CCAR-P Advanced Agentic Architecture Practice Question
A document-processing agent must extract structured fields from thousands of PDFs. Some PDFs are scanned images, some are native text, and some are encrypted. The architect wants one pipeline that routes each document to the appropriate extractor and reports per-document confidence. Which design best fits?
⚠ Common exam trap
The trap here is treating PDF processing as a single uniform step, when routing by document structure is what makes extraction both accurate and cost-effective across mixed inputs.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Pre-classify each PDF by inspecting its structure, route native-text documents to a text extractor, scanned images to OCR, and encrypted files to a decryption step, then validate extracted fields and attach confidence scores.
The three document classes need different handling, so the pipeline should inspect each file and route it: direct text extraction for native PDFs, OCR for scans, and a decryption stage for encrypted files. After extraction, validating fields and attaching confidence scores gives the per-document reporting the architect requires. This approach uses the cheapest viable method for each class and keeps confidence meaningful, rather than forcing one uniform method that is wrong for at least one class.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Pre-classify each PDF by inspecting its structure, route native-text documents to a text extractor, scanned images to OCR, and encrypted files to a decryption step, then validate extracted fields and attach confidence scores.
Why this is correct
Inspecting structure first lets each document take the cheapest and most accurate path: text extraction for native PDFs, OCR for scans, and a decryption stage for encrypted files. Validation and confidence scoring after extraction give the per-document reporting the architect wants. This matches processing to document type instead of forcing one method on all three, improving both cost and accuracy.
- ✗
Reject any PDF that is not native text and return an error, since scanned and encrypted documents are out of scope for automated processing.
Why it's wrong here
Rejecting scanned and encrypted documents leaves a large share of the corpus unprocessed, which defeats the goal of handling thousands of PDFs. The requirement explicitly includes scanned images and encrypted files, so exclusion is not an option. This avoids the routing problem by shrinking the workload rather than solving it, and provides no confidence reporting for the documents it does accept.
- ✗
Convert every PDF to images first, then run OCR on all of them, and skip classification because OCR handles any document.
Why it's wrong here
Rasterizing native-text PDFs discards embedded text and introduces OCR errors that would not occur with direct text extraction. Encrypted files still cannot be rasterized without decryption, so that class is unhandled. Uniform OCR also loses structure such as table boundaries that native extraction can preserve, and per-document confidence becomes less meaningful because all documents share the same error profile.
- ✗
Send every PDF to a single vision-language model and ask it to return the structured fields directly, ignoring document type.
Why it's wrong here
A single vision path wastes cost on native-text PDFs where text extraction is cheaper and more reliable, and it offers no special handling for encrypted files, which will fail regardless of model capability. It also provides no explicit routing or tiered confidence, so per-document reporting is coarse. Uniform processing ignores the meaningful differences between the three document classes.
About these practice questions
Courseiva writes every CCAR-P question from scratch — 262 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCAR-P practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCAR-P exam.