Recruiter
“The output alone does not reveal how well they use AI.”
The result is visible.
Their decisions and verification process are not.
Find and assess AI-native talent that delivers 10x.
Assess how candidates work with AI through practical tasks and process analysis. From talent discovery to skills assessment, get the evidence your team needs to make hiring decisions.
Analyzing the process of
solving practical tasks
Assess AI skills across technical and non-technical roles with job-relevant tasks.
We analyze outputs alongside conversations, tool use and verification records.
A practical development environment with code, documentation and tests.
Candidates complete practical tasks for non-technical roles using documents, AI tools and business applications.
The output alone does not show how well someone works with AI.
Hiring teams need to understand the person’s decisions and verification process behind the work.
Technical reviewer
“We need evidence of what the AI did and what the candidate contributed.”
We need to see how they framed the problem
and revised the AI’s work.
Team lead
“Can they deliver results in our working environment?”
They need to solve problems within constraints
and explain the result.
We found the capabilities behind effective AI use
in practitioners’ experience.
Together with practitioners using AI in their daily work across different roles,
we explored the capabilities that drive results.
“Build with AI while preserving
system rules and safety.”
Developer’s perspective“Before asking AI for answers,
define the problem to solve.”
Planner’s perspective“In AI-generated conclusions,
distinguish evidence from speculation.”
Analyst’s perspective“Separate work to automate with AI
from work that needs human judgment.”
Accounting’s perspective“Evaluate AI suggestions in the context
of our organization and team.”
HR’s perspective“Expand options with AI,
and set criteria for choosing.”
Strategy’s perspectiveWe organized these insights into three assessment dimensions.
Technical
Can they use AI to build something that works?
- Execute and verify the output
- Build maintainable structures
- Manage security and sensitive data
Intent
Can they steer AI toward the goal?
- Define goals and constraints
- Break down and delegate work
- Adjust direction and priorities
Cognition
Do they understand how it works and where it fails?
- Explain structural decisions
- Understand how the output works
- Anticipate limits and failures
In the AI era, talent creates different outcomes from the same conditions.
The difference is how people put AI to work.
- Delegation without direction
- Repeated revisions
목표
- Clear goals
- Partial verification
목표
- Structure the work first
- Verify and adjust direction
구조화검증목표
Assessment grounded in actual work.
We review the work, code, AI collaboration records and follow-up answers. Each competency is scored against defined criteria, with supporting evidence in the report.
Good
~/project/auth
~/project/auth $ codex
›_ OpenAI Codex
directory: ~/project/auth
/help for available commands
Explain the login failure and scope of the fix first.
Fix it while keeping the existing authentication and API.
Test success and failure paths, then report the results.
Fix it while keeping the existing authentication and API.
Test success and failure paths, then report the results.
- Clarifies cause and scope
- States constraints to preserve
- Checks tests and results
Bad
~/project/auth
~/project/auth $ codex
›_ OpenAI Codex
directory: ~/project/auth
/help for available commands
Login is broken. Handle everything else yourself.
Change authentication however you prefer.
Skip testing and just tell me it is done.
Change authentication however you prefer.
Skip testing and just tell me it is done.
- Delegates without a clear scope
- Allows changes without constraints
- Accepts completion without testing
From practical tasks to interview support.
Assess the work and the process, then carry the evidence into interviews.
- 01
Start in a real environment
Start in a browser IDE, without installation. Use real APIs and compute within set time and token limits.
- 02
Artifact & process, auto-collected
Automatically collect outputs, prompts, tool settings, AI conversations and verification records.
- 03
Post-task Q&A
Explain the work and the decisions behind it. Answers are checked against the actual implementation.
- 04
Assessment report · interview support
Receive competency scores and supporting evidence. Candidate-specific questions and follow-ups support the interview.
Cofa Use Cases
From internal hiring to a joint hackathon, assessing skills in practice.
01 / INTERNAL HIRING
Developed within KRAFTON. Now used in internal hiring.
Read the story →Cofa, developed and validated in-house, applied to KRAFTON hiring.
Built for our own hiring needs
Expanded into internal hiring
02 / JOINT HACKATHON
Cofathon
AI Native Battlegrounds
Read the story →An AI Native hackathon by KRAFTON and CJ Olive Young.
[Cofathon official page ↗]A joint AI Native hackathon
Cofa for tasks and capability assessment
Frequently asked questions
Do you only assess tool proficiency?
We also assess how candidates define problems, delegate to AI, verify results and explain their work.
What evidence is assessed?
Artifacts, source code, prompts, work logs and follow-up answers support assessment across technique, intent and cognition.
Do candidates use different AI environments?
Candidates receive the same task, workspace and model so we can compare how they solve problems under shared conditions.
Does it support interviews after assessment?
Report evidence and candidate-specific follow-up questions help interviewers prepare.
What should I prepare for an inquiry?
Share your assessment goal, target roles, expected candidate volume and timeline. We will discuss tasks and scope with your team.