Tacit Intelligence is a critical player for the development of specialized AI agents. While current agents often rely on generic LLM 'reasoning' or basic RAG (Retrieval-Augmented Generation) to make decisions, Tacit provides a way to encode the specific policy and intuition of human experts into the agentic loop. For anyone building agents for high-stakes enterprise tasks—such as automated underwriting, threat intelligence, or industrial optimization—Tacit represents the 'expert brain' that can guide an agent's actions.
In the agent stack, Tacit sits at the intersection of knowledge representation and decision logic. They are essentially championing a move away from black-box outcome matching toward transparent, expert-guided reasoning. This matters because the reliability of autonomous agents depends entirely on the quality of the 'policy' they follow. By capturing tacit knowledge, Tacit allows organizations to build agents that don't just act, but act with the specific judgment of the company's best people.
Tacit Intelligence is a New Zealand-based firm founded on the idea that the most valuable data in any organization is the knowledge that people cannot explain. This concept, known as tacit knowledge, refers to the intuitive decision-making that a veteran insurance underwriter or a special operations commander develops over decades. General large language models (LLMs) are typically trained on public data or outcomes that have been labeled by non-experts. Tacit argues this leads to AI that 'looks' right but fails to 'think' right when faced with the nuance of professional judgment.
The company is led by CEO Joe Cole and a team that previously built Smartabase, a performance software system used in high-stakes environments like elite sports and emergency services. This background in measuring human performance is the foundation of their current work. They are not building another chatbot; they are building a framework to observe experts as they work, record the specific factors they weigh, and encode that reasoning process into a proprietary AI model. The goal is to ensure that a company's internal expertise only compounds over time, rather than disappearing when key employees retire or leave.
The team’s history in elite performance is critical to their methodology. At Smartabase, they worked with organizations where expertise is often a matter of life or death. They noticed a persistent gap in AI development: general models tend to rush toward a conclusion based on pattern matching. In contrast, true experts spend the beginning of any problem-solving task asking questions and navigating uncertainty. Tacit focuses on this early-stage reasoning. By capturing the questions an expert asks in the first ten seconds of a task, they build models that reflect a specific organization's 'gut instinct.'
Based in Christchurch, the leadership team includes Roanne Hurley, PhD, who leads the research into the science of expert cognition, and Alex Dong, a serial entrepreneur who previously sold a company to YouTube's founders. This mix of research-heavy cognitive science and experienced engineering suggests a focus on the technical mechanics of human-to-AI knowledge transfer rather than simple data labeling.
Tacit’s primary critique of the current AI market is the reliance on outcome-based training. They argue that a novice and an expert can reach the same conclusion for different reasons, but the novice's reasoning will eventually fail in edge cases. By training AI to think like experts—understanding the factors and uncertainties involved—the resulting systems are meant to be more reliable in complex environments. This approach is particularly relevant for sectors like insurance, where identifying fraud or assessing risk requires noticing patterns that are not explicitly documented in training manuals. The company operates as a specialized partner for organizations that consider their internal judgment to be their primary competitive advantage.
A framework to capture expert reasoning and deploy it at scale.
Tacit Intelligence Company is hiring