Your team, producing like a team twice its size. Not by working longer — by putting AI to work on the repetitive tasks, and turning your sharpest people into the ones who build the tools that handle it. On your real workflows, with your real data. On-site, hands-on.
Give your team AI and they get faster — call it 2x. The 10x comes when your capable people start building with it: turning the work you repeat into tools that run themselves, fed by what your company already knows. Not better prompts. Better leverage.
Most AI trainers have never shipped a production system. What your team learns is the same method that built the invoice-intelligence system on this site — running live at a marine manufacturer, where it surfaced $261,664.92 in overbilling and price creep in year one. Not theory; what works in production. See the system →
One progression. The first three moves make everyone sharper. The fourth is where output multiplies.
Most AI disappointment is a targeting problem, not a tool problem. Your team learns what AI is for — high-volume, language- or data-heavy, easy to check — and what to keep human.
Not a search box. Give it the goal, the context, the constraints, and one example of good. Then direct the draft instead of doing the work.
Scan for what matters, check harder when the stakes are higher, keep a human in the loop where it counts. This is what makes it safe to hand real work over.
Take a workflow your team does fifty times a month and build it once: a reusable skill, a small automation. A prompt helps once; a system compounds. It's how the software on this site was built.
Fed by your context: each tool runs on what your company already knows — past work, playbooks, your data — so the output sounds like your best employee, not a generic bot.
a-skill/ ├─ instructions.md · how it thinks & decides ├─ knowledge/ · your playbooks, your data ├─ tools/ · the actions it can take └─ guardrails.md · what it must never do
Bring your bottleneck. From day one everyone works on a real task, not a toy — scaled to where they're starting. The whole team leaves with AI doing real work for them; the builders leave with a tool they made themselves.
Copilot, ChatGPT, a few logins scattered around — most operators are already paying for AI and getting a fraction of it, because people treat it like a search box instead of feeding it context. I don't replace what you run; I make it deliver. The method sits on top of whatever's already deployed — no migration, no rip-and-replace, no new tool for anyone to learn.
phosops-starter-kit/ ├─ 00-start-here.md · paste me into any AI — I set up the rest ├─ skills/ · templates to turn workflows into tools ├─ context/ · your company's knowledge, structured └─ ai-policy.md · safe-use guardrails
None of this is notes to file away — it's capability your team keeps and uses Monday morning.
One team or department at a time — roughly 6–25 people, one or two days on-site. We start with the core ideas — how to actually get leverage from AI — then go hands-on for the rest, building against your real work. Then 30 days of support so the first tools actually ship.
The real leverage at the top isn't executives typing faster. It's seeing what to change. A dedicated track for the people who run the place.
When drafting, research, and analysis cost near zero, the question shifts from "how do we do this faster" to "what would we do if it were free." Your leaders leave asking the second one.
A ranked read on your operation — which functions AI actually moves — so budget and attention follow the return, not the hype.
Bolting AI onto an old workflow barely moves the needle. The step-change comes from redesigning around what just got cheap — and spotting where the bottleneck moves next.
Your team learns what's safe to feed AI and how to handle sensitive data — and leadership leaves with a one-page usage policy. No more shadow-AI risk.
After the intensive, a brief to leadership: what your team built, what they can now build on their own, and the bigger systems worth building properly — the ones I can build with you.
Thirty days of office hours after the sessions, so the first real tools actually ship instead of fizzling out.
By the end your capable people can build, ship, and extend real tools on their own; everyone else can put AI to work on what they do every day. That's not a promise about someday — it's what's true when I leave the room.
On-site, one or two days. Both tracks in one engagement. Standalone, or bundled into a build. It scales with team size and days — a fraction of what a Big-4 change engagement runs, and your people keep the capability for good.