AI Technical Specification Assistant

To develop an AI-powered Engineering Specification Assistant
- The solution must:
1. Provide a high-accuracy, citation-based interface for querying Aurora Water's technical engineering specification manuals.
2. Serve both internal staff and external users.
3. Be a browser-based user interface embedded on the City’s website.
4. Be maintainable by City staff as manuals are updated, typically on an annual basis, with appropriate access controls.
5. Be scalable to accommodate additional manual types and city code searches over time.
6. Include a complete project management plan and team that coordinates a successful go-live.
7. Include a training plan for Aurora Water administrative roles and a public-facing training resource
8. Include a support and maintenance agreement for OTSS solutions or a handoff and support plan for professional service custom-built solutions.
9. Comply with all City security, accessibility, and data privacy requirements as specified in the Master Services Agreement and related exhibits, unless mutually agreed modifications are accepted
during the proposal process.
- Accuracy and Response Quality
1. The solution must answer user questions accurately.
2. All answers must cite the relevant section and page number from each specification manual that relates to the user question.
3. The solution must support follow-up questions within a session.
4. The solution must provide answers grounded and traceable to the uploaded specification manuals only. The system must not infer, extrapolate, or draw from outside knowledge.
5. When a query relates to a detail sheet within a specification manual, the solution must provide a direct link to the relevant detail page in addition to the answer.
6. When an answer cannot be found, the solution must direct users to the appropriate Aurora Water Engineering inbox for further assistance.
7. Preferred - the solution should display a snippet or excerpt from the relevant page of the specification manual to provide visual confirmation.
8. Preferred -The ability to show the reasoning or retrieval path behind an answer to administrative staff.
- Browser-Based Chat Interface
1. The solution must provide a browser-based chat interface.
2. The chatbot must be embeddable directly on existing City web pages — users must not be required to navigate to a separate third-party site.
3. The interface must clearly disclose to users that AI is in use and must provide a direct link to the specification manuals for users who wish to reference them independently.
4. The interface must be able to incorporate disclaimer language from the City’s legal team, including notices such as “Submitted prompts are not confidential” or “AI can make mistakes”.