Blokken AIFIELD GUIDE
← Back to home
THE WORK, NOT THE HYPE

Software Engineering with AI.

Explore how AI changes your work — and what to learn next.

A practical guide to
the work ahead
Software EngineeringDebuggingGrowing Practice

Debugging

Analyze sanitized errors and propose possible causes.

The workflow, side by side

Traditional workflow

Before generative AI assistance
  1. 1

    Reproduce the reported issue

  2. 2

    Inspect logs and trace the relevant code

  3. 3

    Develop a root-cause hypothesis

  4. 4

    Modify the code

  5. 5

    Run regression and edge-case tests

  6. 6

    Review and approve the change

AI-assisted workflow

AI contributes. You guide and verify.
  1. 1

    Provide sanitized code and error context

  2. 2

    AI analyzes logs and proposes possible causes

  3. 3

    Engineer reproduces and validates the hypothesis

    AI + YOU
  4. 4

    AI suggests a candidate fix

  5. 5

    Engineer reviews the change and runs tests

    YOU
  6. 6

    Engineer approves and monitors the result

HUMAN CHECKPOINT

Does the proposed solution actually address the root cause?

Your judgment matters

The shift: Reviewing, validating, and integrating AI-generated code. Foundational skills still matter.

Build the skills behind the work.

A practical learning path for Software Engineering.

GO A LITTLE DEEPEROpen only what you need
What changes — and what doesn’tSkills & responsibilities
AI HELPS WITH
AnalyzeReview

Analyze sanitized errors and propose possible causes. The output is a starting point to inspect, not a decision to accept automatically.

STILL YOUR RESPONSIBILITY

Validate correctness, choose tradeoffs, protect users, and approve changes.

Skills to develop

FoundationsProgramming, algorithms, system design, and security.

AI collaborationProviding task-specific context and requesting explicit assumptions.

VerificationChecking a reproducible duplicate-payment report against independent evidence.

Professional skillsCommunicating tradeoffs and taking responsibility.

Where AI can go wrong3 things to check

A plausible but wrong answer

A plausible patch can hide a symptom while leaving the defect intact. It can fail the underlying goal even when it sounds convincing.

Your check

Check idempotency and reproduce retries before accepting a timeout change.

Missing or invented context

AI may fill gaps with unsupported assumptions, which can send the work in the wrong direction.

Your check

Trace claims to original evidence and ask the relevant person about unknowns.

Information shared in the wrong place

Sensitive records or code can cross confidentiality boundaries if supplied to an unsuitable tool.

Your check

Use approved tools, share the minimum context needed, and follow your organization’s rules.

Try a quick exerciseA practical scenario

A payment retry creates two charges. AI suggests increasing the timeout.

AI proposes increasing a timeout after a retry produces two charges. Which check is missing?
Sources & contextEvidence behind this example

These are illustrative workflows, not claims that AI is always better or that every organization works this way. The scenarios and checkpoints are editorial teaching examples.

Reviewed September 2026 · Growing PracticeO*NET — Software DevelopersSupports the role and task baseline; it does not validate our AI workflow sequences.GitHub — Responsible use of inline suggestionsDocuments review and security responsibilities for generated code; capabilities vary by tool and configuration.GitHub — About Copilot code reviewEvidence of an available assisted-review capability, not proof of universal adoption or correctness.How we build these examples