AIMentor — AI Expertise Learning Platform

13 questions · pass mark 80% · 20 minute limit. The timer auto-submits at zero.

Time remaining: 20:00

Question 1easy
Why does connecting AI to business tools (CRM, helpdesk, Slack) matter?
It makes the model biggerAI delivers value inside the systems where work actually happens, not as a standalone demoIt reduces token costIt is required by the EU AI Act
Question 2medium
What is usually the authentication method for business-tool APIs, and its production gotcha?
A simple static API key that never expiresOAuth — you must refresh tokens, and wrong-account consent is a common production failureNo auth is neededA password in the URL
Question 3medium
What is the real demo-to-production gap for business-tool connectors?
The AI model qualityAuth (OAuth refresh, correct-account consent) and per-tool rate limitsThe faviconThe prompt length
Question 4medium
What is the core loop of a business-tool AI connector?
Train → evaluate → deployRead context from the system → run AI → write the result back to the system's fieldsEmbed → retrieve → generateChunk → index → search
Question 5hard
The enrich-on-event pattern combines which Track 9 primitives?
Only a REST callWebhook (verify + ack fast) → queue → idempotent worker with retries → write-back to the source systemOnly a message queueFine-tuning plus embeddings
Question 6medium
When mapping AI output to a CRM, whose schema must you write to?
Your own preferred schemaThe tool's schema — its fields, picklist values, required fields, and formatsNo schema is neededThe embedding model's schema
Question 7hard
How should you handle an AI category that has no matching value in the tool's picklist?
Let the write failMap to a safe fallback value (e.g. 'Other') rather than sending an invalid value that rejects the writeDelete the recordRetry forever
Question 8hard
How do you make writes to a business tool idempotent to survive retries?
Write fasterUse the tool's upsert / external-id features so a retried write updates the record rather than duplicating itDisable retriesUse a bigger model
Question 9medium
For a standard 'HubSpot deal Won → post Slack summary' at low volume, which approach fits?
A custom Kubernetes microserviceiPaaS / no-code (n8n, Zapier, Make) — the connectors already exist and it's non-engineer maintainableFine-tuning a modelA native SDK rewrite of both connectors
Question 10hard
For a high-volume Salesforce enrichment with complex logic and private-KB RAG, what's usually best?
Pure no-code, forcing the RAG into a Zapier stepHybrid — an iPaaS/connector owns the Salesforce auth/trigger; a deployed code service does the complex AI/RAGA spreadsheetRebuilding the Salesforce connector from scratch in code
Question 11hard
Which are business-tool connector pitfalls? (Select all that apply)
Underestimating OAuth refresh and wrong-account consentIgnoring per-tool rate limits under loadMapping to your own schema instead of the tool's fields/required-fieldsUsing an iPaaS platform's maintained connector for a standard integration
Select all that apply.
Question 12medium
The pattern webhook → queue → idempotent worker → write-back to the source system is called enrich-on-____ (one word).
Question 13hard
What class of platform (acronym or 'no-code'/'iPaaS') already ships maintained connectors for standard tools like n8n/Zapier/Make?

0 / 13 answered

Business-Tool Connectors — Final Exam