WorkWise AI is a functional, offline front-end prototype developed by Simphiwe Sethole for the AI Skills Accelerator Programme. It demonstrates how structured prompt engineering can support workplace communication, meeting management, planning, research and conversational assistance.
- Smart Email Studio — creates audience-aware emails in formal, friendly, persuasive or concise tones.
- Meeting Intelligence — extracts key points, decisions, action items and detected deadlines from raw notes.
- Priority Task Planner — ranks tasks using urgency and importance, then creates realistic time blocks.
- Research Brief Generator — produces a source-based summary, insights, recommendations and limitations.
- Workplace AI Assistant — maintains a local multi-turn conversation, recognises workplace intent and routes users to the correct workflow or quality check.
This submission is an educational prototype, not a live large-language-model service. It does not call OpenAI, Gemini or another paid API. Instead, it uses transparent browser-based rules and templates to demonstrate the intended workflows and prompt-engineering structure. No data is uploaded or stored.
- Download or clone the repository.
- Open the project folder.
- Double-click
index.html. - Select a tool, use Load sample or enter your own non-sensitive information, and choose the main action button.
No installation, account, server or API key is required.
Each workflow builds a clear instruction using:
- Role — the professional function the assistant performs.
- Context — audience, topic, notes or task information.
- Task — the exact outcome required.
- Output format — email, summary, plan or research brief.
- Constraints — no invented facts, structured output and human review.
The constructed instruction can be inspected under View prompt-engineering structure after generating an output.
- Do not enter confidential, personal or sensitive data.
- Verify names, dates, facts, deadlines and commitments.
- Treat recommendations as suggestions, not established facts.
- Use qualified human review for legal, medical, financial or other high-stakes decisions.
- The interface clearly identifies itself as an offline demonstration engine.
- HTML5
- CSS3 (responsive, mobile-first adaptations)
- Vanilla JavaScript
- Browser APIs for copy and Word-compatible document download
- GitHub Pages-compatible static architecture
workwise-ai/
├── index.html
├── README.md
├── PROJECT_DOCUMENTATION.md
├── PROMPT_ENGINEERING.md
├── TESTING_CHECKLIST.md
├── OPEN_ME_FIRST.txt
└── docs/
The project includes a completed TESTING_CHECKLIST.md covering offline operation, all five workflows, validation, reset, copy/download behaviour, accessibility, responsible-AI disclosure and GitHub Pages compatibility. A final live-site check should still be completed after deployment.
With approved organisational access and a securely stored API key, a future version could connect the interface to a real language model, add authentication, encrypted storage, user feedback and formal quality evaluation. Keys must never be hard-coded into browser code.
Simphiwe Sethole
BCom Information Systems student | Aspiring Data Engineer