I build systems that handle transactions.
Categorizing them, catching duplicates, reconciling accounts, and flagging the ones a human needs to look at.
If you run more than one company
Logging into two or three bank portals and exporting what's there
Working out which transactions belong to which entity
Finding the transfer between your own accounts that got counted twice
Getting a P&L back that describes a month which ended three weeks ago
quickbooks doesn't consolidate across entities · the work lands on you
Rules run it. You stay last.
Most transactions post themselves. The rest wait for you. In my demo run, 454 lines went from bank feed to a finished P&L in about five minutes.
Your numbers are current while the month is still running, not three weeks after it closed.
Month-end becomes a review — you approve a handful of exceptions instead of rebuilding the ledger by hand.
The work doesn't leave with me. The rules are yours, in your spreadsheet, in plain language you can read and change.
No new software, no new login, no new process to learn.
System Execution Model
Compare fragile manual spreadsheet work with automated transaction routing.
Log into 3+ portals, export CSVs, manually normalize formats.
Fragile spreadsheets; high risk of misattributing entity expense.
Inter-company transfers frequently get double-counted in P&L.
Books close 3 weeks late. Decision makers see outdated numbers.
Real-time transaction stream directly inside your Google Workspace.
Rules you can read and edit. In the demo run, 411 of 454 lines routed without a human.
AI only assists when confident. Unrecognized lines wait for review.
Ledger and dashboard figures update while the month is still active — the full demo run finished in about five minutes.
See it work
Built to show what the approach looks like, not as a product. Your version would look different.
Estimate what would still need you
Enter your own numbers. This applies the ratios measured in my demo system to your volume — it is an estimate built on one dataset, not a promise about yours.
A single run of the demo system — 454 transactions, start to finish in about 5 minutes, including categorization, reconciliation and dashboard figures.
• 411 routed without a human (90.5%)
• 43 held for a decision (9.5%)
One run on my own demo data. Your books will differ — messier categories and unfamiliar vendors mean more lines reach you, at least at the start.
The first two figures apply demo ratios to your volume. The third is simply your own hours × your own rate — what the time is currently costing, not a saving I am claiming. What the system actually returns is the part of it spent on lines a rule could have handled.
I don't store it
How working with me works
Click any step to view details in focus
A conversation
60 minutes, free. You tell me how you handle transactions today. I ask questions. I don't present anything or pitch you.
A written assessment
Within 24 hours, free. Two pages: what your current process looks like, where the manual work sits, what could be automated, and what I'd leave alone. You get this whether or not we work together. If I don't think I can help, I'll say so here — and tell you what I'd do instead.
Paperwork before anything is built
A mutual NDA and a short services agreement. Your data stays in your own Google Workspace and bank connection — I never hold a copy, and you can revoke my access at any time without ending the agreement.
A small first project
Two weeks, fixed price. One workflow, one entity, scoped in writing with the outcome agreed in advance. Half up front, half on delivery. If you continue afterwards, that fee comes off your first month.
Ongoing, if it's working
A monthly arrangement covering rule changes, new accounts, and support. Thirty days' notice either way, no reason needed.
Questions people ask before we start
Can you move money out of my accounts?
No. Bank access is read-only through Plaid, and the system has no payment capability. It can read transactions. It cannot initiate anything.
Do you see my banking passwords?
No. You connect your accounts through Plaid yourself, in your own name. Credentials go to your bank, never to me.
Where does my data live?
In your Google Workspace, on your account, billed to you. I don't keep a copy on a server, a database, or my own machine.
What happens if you disappear?
The system keeps running. The rules are in your spreadsheet in plain language, and you can edit them yourself. Nothing depends on me being reachable.
Can I revoke your access?
Any time, from your own admin console, without telling me. Nothing breaks.
What if the AI categorizes something wrong?
It only touches transactions your rules don't recognize, and only above a confidence threshold you set. Below that, it doesn't decide — it hands the transaction to you with a suggestion. Nothing posts without a rule match or your approval.
Are you an accountant?
No. I don't do tax, audit, payroll, or accounting advice. I build the mechanism that handles transactions; your bookkeeper or CPA keeps the judgment.
Do you replace my bookkeeper?
No. Most of my work makes their job faster — fewer manual entries, cleaner data, exceptions surfaced instead of buried.
You're based in Vietnam. How does that work?
I work US mornings, so we overlap. Everything I build lives in your accounts rather than mine, so nothing depends on where I am. And there's an NDA and a services agreement before I'm given access to anything.
What do you need from me to start?
Read-only bank access, your category list or an hour to build one together, and one person to answer questions. That's it.
How long does a first project take?
Two weeks for one workflow and one entity, scoped in writing before anything starts.
What if it breaks?
You'd see it in the review queue rather than in your books — nothing posts silently. Support response is within two business days, written into the agreement.
Long Nguyen
Seven years in banking and fintech, currently on the systems side of a digital lending product — debt classification, fee calculation, reconciliation, and partner billing where real money moves.
• MSc Finance, Colorado State University
• CFA Level II candidate
• Based in Hanoi · Working US morning hours