Pūnaha enterprise AI search and workflows

Turn organisational knowledge into governed action

Pūnaha is being developed to combine enterprise search, supported AI models and durable workflows in one environment. Its target launch is September 2026. Until general availability is confirmed, the capabilities described here should be read as planned launch scope rather than a live service commitment.

Search across useful enterprise sources

Planned launch connectors include email, generic APIs, MySQL, SQL Server and PostgreSQL. Search is designed to support keyword, semantic and hybrid retrieval while respecting tenant and source permissions.

Choose where models and data run

Planned model options include supported local GGUF models, Ollama and external providers such as OpenAI, Anthropic and AWS Bedrock. Customers will be able to choose customer-hosted deployment or managed Azure or AWS deployment, based on the required balance of sovereignty, operational responsibility and model access.

Build workflows with control

Planned workflow building blocks include prompts, retrieval, models, human approval, conditions, parallel paths, merging, email, chat, APIs and databases. Versioning, validation, publication, durable execution, approval points, failure paths, history and audit records are intended to help organisations move beyond isolated chat experiments.

Designed for accountable adoption

• OIDC single sign-on, role-based access and tenant isolation
• Source permissions carried into retrieval
• Human approval where judgement or accountability matters
• Choice of model and deployment environment
• Execution history for review and improvement

Pūnaha means system in Māori. It is pronounced pū-na-ha. Product tours show representative workflows and should not be read as evidence of customer adoption.

Explore the Pūnaha product site or discuss AI governance and implementation.