I build AI into software that already has a job to do: an assistant that answers from your own documents, a classifier that reads receipts, a summariser that turns a long report into something a person will read. Usually on OpenAI, sometimes Anthropic or Gemini, with the choice made on the task rather than on the brand.
Grounded in your content
A general model that has never seen your documents will confidently invent an answer. The useful pattern is retrieval: the model is given your material at the moment it answers, and told to say when it does not know. That is the difference between a demo and something you can put in front of staff.
What I will talk you out of
AI is not the answer to a process that is broken, and it is a poor fit for anything that has to be exactly right every time. If a rule engine or a form would do the job, that is what I will build. It is cheaper, it is testable, and it does not surprise you.
Cost and data
Every AI feature has a per-use cost and a data question attached. I will tell you what a feature will cost to run at your volume before it is built, and where your data goes when it runs. The free tools on this site are built the same way, which is why several of them sit behind an access code.
Running a business in Canberra? The AI product development in Canberra page covers the assistants, automations and AI products I build for local small and medium businesses.
