Black Swan Consultancy uses artificial intelligence as a core analytical tool for document analysis, strategic research, and advisory work across our practice areas.

Where client engagements involve sensitive commercial, financial, or operational data, we run AI inference entirely on our own hardware. Our local stack is built on Apple Silicon with Ollama, LM Studio, and Qwen large language models. This means your client data never leaves our systems and is never transmitted to third-party cloud services. No data is used to train external models. No conversation history is retained on external servers.

For general research and open-source intelligence work, we use frontier cloud models where appropriate. The distinction between the two environments is maintained deliberately and explained to clients at the outset of each engagement.

This approach reflects a broader philosophy: the same rigour we apply to mining safety systems and M&A due diligence, we apply to data handling. Clients expect judgment, discretion, and analytical depth. Our AI infrastructure is designed to support all three.

Human oversight and accountability

AI at Black Swan Consultancy is a research and drafting tool, not a decision-maker. It supports document analysis, strategic research, and first-draft thinking. It does not replace the judgment calls clients engage us for, on M&A risk, on transformation strategy, on leadership advice. Every AI-assisted output is reviewed and signed off by Chris or Donna before it reaches a client. If AI gets something wrong, that is caught in review, not in delivery.

Records and traceability

We keep a record of where AI has materially contributed to an engagement, consistent with public sector expectations for traceable data and auditable decision-making. For engagements involving government agencies or Crown entities, this means we can account for what was AI-assisted, what was human judgment, and how the two were separated, in line with Public Records Act obligations and the NZ Public Service AI Framework's approach to accountability and traceability.

Māori data and inclusive development

Where our work touches Crown, community, or public outcomes, we recognise that AI-assisted analysis intersects with Treaty of Waitangi obligations and Māori data sovereignty. We do not use AI to generate or infer conclusions about Māori interests, cultural values, or community outcomes without human review grounded in direct relationships and mana whenua input. AI supports our research and drafting; it does not substitute for engagement with the people and communities affected by the work.

Chris Goddard — AI background

Chris's engagement with artificial intelligence predates the modern era of commercial AI by three decades. He completed postgraduate study in AI in the early 1990s at a time when neural networks were largely theoretical constructs and expert systems were the dominant applied paradigm. That foundation gave him an early and durable understanding of the gap between AI capability and AI hype, a distinction that remains commercially useful today.