Beyond the AI Buzzwords
Understanding what makes Veritas Lab different in the document intelligence landscape
Most "AI" solutions today, especially in document intelligence, are really machine learning (ML) pipelines, not general AI. Let's break this down clearly so you understand exactly what Veritas is doing, and where it falls on the AI–ML spectrum.
AI vs ML in Veritas
| Concept | Description | What Veritas Uses |
|---|---|---|
Artificial Intelligence (AI) | Broad field aiming for machines to simulate human-level intelligence (reasoning, creativity, understanding context, etc.) | Not quite — no general reasoning yet |
Machine Learning (ML) | Algorithms that learn patterns from data to make predictions or decisions without explicit programming | Yes, your Veritas pipeline uses this heavily |
Deep Learning (DL) | Subset of ML using neural networks (like BERT, GPT, etc.) | Yes, used in document classification / NER |
Rule-based Logic | If-then logic or pattern rules without learning | Also used — especially in verification |
What Is Veritas Actually Doing?
Veritas is a hybrid document intelligence pipeline made of:
Preprocessing (non-ML):
- OCR via Amazon Textract (no learning, just computer vision)
- Basic layout/geometry parsing
ML-based NLP (Intelligent Parsing):
- NER: using models like spaCy, HuggingFace to extract entities (names, dates, clauses)
- Document Classification: trained models (e.g., BERT, XGBoost) that say "this is a contract" or "invoice"
- Clause Matching: embeddings + similarity (e.g. "is this paragraph similar to known clause X?")
Rule-based Verification (non-ML or shallow ML):
- Does the document contain expected fields?
- Do risk flags exist (e.g., missing clause)?
- Is there a valid signature?
Proof Overlay (UX enhancement, not AI):
- Geometry mapping for bounding box highlights
- Provides audit transparency, not learning
What's NOT AI (But Often Marketed As)
| Feature | Truth |
|---|---|
| OCR with Textract | Not AI — it's deterministic CV |
| Regex extraction | Not AI — rule-based parsing |
| Bounding box overlay | Not AI — frontend rendering |
| File classification by keyword | Not AI — deterministic tagging |
Where Veritas Could Become More AI
Here's how we can evolve from a smart ML pipeline to a more "AI-like" system:
Context-aware parsing
Use large language models (LLMs) like Claude or GPT to understand nuance ("This is a subcontract clause")
Conversational interface
Let users ask questions about documents: "What is the risk in this NDA?"
Auto-summarization
Provide high-level natural language summaries of documents
Memory + adaptation
Model remembers past patterns per client ("PwC uses a different clause template — adjust extraction accordingly")
So what is Veritas?
Veritas is a document intelligence engine powered by machine learning and enhanced with audit-grade logic and explainability.
It's not full general AI — but that's a good thing.
High trust
Explainable ML
Reliable outcomes
Low hallucination risk
Ready to see the difference?
Experience how our document intelligence platform can transform your workflow with the perfect balance of ML intelligence and explainable results.
Request a Demo