🔩 10-AI-Powered Document Intake

Turning messy PDFs, screenshots, and spreadsheets into structured, auditable data — in seconds.

✍️ Written from Riyadh — for founders, product teams, and AI builders in regulated markets.

---

🎧 Listen to this Article

---

📂 From Raw Docs to Insights

Why the hardest part of lending AI isn’t scoring — it’s document chaos.

---

Walk into any credit team and you’ll see the real bottleneck:

1. Borrower uploads a hodge-podge of PDFs, JPEGs, and Excel files.

2. Someone downloads them, renames them, and drags them into folders.

3. Another person re-types numbers into a spreadsheet “just to be safe.”

4. Weeks later, underwriting finally gets “clean” data — that might still be wrong.

We built Qararak’s Document Intake Engine to kill that workflow.

---

🏗️ 1. Ingestion Pipeline Built for Real-World Mess

Accepted formats: PDFs, images, zip bundles, E-statements, even WhatsApp screenshots.

Upload channels: Web portal, API, SFTP, mobile capture.

Every file is stamped, versioned, and sent straight to an on-prem object store — no local downloads, no email chains.

---

🧠 2. AI-Driven Classification (Multilingual & Domain-Specific)

Standard OCR alone isn’t enough, especially when documents mix Arabic and English.

We combine:

• Tesseract + EasyOCR for baseline extraction

• LLM-powered layout parsing to detect tables, stamps, and handwritten notes

• Custom CNN classifier trained on 40+ Saudi financial templates (bank statements, ZATCA tax returns, MOF certificates)

Outcome: “This is a 2023 audited balance sheet (Arabic), 6 pages.” — with 98 % precision.

---

📊 3. Smart Validation Rules — Not Manual Checklists

Once a doc type is confirmed, we fire validation rules in real time:

Example Rule Logic Result Date range check Transaction dates within last 12 m? ✅ / ❌ Completeness All mandatory columns present? ✅ / ❌ Math integrity Assets = Liabilities + Equity? ✅ / ❌

Rules are decision-table driven — business users can add or edit without code.

---

🔁 4. Feedback Loop to the Borrower (or RM)

If a document fails validation:

1. Qararak generates a reason code (“Missing VAT certificate”).

2. Sends an API callback / email template.

3. Borrower re-uploads only what’s missing.

No phone calls. No guesswork.

---

🔐 5. Compliance & Audit Trail

• SHA-256 hash stored for every file version.

• Linked to the borrower’s master record.

• Full changelog: who viewed, validated, or rejected a doc — and why.

• All data stays on-prem (PDPL-aligned), with optional encryption at rest.

---

⚡ 6. What This Unlocks

Pain Point Yesterday With Qararak Today Manual renaming & sorting Auto-classification & routing Spreadsheet re-typing Structured JSON payloads Version confusion Immutable hash & timestamp Week-long doc QA Sub-minute AI validation

Speed goes up, errors go down — and underwriting finally gets clean, trusted data.

---

🧭 Final Thought

Great credit decisions start long before a model runs.

They start the moment a borrower drags a PDF into your portal.

By turning raw documents into validated, structured insights — instantly and securely — Qararak frees your team to focus on what matters: risk, not re-typing.

Next Article

🔩 How We Build AI Differently | Build on Top of Us: Extending Qararak via APIs

---

---

🎧 Explore More

→ Listen to the 🤖AI on the Ground Podcast: Real-world AI powering compliance, credit, and regulated markets in Saudi — decoded for operators.