Use-case library
Thirty concise stories across business workflows, personal agents, and software development. Every story shows the process, human control, setup needed, limits, and source type.
30 stories
Enquiries arrive with incomplete context and follow-up depends on memory.
Enquiry arrives / Capture context / Qualify / Review reply / Follow up and remind
OutcomeQualified enquiries with an approved reply and tracked follow-up.
Publishing becomes inconsistent when every post starts from an empty page.
Brief and material / Research / Draft and visual / Edit and approve / Publish
OutcomeA repeatable content rhythm without losing the owner's voice.
Repeated questions consume time while unusual or sensitive cases need careful handling.
Question arrives / Search sources / Draft answer / Check confidence / Handle exception / Answer and log
OutcomeFaster support with visible escalation instead of unsafe guessing.
A website can explain a business yet still leave visitors unsure what to do.
Visitor arrives / Relevant message / Choose a path / Proof and offer / CTA and conversation
OutcomeA clearer journey from first visit to a relevant human conversation.
Important requests are mixed with routine messages and forgotten follow-ups.
Message arrives / Triage / Extract action / Draft reply / Edit and send / Update task
OutcomeA calmer inbox with decisions and sending kept under human control.
Staff repeatedly copy information from PDFs, forms, and attachments into other systems.
Upload / OCR and extract / Validate / Check exceptions / Save and notify
OutcomeLess manual entry with uncertain documents held for review.
Useful reporting is delayed because data sits in several tools and manual spreadsheets.
Pull data / Clean and merge / Detect change / Build summary / Interpret and decide / Alert or act
OutcomeEarlier visibility with interpretation and action owned by a person.
Links, PDFs, and notes take time to turn into a trustworthy brief or comparison.
Define question / Find sources / Extract evidence / Compare / Draft / Validate and approve
OutcomeA draft with traceable evidence, assumptions, and human approval.
Answers are scattered across documents and experienced staff, so work becomes inconsistent.
Staff asks / Find SOP / Answer with source / Handle exception / Improve SOP
OutcomeMore consistent answers with gaps feeding back into the SOP library.
One process crosses several tools, people, approvals, and repeated handoffs.
Trigger / Validate / Route / AI assists / Approve risk / Update and log
OutcomeA visible workflow with routine movement automated and risky actions approved.
A community example describes a scheduled AI tools task that summarizes inbox activity and posts the result to the agreed team channel.
Schedule fires / Read allowed messages / Group and summarize / Check sensitive items / Post briefing
OutcomeA scheduled summary delivered to an agreed channel for human review.
A self-hosted document workflow can connect AI tools with Nextcloud and LibreOffice.
Request document / Find source files / Draft in LibreOffice / Review document / Save to Nextcloud
OutcomeA document workflow connected to self-hosted storage.
A community story describes AI tools researching public information, building a landing page, uploading it to secured hosting, and sending completion notice.
Research public sources / Draft page / Build locally / Review content and access / Deploy to secured hosting / Send completion notice
OutcomeA deployed draft landing page created from public information.
A localized briefing workflow can combine market data, local news, image cards, and scheduled messaging-channel delivery.
Schedule starts / Collect local sources / Summarize / Render card / Review sensitive claims / Deliver to messaging channel
OutcomeA localized briefing card delivered on a recurring schedule.
Multiple AI tools can use one shared local SearXNG container as a private search layer.
Agent submits query / Search local service / Collect results / Review source quality / Use approved evidence
OutcomeReusable search infrastructure under the operator's control.
Terminal voice input can make agent tasks more accessible when long typed prompts are difficult.
Speak request / Transcribe / Show interpreted task / Confirm important action / Execute and respond
OutcomeA more accessible way to operate an assistant from the terminal.
A technical assistant can combine indexed repositories, debugging tools, and approved protocol documentation.
Team asks question / Search approved systems / Combine context / Draft technical answer / Expert validates / Record useful finding
OutcomeA shared technical assistant with access to team-specific sources.
A hosted AI tools setup can expose a browser integration as a reusable skill for repeatable browser tasks.
Receive browser task / Open approved site / Inspect page / Confirm consequential action / Interact and verify / Return evidence
OutcomeA remote AI tools setup with reusable browser automation capability.
AI tools official workflow starts broad, maps architecture, and then narrows into specific components.
Open repository / Map structure / Explain architecture / Inspect key components / Developer validates model
OutcomeA reviewed overview of architecture, data models, and important conventions.
AI tools can inspect symptoms, locate relevant code, propose a cause, edit, and run verification.
Reproduce issue / Find relevant code / Form hypothesis / Review plan / Edit code / Run tests
OutcomeA focused change with evidence from tests or reproduction.
AI tools support planning a refactor, changing code, and checking behavior with tests.
Identify smell / Map callers / Plan refactor / Approve scope / Apply change / Run regression tests
OutcomeA smaller reviewed diff with preserved behavior.
AI tools can inspect implementation and existing test conventions before proposing new cases.
Inspect behavior / Read test patterns / Propose cases / Approve coverage / Write tests / Run suite
OutcomeA reviewed test set that covers agreed scenarios and edge cases.
AI tools can inspect changes, run checks, summarize the diff, and help prepare a pull request.
Inspect diff / Run checks / Review findings / Fix agreed issues / Approve final diff / Draft PR
OutcomeA PR description with tests, scope, and known limitations.
AI tools documents parallel sessions in isolated git worktrees so concurrent edits do not collide.
Split independent tasks / Create worktrees / Run sessions / Compare results / Approve integration / Merge selected work
OutcomeParallel branches that can be reviewed and integrated separately.
AI tools can inspect a local repository, edit code, and run installed development tools.
Inspect repository / Clarify task / Plan change / Approve plan / Edit code / Run checks
OutcomeA local diff with verification output ready for human review.
AI tools include a local review workflow for prioritized findings before commit or pull request.
Load diff / Inspect context / Identify findings / Verify each issue / Apply selected fixes / Rerun review
OutcomePrioritized findings that a developer can verify and address.
Official AI workflow guidance supports non-interactive AI coding automation in repeatable scripts and pipelines.
Pipeline starts / Load bounded context / Run AI coding task / Collect result / Approve change or finding / Continue pipeline
OutcomeA repeatable result that can be reviewed before affecting the repository.
AI tools support adding screenshots, diagrams, and design references to the initial task.
Attach image / Describe target / Inspect related code / Propose change / Review visual intent / Edit and test
OutcomeA proposed implementation grounded in both repository and visual context.
AI tools can delegate focused investigations to subagents and bring findings back to the main session.
Define investigation / Split work / Run subagents / Collect findings / Validate conclusions / Choose next action
OutcomeConsolidated findings with separate evidence from focused investigations.
AI tools can launch work in a configured cloud environment and later apply the result locally.
Choose task / Confirm environment / Launch cloud work / Inspect result / Approve application / Apply and test locally
OutcomeA remote result that can be inspected and selectively applied to the repository.
