When a internal team says "we need an AI tool," they usually describe a symptom, not the core issue. Your goal as an AI Engineer here is to dig deeper.
Diagnose before building: Ask the essential questions that prevent building the wrong tool; Draw clear boundaries: Know exactly what AI can evaluate reliably versus where human discretion is indispensable; Design for adoption: Create workflows that guide users through diagnostic steps rather than replacing their core judgment.
If you treat internal AI tooling as a real product with real human constraints, skip the cover letter and demonstrate your approach in the challenge below
The Brief
Context
This assessment evaluates your ability to diagnose operational bottlenecks, define AI/human boundaries, ask sharp discovery questions, and design practical AI-assisted workflows.
Below are two distinct business cases representing different operational challenges. Read both cases, then answer the questions at the bottom. Your answers for Questions 1–3 should address both cases (either by comparing them or answering per case), while Question 4 asks for a workflow design tailored to Case 1.
Case 1: Automated Candidate Screening & Prioritization
Context: The HR team receives applicant CVs via email for various roles (e.g., Account Manager, AI Automation Specialist). Currently, a screener manually reads each CV against the Job Description (JD) to make a pass/fail decision.
Because candidate volume is rising, this manual process has become a bottleneck. Additionally, different screeners often score the same CV differently. Leadership wants an automated system to help collect CVs, score them, and tag them with priority status flags (🔴 Low Match / 🟡 Potential Match / ✅ High Match) so recruiters know which candidates to review first.
Case 2: Content Writer Tool
Our Content team uses a structured approach to create educational materials:
Identify the target audience.
Pinpoint the learning gap (what the reader lacks).
Classify the failure type (why the reader struggles).
Select the matching design principles and templates.
Write the outline and draft.
We have an internal knowledge base containing these frameworks, taxonomies, and templates. However, writers frequently skip steps 1 - 4 and jump straight to writing outlines. This leads to generic content that fails to solve the reader's underlying issue.
The Task
Please answer the following questions:
01. Problem Diagnosis & Stakeholder Discovery
- Root Cause vs. Symptoms: For both Case 1 and Case 2, identify the core root problem versus the visible symptoms.
- Discovery Questions: List at least 3 essential clarifying questions you would ask stakeholders for each case before designing or building a solution.
02. AI Capabilities & Human-in-the-Loop Boundaries
- Limits of AI: Across both cases, which specific evaluation or execution tasks can AI handle reliably, and which tasks must remain under human judgment? Explain your reasoning.
- Critical Step Identification: In Case 2 specifically, which of the 5 steps in the writing methodology is hardest for AI to automate or evaluate? Why?
03. System & Interaction Design Strategy (Case 2 focus)
- Co-pilot vs. Automation: In Case 2, would you design the AI to automatically assign the failure type for the writer, or guide the writer step-by-step to diagnose it themselves? Justify your choice.
04. Workflow Architecture (Case 1 Focus)
- Workflow Design: Sketch a simple step-by-step workflow (Input $$\rightarro$$ Processing Steps $$\rightarro$$ Output) for the CV screening challenge in Case 1. Clearly highlight where human review, quality gates, and final decision-making occur.
The Rules
Submit Your Challenge
Based in Vietnam.
Collaborating worldwide.
We like people who can stay sharp under pressure, handle revision, and move the work forward without chaos.
