Explore how professionals in Software Engineering use AI as a tool. Real use cases, practical examples, and community stories.
24 use cases in Software Engineering
Use AI coding agents and assistants to generate boilerplate, scaffold features, implement solutions end-to-end, and execute large-scale changes in one coherent agentic pass — from quick prototypes and API integrations to full migrations and review-ready PRs. Adding structure upfront (briefs, constraints, approved plans) makes AI output more predictable and the iteration process reusable.
Use AI to analyze bugs and production issues, interpret logs and distributed traces, suggest root causes, propose fixes, and surface edge cases that strengthen implementation and testing — accelerating incident triage and root-cause analysis (RCA).
Use AI — individually or as parallel subagents — to frame problems, explore architecture alternatives, write and refine specs, map existing codebases, and produce step-by-step implementation plans before writing a single line of code. Structured upfront investment reduces ambiguity and makes execution far more predictable.
Use AI coding assistants inside the IDE and terminal to accelerate day-to-day development — generating inline suggestions, multi-file edits, tests, refactors, and shell commands without leaving the dev environment. Combine this with version-controlled workflows (feature branches, PRs, CI gates) so all AI-generated changes are auditable and safely reversible.
Use AI to generate unit and E2E tests, expand coverage, and provide automated PR feedback — reducing manual test authoring and surfacing issues before human review.
Use AI to both communicate more clearly — drafting PR descriptions, Slack updates, incident notes, and postmortems from rough thoughts — and to automate the daily context-gathering work: pulling PRs, issues, and doc changes into standup-ready summaries, and enabling conversational queries over project data.
Create and maintain shared context (docs, repository context files, ADRs) and persistent workspaces (projects/memory/shared files) so AI assistants generate consistent, on-standard output and retain key decisions and constraints over time instead of resetting every chat.
Shift quality assurance from primarily human review/knowledge to infrastructure-driven verification (automated checks, correctness/reliability guardrails) and release-readiness workflows — including AI-assisted PR review, logs/metrics review, edge-case checks, regression diffs, and production-state validation — to safely handle a much higher volume of AI-generated changes.
Use an AI coding assistant as a force multiplier to build, ship, and iterate on software as a solo developer — whether building a personal operating system that connects workplace tools (Slack, Linear, Notion) or shipping a full niche product that previously required multiple roles (dev, design, PM).
Maximize AI productivity by routing tasks to the model best suited for them (Claude for deep reasoning, Gemini for Workspace, ChatGPT for coordination) and by configuring layered personal instructions so AI output consistently matches your team's conventions, review rubrics, and decision frameworks.
Use AI to compare rendered output against an existing site, identify missing metadata (e.g., OG/Twitter tags), and surface accessibility issues (e.g., adding “skip to content”), then apply targeted UI/CSS changes based on feedback.
Use AI to plan and implement deployment steps for a VPS (DigitalOcean), including DNS/SSH setup, Ansible configuration, reverse proxy setup, hardening guidance, and troubleshooting issues like HTTPS visibility discrepancies.
Leverage AI to quickly create supporting software artifacts like landing pages and internal tools that would normally require additional team capacity.
Use AI to propose and implement refactors across larger codebases, including structural and architectural changes.
Export Cursor usage data and analyze token composition (e.g., cache reads vs. generation) to identify cost drivers like repeated context reloads, then adjust workflow (summarization/resetting sessions, context layering) to reduce spend without sacrificing output quality.
Apply a deliberate workflow (switching between different “modes” of AI assistance) to choose when autonomous agent work is worth it versus when direct manual edits are faster and clearer.
Use Claude Code to read/search an Obsidian vault and help with organization (backlinking, reformatting), retrieval (surfacing relevant notes and connections), and triage (processing fleeting notes by adding context and cross-referencing with operational systems like Notion).
Creates an automation/agent (“slack-cleanup”) that analyzes Slack channel activity over time and cross-references business context (e.g., Salesforce) to recommend which channels to leave, reducing noise.
Automates syncing and indexing of meeting transcripts (e.g., Granola) and documents (e.g., Notion) so context can be searched and reused across future conversations and work.
Uses an AI assistant with persistent memory of initiatives/targets plus a weekly check-in to recommend what to focus on beyond just calendar-based scheduling.
Generate standalone, disposable Obsidian vaults where an LLM discovers entities and writes interlinked markdown notes to map a new domain (e.g., an industry landscape or a concept graph) for exploratory navigation.
Runs fast, “cheap” one-off analyses (e.g., competitive deep-dives) grounded in internal data such as win rates and Gong call snippets/quotes to surface gaps and hypotheses quickly.
Documents LLM usage directly in Git history—sometimes via inline comments, but primarily via per-commit `Co-authored-by` trailers that record the model/provider—so teams can later identify where AI-generated code entered the codebase (useful for review, provenance, and potential legal/indemnity requirements).
Uses ChatGPT/Claude in a back-and-forth to convert manual processes into clear checklists/runbooks and sometimes generate an automation script.