptions, limitations and confidence levels in deliverables
• Use of multiple evidence sources rather than relying on any single AI output
• Context Engine review of evidence quality, source reliability and missing perspectives
• Decision Transparency Lab analysis where power, accountability, system constraints or automated decision systems materially shape the decision environment
• Kaupapa Methodology Module activation where Māori interests, Māori data, mātauranga Māori or Te Tiriti obligations are in scope

AI tools we use

Pūtake Labs currently uses the following AI tools in its work:

• Anthropic Claude, for analysis, synthesis and simulation support
• OpenAI, for analysis support, drafting and content production
• Google Gemini, for content generation and analysis support

Tool selection is based on capability, data handling practices and fitness for purpose. We regularly review the tools we use and their policies.

Client rights

Clients have the right to:

• Know when AI tools are being used as part of the analysis
• Request that specific data not be processed through AI tools
• Request information about which AI tools were used in their engagement
• Establish data sovereignty and governance arrangements before AI is applied to culturally significant data
• Ask for human explanation of AI-supported analysis that informs recommendations

Limits of AI-supported analysis

AI-supported analysis can assist with pattern recognition, synthesis and scenario testing, but it can also produce errors, omissions, overconfident interpretations or unsupported outputs. Pūtake Labs does not treat AI output as evidence on its own.

AI-supported findings must be checked against source material, engagement context, human judgement and the decision conditions that matter to the client.

Related policies and services

This Responsible AI Policy should be read alongside the Pūtake Labs Privacy Policy. You can also read more about the wider method through How We Work and The Labs.

For practical AI implementation after decision assurance, see Changeable. For individual AI capability building, see Zero to AI.

Review and updates

This policy is reviewed regularly as AI capabilities, tools and best practices evolve. The current version will always be available at this page.

For questions about this policy, contact kiora@putakelabs.co.nz.

Last updated: July 2026.