AI-assisted testing

Advanced AI-Assisted Coding for Developers

Thalia Barrera

AI Engineering Curriculum Manager, DataCamp

Continuing with Atlas

A software developer reviewing code with security shield

Integrating Atlas into Wayfarer Labs
✅ Must ensure code is tested, secure and resilient
🤖 AI accelerates evaluation and testing of unfamiliar code

Why AI-assisted testing?

Uncover untested paths
Find edge cases
Detect potential vulnerabilities

🤖 Start by asking AI to assess testing maturity of the codebase

A magnifying glass examining code with bug icons

Assessing testing maturity

You are a senior QA engineer and software architect.

Assess how well tested this codebase is and identify the highest-risk gaps. Score the following on a 0-5 scale:

Coverage of critical paths

Edge-case handling

Regression protection

CI/CD automation

In addition, identify the test types currently present: unit, integration, end-to-end, regression, property or fuzz tests.

Testing maturity results

Establishing a baseline

Analyzing coverage results

Testing strategies

Testing strategies diagram

✔ Exploratory → understand behavior
✔ Functional → verify outputs
✔ Regression → protect existing behavior
✔Automated → fast, repeatable feedback

Building a test suite

What is a test suite?

👉 A structured collection of tests, data, tools, and rules that ensure system behavior remains consistent over time.

🤖 Test suite prompt:

You are a senior software engineer. Generate unit and integration tests for the critical paths in this codebase. Include:

Normal cases

Edge cases

Failure scenarios

Clear assertions and fixtures

Also provide a list of all generated tests and a one-sentence description of each.

Verifying coverage improvement

List of AI-generated tests:

New code coverage:

Runtime protections

Injecting custom data as input ➡ Atlas ❗

🤖 Runtime protection prompt:

When injecting or handling custom data, add runtime protections:

Validate inputs before processing (types, required fields, unexpected values)

Check schemas at every pipeline boundary, never assume upstream data is clean

Return clear error messages that identify what failed and where

Apply safe defaults for optional fields rather than leaving them undefined

CI automation

Let's practice!

Advanced AI-Assisted Coding for Developers