Experience & integrations

Create a responsive, mobile-friendly reference interface for Point Preserve.

The initial experience should prioritize speed and clarity.

Include:

•tasks assigned to me;

•tasks created by me;

•tasks awaiting my review;

•overdue tasks;

•blocked tasks;

•recent activity;

•active projects;

•recurring maintenance due;

•quick voice capture;

•quick photo task;

•quick text task.

•original request;

•source artifacts;

•structured task;

•assignee;

•status;

•checklist;

•due date;

•location and asset;

•comments;

•voice updates;

•evidence;

•review history;

•audit timeline;

•linked knowledge;

•related tasks.

A user should be able to:

1.tap a voice, camera, upload, or text action;

2.provide the input;

3.review the AI interpretation;

4.correct important fields;

5.confirm task creation.

Do not require users to complete a long form before capturing an urgent issue.

Use progressive enrichment after initial capture.

Show:

•work submitted for verification;

•required evidence;

•checklist status;

•before and after artifacts;

•reviewer actions;

•revision history;

•final approval notes.

Support:

•project summary;

•task hierarchy;

•task board;

•progress;

•owners;

•due dates;

•risks;

•costs when authorized;

•evidence;

•lessons learned;

•generated knowledge.

Reuse existing connectors and shared integration services.

Evaluate integration points for:

•Google Drive;

•Gmail;

•Apple Note exports;

•uploaded files;

•mobile share sheets;

•calendar reminders;

•communications and notifications;

•vendor/contact records;

•purchasing or expense systems;

•OwnerRez or property systems where relevant;

•future Apple ecosystem connectors;

•Supabase;

•Vercel;

•GitHub;

•Good Combinator Knowledge and Evidence services.

Do not create unnecessary direct integrations inside the core task service.

Use adapters and versioned contracts.

AI may assist with:

•transcription;

•summarization;

•task extraction;

•project extraction;

•categorization;

•priority recommendation;

•duplicate detection;

•checklist generation;

•assignment recommendation;

•due-date extraction;

•evidence completeness checks;

•knowledge proposal generation;

•SOP suggestions;

•operational question answering;

•pattern detection.

AI must not:

•conceal uncertainty;

•fabricate instructions;

•create unsupported commitments;

•approve its own consequential work;

•invent evidence;

•mark work verified without the required authorization;

•publish unreviewed material as canonical policy;

•expose records outside the user’s permissions.

Material AI outputs should include:

•confidence;

•provenance;

•model or processing metadata where required;

•human-review status;

•links to source artifacts.