Not AI in a report.
AI that does the work.
FloatAI builds AI systems that read, check and produce the documents a business actually runs on — quotations, purchase orders, certificates, reports.
We work with small and mid-sized manufacturers, traders and distributors in Korea, and with the overseas sites they run.
- Based in
- Seoul, Korea
- Part of
- MoatAI Group
- We build for
- Manufacturing · Trade · Distribution
- Your systems
- Kept, then connected
- At handover
- Your team runs it
AI strategy consultancy
Small and mid-sized companies
ERP · spreadsheets · email · messengers
Operating docs + training included
What we build
Six kinds of work we keep meeting
Industries differ. The office does not. These are the six we run into again and again — and what changes when they are automated.
Reading and re-typing
Order forms, requests, bank statements, receipts
- Today
- The same information arrives as a spreadsheet, a PDF, a fax, a chat message — and someone types it in again.
- After
- Whatever the format, the AI reads it and files it. Every value links back to the line it came from in the original.
Nobody re-types anything.
Calculating and deciding
Quotations, material requirements
- Today
- Someone digs through a price-list spreadsheet, and one veteran does the rest in their head.
- After
- The AI matches the request against the price list and drafts the quote. People review and adjust.
Hours per quote becomes 10–20 minutes.
Producing documents
Certificates, test reports, compliance packs, grant reports
- Today
- The same content is rewritten into a different template for every customer and every programme.
- After
- Data goes in; each template comes out filled. People check it and sign it.
More customers, same paperwork load.
Taking in and answering
Phone and chat enquiries, service requests
- Today
- Notes are taken by hand, and on busy days they are not taken at all.
- After
- Enquiries are logged, classified and routed automatically. A repeat caller arrives with their history attached.
Nothing gets dropped, and the history becomes company property.
Watching and collecting
Tenders, public notices, price movements
- Today
- Someone checks a few websites whenever they get a spare minute.
- After
- Anything matching your criteria is collected overnight and waiting, sorted, in the morning.
Nobody has to watch, and nothing is missed.
Keeping to deadlines
Statutory deadlines, contract renewals, scheduled inspections
- Today
- A calendar, and somebody's memory.
- After
- Deadlines are computed from the issue date on the document itself, and alerts arrive as they approach.
The ones that cost you a fine get kept first.
AI Agent, Trusted
Agents do the work. Engineering keeps them from being wrong.
We build AI agents that carry a task through on their own — intake, lookup, draft, approval request. And anyone can build a demo: making it not be wrong in daily operations is the engineering, and that is where our time goes.
They read and understand
Request forms in every layout, handwritten notes, drawings that arrive as a phone photo. It does not have to be a fixed template.
They judge, not just extract
“Is this enquiry a repair or a sales lead?” “Which line of the price list is this item?” The AI makes the first call; a person confirms it.
They carry several steps
Nobody presses a button between stages — the agent continues. What it is doing right now is always on screen. We do not ship black boxes.
Every output carries its evidence
Any value the AI produced links back to where it came from in the source document. If it is not in the source, it does not invent one — it says so.
Low confidence goes to a person
When a case is ambiguous, the AI does not decide it. It lands in a person's review queue.
Money and documents need human approval
Sending a quote, placing an order, issuing a certificate — anything that leaves the building is confirmed by a person.
We test on your real data first
Before anything goes live we run your actual documents through it, show you the accuracy, and start from there.
Where AI isn't needed, we don't use it. Plain calculation and transfer go to ordinary automation — faster, cheaper, and it does not get things wrong. AI goes only where something has to be read, understood and judged.
Numbers are computed by code, not by a model. The AI reads the messy source; verified formulas compute the amounts and quantities. That is why the arithmetic holds.
Where we come from
Built out of an AI consultancy, for the companies it couldn't fit
FloatAI is a subsidiary of MoatAI, an AI strategy consultancy that has designed and built AI transformation for large organisations — asset managers, securities firms, industrial engineering groups.
That work made one thing obvious. AI that removes repetitive work does not care how big a company is, but the way a large organisation adopts it does not transfer to a company of forty people. FloatAI exists to rebuild the same technical base and the same methods at a speed and a cost a mid-sized company can carry.
Large-enterprise AI transformation stays with the parent. Working-level automation for small and mid-sized companies is ours.
A Korean asset manager — automating operational work
Department-wide demand surveys, training sessions and interviews with the people doing the work selected the tasks. Classification and summarisation of market research, and first drafts of reports, were automated and validated in a pilot. FloatAI's core team worked on requirements and delivery management at the time.
01
Korea Investment Management
02
Hanwha Securities
03
Samsung Securities
04
NH Investment & Securities
05
Eugene Investment & Securities
06
IBK Investment & Securities
07
OKU System
08
Nubicom
＋ undisclosed clients
MoatAI has worked on real problems with companies across several industries — from current clients to firms it built products and services with in the past. FloatAI starts on top of that experience.
How we work
Start small. Finish with your team running it.
01
Diagnose
Discover
We find the automation opportunities through on-site interviews and your actual working data. We don't only look at the scope you asked about — if work that is already automated in another industry applies to you, we say so.
You get
- A list of the repetitive work
- Impact × difficulty priority table
- A fixed-price quotation
02
Design
Design
Highest impact first, designed small.
You get
- A prototype you click through on your own real work
03
Build
Deploy
We test alongside the people who do the work, and attach it to the real process.
You get
- Something that works, every week
04
Hand over
Drive
We hand it over so your own staff can run it and change it.
You get
- Operating documentation
- Training for your owner of the system
- A maintenance agreement, if you want one
Stage by stage
We confirm the result of each stage with you, then decide the next one together.
Careful with your data
Your information is handled only within the scope we agree on.
No lock-in
It keeps running without us. You are not tied to us.
Working together
The three conversations we usually have
We are a small company and we say so. These are the situations where talking to us is worth your time.
You run the company
There is paperwork that repeats, and a team too small to absorb it. We start by looking at the actual documents — and if automation is not worth it for you, that is what we will tell you.
You serve companies like that
Consultancies, system vendors, industry and trade associations. You hold the customer relationship and the domain knowledge; we build the automation layer underneath it.
You operate across a border
Head office in one country, production or sales offices in another. The same document gets entered twice, once on each side. We design for the whole loop, not one office.
FAQ
Before you write to us
Yes. We work in Korean and English, and the case we are built for is a company whose plant or sales office sits in another country — where the same document is entered on both sides of the border. Tell us your setup and we will say plainly whether we are the right fit.
No. Most of the companies we work with have none. We build it; then we hand over how to run it and how to change it.
No. We leave them where they are and connect to them first. If a system is closed off, we work around it — an upload format, a watched mailbox, a shared folder.
FloatAI is a subsidiary of MoatAI. The parent handles AI transformation for large enterprises and financial institutions; we handle working-level automation for small and mid-sized companies. We share the technical base and the experience — but on a FloatAI project, the contract, the delivery and the result are FloatAI's responsibility.
Which pile of paperwork hurts most?
Tell us what you're looking at
Write in English or Korean. An engineer who has sat in that office reads it — not a sales team. If it is useful, we will set up a 30-minute call and pick out three tasks worth automating, at no charge.
If automation is not worth it for your company, we will say so.