AI automation for work that repeats.

The job comes first.

Keen Concepts is an Australian consultancy for AI and automation. An automation engagement starts from a piece of work the business already does, not from a product that needs a use case invented for it.

A useful scope is one queue, one handoff, or one report. Who starts it, what they look up, what they write, where it is filed, and what still needs a person. If that path is not clear, automation spreads the confusion.

What gets built.

The build follows the path we agree in discovery. Typical pieces:

• A workflow for a repeatable job, with the source documents and systems named.
• A model choice: a cloud model, a managed service such as Microsoft Azure AI Foundry or AWS Bedrock, or a model on hardware you own.
• Connectors into the systems the job already uses, including a custom MCP when an assistant needs controlled access.
• Limits on what can be read, what can be changed, and what waits for a person.
• A pilot with the team that does the work, then an agreed retainer for ongoing maintenance.

The commercial shape is the same as every other engagement: one upfront fee, depending on scope, then maintenance. See how an engagement runs.

When this is the right brief.

It fits owners and teams who can point at the work: the inbox, the claims, the quotes, the weekly pack, the follow-ups that slip. It also fits when a previous automation stalled and needs an honest read on whether to finish it or replace it.

It is a poor fit when the ask is a general assistant for the whole company, with no job named. We would rather narrow the brief than install something nobody trusts on a Tuesday afternoon.

Related.

Private and local AI is the page for work that cannot be sent to a public model. Situations lists the wider mix, including integrations and bespoke platforms. Notes covers the decisions that usually come up before a scope is agreed.