A no-login, zero-setup web tool for Thai office teams. Drag invoices, receipts, contracts, and ID scans in — DocPlus reads each one, classifies it, pulls out the key fields, and suggests where it should be filed. Thai-first. Thai data residency available. Nothing to install.
Every back-office in Thailand has a filing person. Not officially — their job title says "accounts assistant" or "admin coordinator" or "HR officer" — but somewhere between their second and third year on the job, they quietly inherited the task of opening every PDF that lands in the shared inbox and deciding where it goes.
They receive between 200 and 800 PDFs a month. Invoices from vendors. Receipts attached to staff expense claims. Draft contracts forwarded by legal for countersignature. Government notices in Thai script. ID scans for KYC. Bank statements for reconciliation. Commercial invoices for shipping clearance.
They open each file. They decide what it is. They move it to the right folder. Often they rename it. Sometimes they copy an amount or a reference number into a tracking spreadsheet. Then they repeat. A ten-document morning takes forty-five minutes. A backlog after a public holiday takes a full day.
Enterprise DMS platforms expect a taxonomy workshop, a permissions matrix, an IT integration plan, and a rollout training programme. Six months before anyone sorts a document faster.
Off-the-shelf tools trained on English and Latin scripts stumble on Thai-language invoices, handwritten Thai receipts, and the Thai national ID card format. Accuracy under 70% means the tool is creating work, not saving it.
IT is already stretched. Nobody wants to issue 20 new logins, configure SSO, and train staff on another dashboard just so invoices get sorted faster.
"The person doing this work knows the classification in two seconds of looking at the document. They've just been asked to do it one file at a time, a thousand times a month."
Conversation with an AP lead at a Bangkok distributor · October 2025
DocPlus inverts the filing model. Instead of making the operator navigate a DMS taxonomy before they even know what's in front of them, DocPlus classifies the documents first and suggests where each one goes. The person confirms, corrects, or overrides. Filing happens in one action. There is no account. There is no project. There is no admin setup. The tool exists at a URL — the same way a calculator or a file-converter does.
The operator uses DocPlus in two seconds, not two days. No registration flow, no password, no team invite. Open the URL on any browser. Drag files in. See results. For shared-terminal use (front-desk, accounting workstation), a "Clear All" button resets the session for the next person.
DocPlus was trained on Thai-language invoices, Thai ID cards, and Thai-script government forms from the first line of code. Printed Thai text reads at ~98% character accuracy; handwritten Thai at ~91%. English is also supported. The tool doesn't treat Thai as a localisation concern — it treats it as the primary.
Default: Google Gemini (fastest, lowest cost, suitable for most commercial documents). Swap to NIPA Cloud AI for Thai data residency — required by banks, government agencies, and organisations handling regulated document types. Same interface, different backend. The operator doesn't see the swap.
This isn't a DMS. It's the sort-and-extract layer that lets humans move through paperwork at human speed.
Landing state. One page. One action.
A single page. No navigation. A large drop zone centered on the canvas. A subtle toggle in the top-right switches between Gemini (default) and NIPA Cloud AI (Thai data residency). A small help icon explains the toggle in one sentence. That's the entire landing UI. No login prompt. No "Get started." No marketing banner. The operator is already in.
Drag a single file or a hundred. PDFs, JPGs, PNGs, TIFFs.
Drag files individually or in bulk. Maximum 100MB per document; up to 50 documents per batch during pilot (horizontal scaling removes this limit in production). A progress bar appears per file as OCR runs — usually 2–5 seconds per document for text-dense PDFs, slightly longer for high-resolution scans. No format conversion required.
Each document becomes a card. Type, confidence, extracted fields, suggested folder.
Each uploaded document appears as a card in the results grid. Document type label, 0–100% confidence score (green above 90%, amber under 70%), extracted fields, a suggested folder path derived from your taxonomy, and a thumbnail preview. Click to open the full document in a side panel.
One-click accept. Two clicks to override. Type to correct a field.
Each card has three actions: Accept (green tick) confirms the classification and locks in the extracted fields; Override opens a dropdown of alternate document types and records the correction; Delete removes the document. Any extracted field can be edited inline. Click the value, type the correct one, tab to save. Corrections feed back into the model's fine-tuning pool (opt-in per-org).
Export to Excel or PDF. Or copy the filing path and move the files yourself.
Excel (XLSX) or CSV — one row per document, every field as a column. PDF report — printable, grouped by type, audit footer per session. Copy suggested filing paths — one button copies all paths as plain text for manual filing to SharePoint or a network share. DMS integration (SharePoint, Drive, OneDrive, S3, custom REST) is on the pilot-two roadmap.
One button resets the session. Nothing persists unless exported.
DocPlus is session-only by default. Documents are processed in memory, results are held for the duration of the browser session, and a single "Clear All" button wipes everything. The next person at the same terminal starts with a blank page. Session persistence is opt-in per deployment (for orgs that want per-user history across sessions). Session-only is the default because most office use is shared-terminal and the fastest privacy story is "we don't keep it."
The operator uses the tool in two seconds, not two days. No registration, no password, no team invite. Open the URL on any browser, drag files in, see results. Designed for the reality of shared-terminal office use.
Invoices, receipts, tax invoices, contracts, ID cards (Thai national ID front and back, MRZ for passports), bank statements, commercial invoices, purchase orders, HR forms, government notices, delivery orders, credit notes. New types added per-pilot.
Reads Thai-language text as the primary language, not a localisation afterthought. Extracted fields include vendor name (Thai script), tax-invoice number, issue date (Buddhist calendar auto-converted), VAT amount, total, names from Thai ID cards.
Every classification comes with a 0–100% confidence score so the operator skims the results and focuses review on low-confidence cards. Based on pilot data, ~85% of cards score above 90% and can be accepted in one click; the remaining 15% benefit from human review.
DocPlus suggests where each document should go — e.g., /AP/2026/March/Vendors/Acme-Corp/. Taxonomy is seeded during pilot week 1, refined over weeks 2–4 as the system watches which suggestions the operator accepts or overrides.
Default: Gemini (fastest, lowest cost, general commercial use). Swap to NIPA Cloud AI for Thai data residency, required by banks, government agencies, and regulated enterprise. Same UI, different backend. No operator retraining required.
One button wipes the session. Designed for shared-terminal use — front-desk KYC workstation, AP batching terminal, HR intake PC. The next person always starts fresh. Session persistence is opt-in per deployment.
Every classification is logged with model version, AI provider, confidence score, operator action (accept / override / delete), and timestamp. Sessions export as a CSV audit log per regulator request. Defensible for ISO 27001, SOX, Thai PDPA.
A single 50-person back-office that processes 1,500–3,000 documents a month recovers roughly 20–30 staff-hours per week when DocPlus replaces manual sorting. That's the headline. The detail is on how the work gets done, not just how long it takes.
| Metric | Before DocPlus | After DocPlus (8-week pilot) |
|---|---|---|
| Seconds per document (sort + file) | 180–300 | 10–20 |
| Misfiles per 1,000 documents | 30–60 | under 5 |
| Staff hours / week on sorting (50-person org) | 20–30 | 3–4 |
| First-time-right classification | 60–70% | 90–95% |
| Backlog clearance after a public holiday | 1–2 days | 1–2 hours |
| Audit trail completeness | 20–40% | 100% |
Figures derived from internal pilot benchmarks and our founder's prior OCR work on Thai and Myanmar-script documents (98.27% accuracy on printed text, 91.56% on handwritten characters). Individual pilot results vary by document mix, existing workflow, and team size.
In operational terms, a mid-sized back-office recovers the equivalent of one full-time salary's worth of sorting labour per quarter. In compliance terms, the audit trail closes a gap that typically costs hours of forensic reconstruction during regulatory review. In headcount terms, the filing-person goes back to doing the work their actual job title describes.
| AI providers | Google Gemini (default); NIPA Cloud AI (Thai data residency) |
|---|---|
| Supported formats | PDF, JPG, PNG, TIFF (OCR runs automatically before classification) |
| Languages | Thai (primary), English. Myanmar and Burmese on roadmap. |
| Max document size | 100MB per file |
| Max batch size | 50 documents per upload (MVP); horizontal scaling removes this limit post-pilot |
| Processing time | 2–5 seconds per typical document; 5–15 seconds for large high-resolution scans |
| Session model | Session-only by default; per-user persistence opt-in |
| Deployment | Web (hosted) or on-premises via Docker Compose |
| Stack | Laravel 13 + Livewire 4 (+ Volt) + Preline UI 4 + Tailwind CSS 3 + Alpine.js + PostgreSQL 18 |
| Infrastructure | Any modern VPS; minimum 2 vCPU / 4GB RAM for single-instance deployment |
| Scale | Single instance handles ~5,000 documents/day; horizontally scalable beyond that |
| Integrations (roadmap) | SharePoint, Google Drive, OneDrive, S3, custom REST API |
DocPlus is built to clear a Thai-enterprise procurement review without custom exceptions.
NIPA Cloud AI mode keeps all document processing and temporary storage within Thai infrastructure, satisfying Thai PDPA and the residency requirements of BOT-regulated financial institutions and Thai public-sector procurement guidelines.
By default, DocPlus holds processed documents and extracted fields in memory for the duration of a browser session. Closing the tab or pressing "Clear All" deletes all data. Optional persistence is available per-deployment.
Every classification and operator action is logged with timestamp, model version, AI provider, and outcome. Sessions export as CSV on request. Retention configurable from 0 days (default) to 7 years.
Because there is no login, there is no user-account database, no password hash store, and no PII beyond the documents themselves (which are session-scoped by default).
TLS 1.3 in transit; AES-256 at rest for any persisted data; encrypted file-level storage on NIPA Cloud AI infrastructure when that mode is selected.
Audit-trail structure is designed for standard compliance reporting. Formal certification is on the roadmap for production deployments; pilot deployments can operate within existing client ISO / SOX envelopes.
DocPlus deploys as a pilot with a written KPI and an honest ending. Either the measured outcome clears the KPI and we scale, or it doesn't and we part ways with the pilot report in your hands. No open-ended contracts.
Operational walkthrough with the team that currently does the sorting. Identify the top 3–5 document types that would deliver the most relief. Write the KPIs. Fixed pilot price confirmed.
DocPlus deployed to one site — one AP team, one HR desk, one KYC workstation. Real documents, real operators, real workflow. The taxonomy is seeded in week 3 and refined continuously through week 6. Weekly reviews are working-software demos, not slide decks.
Measurement against the written KPIs. Honest gap analysis. Recommendation: scale deployment, continue pilot with adjustments, or walk away. All three endings are valid.
If the pilot cleared the KPI: roll out to additional teams, deploy on-premises or to a dedicated NIPA tenant, enable SharePoint / Drive integration. If it didn't: we hand over the pilot findings and part ways.
Fixed-price pilot. Walk-away clause. Roadmap influence for design partners. No surprises.
Eight weeks, fixed price. Deploy to one site or one team. All costs confirmed in week 0.
Hosted multi-tenant. Priced by monthly document volume. For teams past pilot.
Annual licence, unlimited documents, on-premises deployment available.
In Gemini mode: documents are sent to Google's Gemini API. Google's standard API terms apply. Suitable for non-regulated commercial documents. Documents are not used for model training under Gemini's enterprise API.
In NIPA Cloud AI mode: all processing happens on NIPA Cloud AI infrastructure within Thailand. Documents never leave Thai borders. Suitable for regulated documents (banking, government, healthcare), Thai PDPA-compliant by design.
In either mode: DocPlus itself does not store your documents by default. Session-only processing. Closing the tab deletes everything.
The demo takes ninety seconds. No account, no commitment, no "our sales team will be in touch." If it works on your documents, you'll see it in the first thirty seconds. If it doesn't, the tool tells you (low confidence scores, missed fields) honestly.
DocPlus didn't start from scratch. The classification engine, the Thai-script handling, and the extraction accuracy are all built on our founder's prior enterprise work across Southeast Asia (2017–2026) — including proprietary OCR systems for Thai and Myanmar-script documents that achieved 98.27% accuracy on printed text and 91.56% on handwritten characters. Three million identity attributes have been processed through systems our founder has previously shipped.
That history is why DocPlus is Thai-first instead of Thai-retrofitted, and why the accuracy numbers on this page are calibrated to real-world pilot expectations rather than marketing-grade benchmarks.
Read the full track record →