I build and run a private AI assistant desk for one busy professional or a small practice. Email, calendar, admin and follow-ups come back to you as drafts and decisions — not another tool to learn.
Inbox triage, calendar defense, admin follow-ups, research, and errands that don't spend your money.
The last word. Nothing is sent, booked, purchased or signed up for without your explicit yes.
A healthcare operations executive who already runs a full assistant desk of his own, every day.
There's no shortage of AI inboxes and schedulers. Someone still has to set them up, keep them current, and watch what they send. Usually that someone is you.
A subscription can draft a reply. It won't learn your house rules, protect your Friday, or chase the vendor who went quiet — unless you keep prompting it.
Managed executive-assistant services commonly run $3,000+ a month, plus the time to manage a person. Great for some. More seat than many people need for admin overflow.
A medical office can't treat patient messages like ordinary mail. The desk has to know what's operations work and what belongs in the clinical system.
I do the design, the setup and the daily running. You get a short brief and a short list of decisions. If managing it would take you longer than doing the task, it doesn't belong here.
Twenty minutes on what's overflowing, what must never be touched, and whether this is the right shape at all.
A written questionnaire: inboxes, calendar, voice, approvals, and a hard off-limits list — including any patient information.
I set up the assistants, labels, briefs and routines around how you already work. You grant only the access we agreed on.
Work arrives as drafts, options and a regular brief. I run the desk; you keep approval on every message, booking and purchase.
“There are no client quotes on this page, because I haven't taken a client yet. What I can show you is the desk I run for myself.”
I operate a roster of about 29 AI assistants under one set of rules. They triage email, coordinate a calendar, deliver recurring briefings and take on single specialized jobs — always as drafts and decisions, never as messages sent on their own. I run it alongside a full-time role in healthcare operations, which is the experience I bring to a practice owner's admin desk.
This describes my own setup. It is not a claim about clinical software, payer contracts or patient care, and it is not another client's case study.
Platform costs are included. 90-day start, then month-to-month with 30 days' notice.
No. As scoped today, the desk supports non-clinical operations — vendor follow-ups, internal admin, the owner's own calendar. It doesn't read charts, process claims, or work in patient inboxes. That would need a separate written HIPAA arrangement and a compliant setup. Anything that looks like patient information is returned to you untouched.
Only after you approve the exact message, or under a narrow written exception you define. The default is the rule I use on my own desk: nothing sent, posted, booked, purchased or signed up for without your yes.
A tool still leaves you to configure and supervise it. An assistant is a hire. This is a desk I build and run for you — the assistants plus an operator who already works this way. You're not managing a product or a new employee.
Only what the work requires, written down in advance — usually specific email and calendar accounts. You grant it, you can revoke it, and there's a do-not-touch list. No banking, brokerage or clinical-system access in the standard desk.
After the fit call and questionnaire, most desks can be stood up within the first two weeks, then tuned. The first 90 days are for getting the house rules right, not for adding every possible job at once.
Anyone who needs a patient-facing receptionist, a clinical or billing specialist, legal or tax advice, or an assistant that spends money on its own. Also not a fit if you want a bot that sends freely without approval.
Twenty minutes is enough to tell whether this fits. If it doesn't, I'll say so.
Three founding spots. Eastern Time. Remote-friendly.