From a working workflow to a platform

A workflow that ran once in a pilot is not yet a platform. The platform is the identity the workflow uses, the list of actions it may take, the log of what it did, and the path a person uses when the outcome is wrong. Those pieces are the same ideas as the Automation and Agentic AI practices, sized to the workflow you already have.

You should not have to rebuild the first workflow to add a second one. The audit named the boundary. The pilot proved one path inside it. Scaling adds another path, or a private place to run the same path. This is part 2 of From One Workflow to a Factory. Part 1 is Start with one workflow.

What stays the same

The decision does not get a new design because the deployment changed. Hosted, private cloud, or on premises are places to run the path you already agreed. The data stays in Australia, under your control. Approvals and the audit log stay with the workflow. A decision that changes money, access, or a record about a person still stops for a person.

If the agent is wrong, the owner can reject the outcome. The path that does not use the agent stays available. That is the same rule as in the pilot. A platform that drops it has not grown. It has forgotten the constraint.

When a factory is the right conversation

An AI factory is the system of compute, data and facility for an organisation that needs its own accelerated computing. Most workflows never need that. When one does, the AI Factory practice is where the conversation moves. The workflow you already run is an input to that design. It is not something you throw away.

The small business automation page is the engagement that starts with one workflow. This series is general information. It is not a proposal, and it does not set a price.