✓ Authorized genuine licensing in Vietnam Hotline: 0866 398 102 contact@foxitpdf.vn

Automating PDF Workflows with Foxit: Batch OCR & AI Agents

A practical guide to PDF automation with Foxit: batch OCR and conversion, Action Wizard, PDF JavaScript, AI Assistant and document AI agents built with LLMs.

If your accounting team scans hundreds of invoices a month, your legal department keeps contracts as image-only scans, or your back office still retypes data from PDFs into business software, this guide is for you. Drawing on our experience deploying Foxit for enterprise customers in Vietnam, we walk through the full automation ladder: batch OCR and conversion inside Foxit PDF Editor, Action Wizard, JavaScript in PDFs, the AI Assistant, and finally document AI agents built on Foxit’s developer APIs — with honest notes on where each tool stops.

Step 1: Batch OCR — the foundation of every PDF workflow

Scanned PDFs are just pictures. Until OCR adds a text layer, nothing downstream — search, AI summarization, data extraction — can work. Foxit PDF Editor handles this at scale:

  1. Open the Convert tab and choose OCR → Multiple Files (for a single open document, use Recognize Text → Current File).
  2. Click Add Files / Add Folder / Add Open Files — adding a whole folder is what makes this a batch tool.
  3. Set the output folder and file-naming rules.
  4. Pick the OCR language (Foxit’s engine supports Vietnamese with full diacritics alongside English and many other languages; you can select several for bilingual documents).
  5. Choose the output type:
    • Searchable Image Text keeps the original page image and adds an invisible text layer underneath — our recommendation for legal records, since the document’s appearance never changes.
    • Editable Text converts recognized content into editable text; combine it with Find All Suspects to review words the engine was unsure about.

A field note from a logistics customer in Ho Chi Minh City: at 300 dpi, Vietnamese OCR accuracy was good enough for reliable full-text search across their archive. Below 200 dpi, rescanning the originals was faster than fixing OCR output.

Step 2: Action Wizard and batch conversion — automation without code

Beyond OCR, Foxit PDF Editor converts files to and from Office formats in bulk, and the Action Wizard (a Pro/Editor+ feature) lets you record a sequence of commands — for example OCR → apply a watermark → password-protect → save to an archive folder — and run it against any folder of files, in order, with no scripting. Actions can pause for user input at steps that need judgment, and support custom commands for advanced cases.

One pattern we configure often for administrative teams: staff drop the week’s scanned correspondence into a folder, then run a single action that OCRs everything in Vietnamese, stamps a reference footer, and saves the output to SharePoint.

JavaScript inside PDFs

Foxit PDF Editor also supports JavaScript for PDF, similar to Adobe Acrobat, under Form → JavaScript: Document JavaScript (document-level functions stored in the file), Document Action (event-driven scripts such as Document Did Save), a JavaScript Console for testing, and Edit All JavaScripts — with optional external editors like Sublime Text or WebStorm. The classic enterprise use case is self-calculating forms:

// Document JavaScript: auto-total a payment request form
function computeTotal() {
  var total = 0;
  for (var i = 1; i <= 10; i++) {
    var f = this.getField("line" + i + ".amount");
    if (f && f.value) total += Number(f.value);
  }
  this.getField("grandTotal").value = total;
}

PDF JavaScript is the right tool for logic inside one document (forms, validation, navigation) — not for processing thousands of files. For that, read on.

Step 3: Foxit AI Assistant — reading, summarizing, Q&A

Foxit AI Assistant works across Foxit PDF Reader, PDF Editor and Editor+ on a credit model: every Foxit account gets 300 free AI credits per month, and the AI Assistant subscription raises that to 2,000 credits per month, poolable across a team via the Foxit Admin Console. Verified feature set (names as Foxit documents them): Chat with your PDF (natural-language Q&A over one or several documents, including images), Summarize, Translate (30+ languages, side-by-side view on desktop), Read Aloud, Smart Podcast, AI Bookmark, and — on Editor/Editor+ only — Edit by chat (redact sensitive data, reorder pages, apply watermarks by typing an instruction) and PDF Action Inspector, which scans for embedded JavaScript and data that survives redaction.

AI featurePDF Reader (free)PDF EditorPDF Editor+
Chat, Summarize, Translate, Read Aloud✔✔✔
Local AI mode (Windows/macOS)✔✔✔
Edit by chat—✔✔
PDF Action Inspector—✔✔
AI Bookmark, Generate Image—✔✔

The prompt pattern that works best for contract review, in our daily use: ask for a structured summary — parties, scope, payment terms, termination, penalties — “citing the clause number for each item”. That last instruction turns the summary into a verifiable index instead of a paragraph you must take on faith.

Privacy note: for sensitive documents, Foxit’s Local AI mode (via Microsoft Foundry Local) runs inference, processing and storage entirely on-device, works offline, and consumes no credits. For cloud processing, review Foxit’s Privacy Policy and AI User Guidelines first.

Step 4: Document AI agents — OCR plus an LLM

The highest rung of the ladder is an AI agent: a pipeline that ingests files, OCRs them, extracts structured data with a large language model, validates the result, and writes it into your business system — escalating to a human only on exceptions. Foxit supplies the building blocks:

  • Foxit PDF SDK (desktop/server) with an OCR add-on that can run as a command-line tool for scheduled, headless batch pipelines.
  • Foxit PDF Services API — REST endpoints for conversion and text extraction that can be chained (one job’s output feeding the next), plus Foxit’s Document AI layer, which extracts tables, key-value pairs and form fields, and isolates specific values such as invoice numbers or contract dates based on your prompt — including from scans and photos via built-in OCR.
  • Foxit eSign API — a REST e-signature service to close the loop on contracts: OAuth 2.0 authentication, a createfolder call that defines signers and returns an embedded signing session for an iframe in your own app, and a folder_executed webhook when the document is final.

A reference architecture for invoice extraction

For a distribution customer processing roughly 3,000 supplier invoices a month, we built four layers: ingest & OCR → text/layout normalization → LLM extraction against a fixed JSON schema → code-based validation (tax-ID format checks, line totals vs. grand total, vendor-master matching). Clean records post automatically; anything flagged goes to a human review queue. Two prompt rules proved decisive: fields not found must be null — never guessed, and the model must grade its own confidence using arithmetic cross-checks, which drives the routing.

Practical accuracy rules we hold to:

  • Benchmark on 100–200 ground-truth documents before go-live; fully automate only fields that hit 98–99%.
  • Validate with ordinary code, not AI — totals, ID formats and dates are cheap to check deterministically.
  • Keep a human in the loop: the goal is removing 70–90% of manual effort, not 100%.
  • Log everything (source file, OCR text, extracted JSON, reviewer) for audit.

For contract repositories the better pattern is RAG (retrieval-augmented generation): OCR the archive once, chunk by clause, store embeddings in a vector database, and answer questions with citations back to the source clause. At small scale, AI Assistant’s multi-document chat covers this without custom infrastructure.

Choosing your level of automation

NeedBest fit
Searchable archive, occasional batchesFoxit PDF Editor (batch OCR/convert)
Recurring multi-step document routinesEditor+ / Pro with Action Wizard
Summaries, Q&A, translation on demandFoxit AI Assistant
Scheduled server-side OCR, ERP integrationFoxit PDF SDK / PDF Services API
Automated signing inside your own appFoxit eSign API

Our rule of thumb from deployments: start with PDF Editor, and move to SDK/API only when volume outgrows human operation — typically beyond 1,000–2,000 documents per month or when real-time processing is required.

FAQ

Can Foxit OCR handle Vietnamese scanned documents?

Yes. The OCR engine supports Vietnamese with diacritics; select it (optionally alongside English) in the Select OCR Engine dialog. Scan at 300 dpi or better for the best results.

Does Foxit PDF Editor support JavaScript like Adobe Acrobat?

Yes — Document JavaScript, Document Actions, a JavaScript Console and a full script editor under Form → JavaScript. Most standard form-calculation scripts port across; Acrobat-specific APIs should be retested in Foxit’s console.

Is Foxit AI Assistant safe for confidential files?

Use Local AI mode on Windows/macOS: Foxit states processing stays entirely on your device and works offline. For cloud mode, check Foxit’s AI User Guidelines against your internal security policy.

When do I need the Foxit PDF SDK instead of the desktop editor?

When documents must be processed without a person at the keyboard — scheduled server jobs, ERP/accounting integration, or volumes in the thousands per month. The SDK’s OCR tool and the PDF Services REST API are designed for exactly that.

Talk to us

FoxitPDF.vn supplies genuine Foxit licenses in Vietnam and helps teams design OCR and extraction workflows end to end. See our pricing or contact the technical team — hotline 0866 398 102 — for a demo on your own documents. Vietnamese readers can also see our in-depth guides on batch PDF automation and invoice data extraction with AI agents.

Compiled from manufacturer documentation linked below. Examples and code require testing on copies of your own data before production use.

References:

← All news
Call us 0866 398 102 Zalo Chat on Zalo Send email