Pūnaha enterprise AI search and workflows

From a question to an accountable action

Pūnaha is Hot Desk's enterprise AI search and workflow product. It brings organisational knowledge, supported AI models and workflows together in one system.

The word pūnaha means system in te reo Māori. The product name is sometimes typed as Punaha when a macron is unavailable; we retain the correct spelling, Pūnaha, throughout this site.

A demonstration can start with an approved set of documents, retrieve the material relevant to a question, show the sources used for an answer and pass the work into a workflow. That workflow can pause for a person to approve or reject the next step before another system is called.

Look beyond the search box

When comparing enterprise search tools, ask which sources people can search, whether existing permissions still apply, which AI model receives the information and what the system is allowed to do next. Pūnaha brings those decisions together while leaving the organisation in control of the design.

Search knowledge without hiding the source

Pūnaha supports semantic and hybrid retrieval across tenant-aware knowledge stores. It can work with common text, email, Office and OpenDocument formats, as well as RTF, EPUB and PDFs that already contain readable text.

The organisation still decides which sources are authoritative, who may use them and how current information is maintained. Image-only PDFs need OCR before Pūnaha can use their text.

Choose where it runs and which models it uses

Pūnaha can run on customer-controlled Windows or Linux infrastructure or in Pūnaha Cloud, subject to the agreed architecture. Model options include local GGUF execution through llama.cpp, OpenAI and supported OpenAI-compatible endpoints, Azure OpenAI, Anthropic and Google Gemini.

Model choice is not only a feature decision. The data path, provider terms, performance, support and the information being handled all matter.

Some buyers describe this requirement as a private AI platform. We use that term carefully: customer-controlled hosting is one part of privacy, alongside identity, permissions, model data paths, logs, support access and day-to-day operation.

Workflows you can inspect

Teams can build visual workflows using prompt, retrieval, model, approval and merge steps. Developers can add conditions, generic REST calls, read-only MySQL access and web-search backend nodes. Each workflow can be validated, versioned and published, with a durable record of its execution.

How to see Pūnaha today

Pūnaha is available through a demonstration and a guided implementation discussion. Product tours use representative or fictional information rather than customer data. Online self-service purchasing is not yet available, and we do not publish customer references at this stage.

Explore Pūnaha, review the technical detail or talk to us about AI governance and implementation.