I came to AI after 30 years in marketing, not the other way around. When business owners started experimenting with Claude and ChatGPT, I kept seeing the same problem: the tools did not know the business. So I created the AI Brain for Business. It captures the strategy, offers, audience and goals in Business DNA, then the positioning, voice and communication rules in Brand DNA, and connects that knowledge to repeatable workflows. What makes my work different is that I do not sell a folder of prompts. I go into the business, find the bottleneck, extract what the owner and team already know, organise it so AI can use it, then automate only the tasks that should be automated. The human still owns judgement and final approval.

Most of my new business still comes through referrals, existing relationships and people who have seen me teach or work in a room. I would like more of it to come from people who discover my thinking before they know my name, through useful long-form content, search and AI assistants. That is a slower authority-building job, but it is also more scalable and gives a potential client the chance to understand how I work before we speak.

I do not divide work into AI tasks and human tasks. I break each job into steps. For my member community sessions, AI handles low-risk production work such as updating sessions, organising uploads, preparing collateral and drafting communications. I test and iterate before giving it more authority, and anything major stays behind my approval. A major part of my Anthropic and Claude training has been understanding what AI can appropriately support, what must remain human, and where approval gates belong. An article is the clearest example. AI supports the research, outline, headings, search optimisation, proofreading and source-based fact-checking. As a solo operator, I do not have someone at the next desk to give it a second read, so I use separate sub-agents to challenge the draft and provide an independent second pass. I still write the article because the perspective, opinion and lived experience have to be mine, and I verify the final facts. AI handles the scaffolding. I remain the author and accountable publisher.

Yes, but the honest answer is that it is early and messy. On 24 February 2026, a first-time Australian prospect attributed finding me to AI search during a recorded meeting. We did not capture the exact query, platform, answer position or cited source, so I do not claim an AI system ranked me as the best choice. Later, Google Analytics recorded 38 AI Assistant sessions between 28 April and 26 July. That is a small number, not a trend, but it proves the channel now exists and can be measured. I am treating it as a baseline, not a victory lap.

Start with one real business bottleneck, not a shopping list of AI tools. Choose a repeated task, capture the business knowledge and standard it needs, run a small test, and decide where human checking and approval must stay. Measure whether it saved time or improved the work before automating more. I would ignore the pressure to build agents for everything, chase every new model or buy giant prompt packs. The useful work for the next 12 months is less glamorous: better business context, clearer source information, tighter permissions and a person who remains accountable for the result.