Let AI create more opportunities for you.
Your team is spending its week on work that doesn’t need them. I build the things that take it back. Everything on this page is a working example, not a slide.
What are we building today?
The prompt is the easy part.
See for yourself.
Working through it
QR Code Generator
✓ built just nowThe mark in the middle is mine, because this is my demo. A real build carries yours.
The building is the easy 10%. Knowing what to build, and making it fit the way a business actually runs, is the 90%. That 90% is the job.
How I work
No new software. I build inside what you already have.
Microsoft 365, Google, Slack, Claude. I connect to what is already there and build the workflows on top, so the budget already being spent goes further. No new platform to buy.
Connected tools: Cursor, Claude, Microsoft 365, Codex, Salesforce, Google, Slack.
Where your data goes
Nobody hacks you. Somebody pastes it into the wrong tool.
Almost every leak starts with someone trying to get their work done faster. Every tool they sign up for is another company holding your data, so I count them before anything gets connected.
Reading is transmitting
An agent opening a file on your laptop sends that file off your laptop. "It's just in a folder on my computer" is false comfort.
One tool, three companies
The assistant is one vendor. The model underneath is another. The host is a third. Count them before you count features.
Resolving the processors behind a single coding tool returns three: a code assistant handling the interface, a model provider running inference, and a cloud host holding storage. Every one of them has to be named in your data processing agreement.
Size the tool to the data
Aggregate data can go anywhere, and should. Other people's personal data never meets a tool that acts on its own.
A policy check weighs two dials against each other. Aggregate data scores low on sensitivity and gets full agent access. Internal notes score medium and get access scoped to one folder. Personal data scores high, which drops the autonomy ceiling to almost nothing: chat only, no agent access.
Extract once, by hand
The risky step happens once and deliberately. De-identify first, then let the fast tools loose on that.
Rent vs own
Fewer vendors, smaller bill, less exposure.
It is also the cheaper end of that trade. The same job, rented and then owned, and what it does to the bill. Figures are illustrative, not client results.
Customer journey
Automating a broken step just breaks things faster.
One lifecycle end to end, with the number each step is accountable for. Yours looks different. The habit does not.
“Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise.”
Decisions
Search became answers. Most dashboards never noticed.
People ask an assistant now instead of typing into a box, which is what AEO and GEO are actually about. Total discovery holds steady while the mix inverts underneath, so a report built for the old front door will tell you nothing changed.
Total discovery is flat. The mix inverted. Nobody lost demand, the front door moved.
Traffic is collapsing. Cut the content budget.
Input
Wrong data analyzed well is more dangerous than no data. It arrives with confidence attached.
Relevance
The right metric is the one that would change what you do. Everything else is reporting.
Drift
A metric can keep its name and quietly stop describing your market. Re-ask what it measures.
FAQ
Not as something that replaces anyone. As something you buy time and reach with. The useful question is never what AI can do, it is what you are doing that does not need you. Those are the hours it hands back, and it can give someone a skill on Tuesday they did not have on Monday. The opening is in what you do with that. It was never in the tool.
Because a model cannot tell the difference between knowing something and being able to produce a sentence shaped like the answer. It fails hardest where it cannot see: a page it is blocked from fetching, a market with no English coverage, your own numbers. Ask any AI how to rank on Korea’s search engine and watch it invent something. The fix is almost never a better model. It is giving it the thing it was missing.
More companies touch it than most people count. One AI assistant is usually three separate businesses: the interface you log into, the model underneath it, and whoever hosts that. Each one is a processor and each one belongs in your DPA. The second thing worth knowing is that reading is transmitting. The moment a tool reads a document in order to answer a question about it, that document has left your building.
No, and if that is the goal I am the wrong person to call. AI has no idea what your organization considers important. It cannot look at thirty things and tell you which four stop a person or a deadline today, unless somebody writes that rule down first. That somebody works for you. What I build takes the part that never needed their judgment, and they get the week back.
It starts with watching how your team actually works, not how the process document says they work. The gaps are almost always in the handoffs, not the tasks. Then I find the lowest-cost way to close them that does not weaken anything on the security side, because the fast fix and the safe fix are rarely the same one. The goal is more output from the same people, not fewer people.
Almost certainly not, and that has been true of nearly every job I have taken. What carries over is the method, not the sector. You work out what the audience actually does before you decide what to build, and that part is the same whether you are selling season tickets or school places.
Something I have not answered? Ask me directly.
Your industry probably isn’t on this list.
It wasn’t on mine either.
- PGA Tour
- EF Education First
- New York Life
- Tampa Bay Rays
- LX Pantos
- U.S. House of Representatives
About & contact
Let's talk.
I'm Steve Nam, an AI Operator. Growth and operations across very different industries, the last three years building the workflows that make real teams faster. If you're working out how AI fits yours, I'm easy to reach.