CCAR-P Advanced Agentic Architecture Practice Question
You are architecting a customer-support agent for a SaaS platform. The agent must answer billing questions using a live invoice API and general policy questions using a static knowledge base. The invoice API is fast but occasionally returns stale data; the knowledge base is large and slow to search. Your design gives the agent a dedicated 'billing' sub-agent and a 'policy' sub-agent, coordinated by a router agent. After deployment, you observe the router sending nearly every query to both sub-agents in parallel, causing high latency. Which architectural change best addresses this while preserving answer quality?
⚠ Common exam trap
The trap here is treating 'router fans out too often' as a routing-logic bug to be replaced with deterministic rules, when the real fix is to expose the router's confidence so the orchestrator can gate fan-out.
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
✓
Have the router emit a structured routing decision with a confidence score, and only fan out to both sub-agents when the confidence falls below a configured threshold.
The router is being overly cautious and fanning out on every query, which multiplies latency. Adding an explicit confidence signal to the routing decision lets the orchestrator fan out only when the router is genuinely unsure. High-confidence routing keeps the common case single-hop, while the fallback preserves correctness on ambiguous queries. This is the only option that changes the router's decision policy rather than replacing it with a brittle rule or an unrelated caching layer.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Merge both sub-agents into a single agent with access to both the invoice API tool and the knowledge base search tool, and let it choose tools freely.
Why it's wrong here
Collapsing the sub-agents removes the separation that let you tune each retrieval path independently, and it reintroduces the original problem: a single agent with two tools will often call both to hedge. It also makes the stale-data problem harder to contain, because the billing tool's freshness caveat is no longer scoped to a dedicated sub-agent.
- ✓
Have the router emit a structured routing decision with a confidence score, and only fan out to both sub-agents when the confidence falls below a configured threshold.
Why this is correct
This preserves the router's semantic judgment while adding a confidence gate. High-confidence queries go to a single sub-agent, cutting latency; genuinely ambiguous queries still fan out, preserving answer quality. The threshold is tunable per environment, giving you an explicit latency/accuracy dial rather than an all-or-nothing routing rule.
- ✗
Cache the router's routing decisions for 24 hours so repeated identical queries skip the router on subsequent requests.
Why it's wrong here
Response caching helps only with repeated identical queries, which are rare in support traffic, and it does nothing about the first occurrence of each query. It also risks serving stale routing for queries whose intent shifts, such as a user following up on a prior billing thread. The observed problem is fan-out on novel queries, which caching does not address.
- ✗
Replace the router with a fixed decision tree that maps every query containing the word 'invoice' to the billing sub-agent and all other queries to the policy sub-agent.
Why it's wrong here
A fixed keyword decision tree is brittle: it misroutes paraphrases such as 'why was I charged twice' and cannot express confidence thresholds. It also removes the router's ability to escalate ambiguous queries to both sub-agents, so it may reduce latency but degrades answer quality on the very cases the router was designed for.
About these practice questions
One of 262 original CCAR-P practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. 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.