Build AI into the work a business actually does.
I work from a business problem toward a usable system: the information people need, the actions tools can take, and the decisions someone must own.
Knowledge that supports the next task
For a survey and data firm in Australia, I provide ongoing technical leadership and develop industry-specific workflows. These turn training videos into procedures, retrieve technical knowledge, and connect AI agents to business tools through the Model Context Protocol.
From an inquiry to useful follow-up
In my own ventures, I have built Magic Receptionist onboarding, call and lead workflows, CRM connections, partner onboarding, referral attribution and commission visibility. I am adapting the platform with a launch partner toward remarketing and lead generation.
I’m preparing these products for launch with partners.
Design and software belong together
For Magic Sites, my development work includes website editing, preview, publishing and version history. For BestReviewedSantaFe, I implemented lead capture, source attribution and CRM delivery tracking. The interface is only useful when the handoff behind it works.

One workflow, from question to usable procedure
Here is an illustrative training-to-procedure workflow, using synthetic content. It shows the design questions behind my industry-specific knowledge work.
- Start with the task. A team member asks how to prepare a field job.
- Retrieve the source. Find the relevant training segment and retain its timestamp and document version.
- Prepare the next action. Draft a checklist with a source link beside each instruction. Leave unsupported steps as questions.
- Review with a person. A knowledgeable team member checks the procedure before it becomes team guidance.
- Save the result. Keep the reviewed procedure, its source references and the revision together so the next person can use them.
A useful result is a procedure someone can follow and check. Retrieval alone does not establish that result.
What agents need to act responsibly
My infrastructure work includes myMCP, Magic Connect and Agent Vault. CARD Network Architecture explores linked knowledge and traceable evidence, with reference implementations in development. The practical questions are context, access, provenance and how a result can be checked.