Stop re-explaining yourself and your company to AI.

• You use AI every day, mostly in a chat window. It helps, and you know it could do a lot more. The trouble: everything it needs to know is in your head, and it starts from zero when the chat closes.
• The output is uneven. Some days it nails it; other days it comes back generic and you rewrite the whole draft because it reads like AI.
• You'd like to go further, into agents and workflows that run on their own, but the business takes every hour and most of your headspace. A serious setup looks like a project you can't staff, and the pace of new tools is its own kind of noise.
• You want your team using AI well too, and you'd like to be the one showing them how.

The bottleneck usually isn't the model. It's that the model doesn't know you. This is the fast lane: your full context set up once, in your own account, so what AI gives you is usable the first time.

And when you are ready to bring your team along, that is a conversation we can have next.

1. 1 · Intake and scopingYou leave with a plan.

   In a live interview, we cover how you make decisions and who matters to them, your top priorities this quarter (which everything else ties back to), where your week actually goes, sorted into what should save you time and what should expand what you can do, and the two workflows most worth setting up first. So we don't start from scratch, the foundation also draws on a prompt you run on the AI tools you already use, which extracts what they already know about you (you review the report and strike anything before it reaches me), and on what is already public, like your website and LinkedIn. We write down where you start, so we can see what changed later.

2. 2 · Your knowledge baseIt knows your context.

   How you think and decide, your company, your priorities, and a relationship vault: a file for every person, company and meeting that matters, so its understanding of your network can grow with every conversation. An interactive setup prompt drafts it from the intake and your own documents, then asks you to confirm or correct each part as it goes. You also set up the instructions that tell Claude where your knowledge base is and how to use it: a CLAUDE.md file, Claude's global instructions, and your project instructions in Cowork. Everything else reads from it.

3. 3 · Your voiceIt writes like you.

   A drafting skill trained on your own writing. You correct its drafts, every correction is logged, and the skill is updated from them.

4. 4 · Your first workflowOne workflow, running.

   The first of your two, set up by you in your account and used in your real week before the next session.

5. 5 · Your second workflow, and the read-outTwo workflows running, and a one-page read-out.

   The second workflow set up, and my recommendations for where to take your AI OS next: the next steps you could take on your own, so you can decide whether to keep going yourself or keep working together. If I notice other gaps along the way, I will name those too.

6. Three weeks later · Impact session, includedWhat you use and how, and what to tune.

   We look at how you are actually using it against where you started, and fine-tune what needs it.

Every session starts from what happened when you used the system that week. The sessions fit your calendar; plan on four to six weeks from intake to read-out. You can bring your COO or chief of staff into the sessions with you, or have them lead the setup in your place. Every workflow you add after this starts from the same foundation.

Beyond these numbers: branded HTML proposals, my website's design and updates, and content. I now publish a point of view every week, where before I published only when I changed jobs. That is the part AI adds to what I can do, not the time it saves, and none of it is counted above.

Monthly averages, August and September 2026; both hour figures are my estimates against my own before-AI baselines. Work per hour of direction, including scheduled routines: September 2026 (3.3) and April to September (5.2). Counting only the work I direct, it is about 3 (2.9 and 3.3). Measured nightly from my own session logs, which undercount Cowork.

• The account is yours, in your name or your company's, on a standard Claude plan. Pro, at $20 a month, is enough to start; heavy daily use may call for a Max plan, from $100 a month.
• Your knowledge base is plain files you own. They live as markdown files in a dedicated AI workspace folder you set up in your Google Drive, and Claude is given permission to read and update them. The files work with any AI tool and move with you. I use Claude because of how it handles security; the skills and routines are Claude's, and would need setting up again elsewhere.
• You approve what it knows about you. The files about you and your company go in only after you have corrected and signed off on them.
• You are the only operator. I guide on a call and never log into your account.
• What I see: by design, only the reports your prompts produce. The prompts strip names, companies, figures and dates inside your account, and you see each report before you send it to me. I never see your inbox. Board and investor material reaches me only as a template.
• Regulated records stay out. Patient health information, payment card data and student records are not connected to the setup. It works from your own communication, calendar and documents.
• Connectors are yours to grant, one at a time (calendar, mail, files), and yours to revoke.
• It drafts and never sends, and asks before it deletes. Its authority is written in a file you can read and change.
• When a source is missing, it says so. Every analysis is set up to stop and name what it is missing rather than fill the gap with a plausible guess. It is the first thing I would guide you to set up.