How I help

Where AI fits in your work — and setting it up once we know.

I don't arrive with a product to install. I arrive with a way of finding where AI belongs in your work, and the hands to set it up once we know. This is the work, organized by the problems it solves.

Finding where AI fits, before buying anything

The first job is a map of your own work: what you do repeatedly, what only you can do, where the hours actually go. Walk one piece of work from "yes" to "paid" and time the waits — the waits are where the money leaks. From that map comes a short list of places AI helps, a shorter list of places it shouldn't go near, and the privacy ground rules for your profession.

That's the AI Clarity Session: ninety minutes and a written roadmap. You leave with the plan whether or not we go further.

Research and analysis

Reading is most of thinking, and most of it sits on your desk before you ever get to judge. Set up well, AI reads the long thing first — a hundred pages of material, a rule change, a client's whole history — and comes back with what matters, where it found it, and what it isn't sure about.

The rule I set: it shows its work so you can check it, and you do check it. Research you can't verify is a rumor.

Writing and client communication

You already know what you want to say; the cost is in the saying. I take the writing you do over and over — client letters, proposals, memos, follow-ups, the explanation you've given a hundred times — and build the AI a set of instructions that capture your voice, your structure, and your rules about what you never say.

The result is first drafts that sound like you, waiting for your edit. Nothing goes out under your name without your judgment on it.

Document-intensive work

Contracts, reports, plans, applications, filings. The work is summarizing, comparing, checking, and marking up — and much of it follows standards you've held so long you've stopped noticing them.

I write those standards down with you and turn them into a first pass the AI runs: this agreement against your usual positions, this report against your checklist, this version against the last one. Your review is the last step, and it's built in. There's no way to skip it.

Knowledge management and your own knowledge system

Twenty years of work is sitting in your files, your email, and your head, and most of it can't be found when you need it. A knowledge system puts your notes, past deliverables, and reference material somewhere the AI can search — so "what did I tell the last client who asked this?" gets answered in seconds, in your own words.

This is the part of the work that compounds. The longer you run it, the more of your judgment is written down and working for you. Everything lives on accounts you own.

Meeting preparation

Preparing well for a client is expensive, so it often doesn't happen. A prep routine gathers what you already know about the person, the history, the open questions, and the last three things you promised, and puts it on one page before the call. After the call, the same routine turns your notes into a summary and a drafted follow-up. The human step is you reading both.

Choosing tools: ChatGPT, Claude, and the rest

You don't need fifty tools. You usually need one professional-grade subscription — ChatGPT or Claude, chosen for your work and your confidentiality obligations — configured correctly, on an account you own. I tell you which and why, set it up, and give you a plain verdict on anything else that gets pitched to you. "Skip it" is a complete answer, and it's the one you'll hear most.

Workflow design and automation

A routine is a job with a trigger, a set of steps, and a place where a person says yes. I design them in that order: make the job right by hand, make it simple, and only then make it automatic. What gets automated is the gathering and the drafting. What never gets automated is the decision.

Every routine I set up produces its work into a folder you control and waits for your approval before anything leaves it. Change one thing at a time, and know how to put it back.

Private and local AI

Some work shouldn't touch a cloud service at all — because of a duty you carry, a client's requirement, or your own judgment about what's yours. For that there are business-grade agreements that keep your data out of training, settings that keep it out of retention, and, when it's warranted, AI that runs entirely on a machine you own, where nothing leaves the building.

I run a private setup myself. I'll tell you honestly when you need one and when you don't. Most people don't; the ones who do can't afford to guess.

Privacy, confidentiality, and control

Every engagement starts with the question that should be blocking you: can this touch confidential client information? We settle it first — what can go where, under which agreement, with which settings — and the answer becomes the walls around everything we build.

The walls are simple to state. Your data stays yours. Your voice stays yours. Everything runs on accounts you own. Nothing goes out under your name without your judgment on it.

AI adoption, at your pace

None of this needs a transformation program. The right order is one map, one routine that pays off within a week, then the next. When it's running without me, that's the finish line — and then a retainer, if you want one, for the months when the tools change and you'd rather not track them yourself.

What I don't do

• Build anything your clients or your audience see. It all sits behind you, in your back office.
• Run your systems or host your data. Everything is on accounts you own, and you can run it without me.
• Sell you software, or take a referral fee for recommending it. The verdict is the product.
• Automate a decision. The gathering and the drafting, yes. The yes itself, never.