Table of Contents

1. Executive Summary

SABI receives a significant volume of bid enquiries via email at estimation@sabi.ae. Currently, the process of identifying, classifying, processing attachments, extracting project data, calculating MEP service estimates, and preparing quotations is largely manual — consuming considerable time and introducing risk of human error or missed opportunities.

Bettroi proposes the development of SABI – an AI-Powered MEP Email Processing and Estimation Automation System — a purpose-built intelligent platform that automates the end-to-end bid workflow: from email ingestion and classification through to quotation preparation and dispatch.

The system is designed to dramatically reduce the time from enquiry receipt to quotation submission, increase bid capacity, and ensure consistent, audit-ready estimation methodology — while keeping SABI's estimating team in control at every critical decision point.

The solution will be delivered as a web-based application accessible from any device, with a clean dashboard for managing the bid pipeline, reviewing AI-extracted data, and approving quotations before dispatch.

2. Scope of Work

2.1 In-Scope

• Email Monitoring: Connect to estimation@sabi.ae via Gmail API / IMAP; monitor inbox in real time or scheduled intervals
• RFQ Detection: AI-based keyword detection for enquiry signals ("Please quote", "Waiting for your best price", "will be interested", "send your rates", "request for quotation", and similar)
• To-Bid List Management: Automatic addition of qualified enquiries to a central bid pipeline dashboard
• Enquiry Classification: AI classification into: Priority-Top, Priority-General, New, To Be Ignored — with override capability for estimators
• Attachment Enclosure: Transfer of email body and attachments to a project-specific folder; support for PDF, DWG, ZIP file types
• ZIP Extraction: Automatic unzipping of compressed attachments; recursive extraction supported
• Drawing List Generation: Catalogue of all drawings by file type (ACAD .dwg and PDF), with file names and page counts
• Project Parameter Extraction: AI extraction from drawings/documents of: number of floors, parking floors, typical floors, area per floor (m²/ft²), total area, typical floor-to-floor height
• Service Identification: Detect MEP services to be priced (Air Conditioning, FAHU, Plumbing, Fire Fighting, Electrical, BMS, etc.) from drawing titles and BOQ headers
• Thermal Load Analysis – Air Conditioning:
  • Open and parse thermal load schedule drawings (PDF)
  • Extract total calculated KW from bottom-of-column summation
  • Identify FAHU total KW separately
  • Calculate AC unit total (Total KW minus FAHU KW)
  • Identify AC system type (VRF, Chilled Water, Split, etc.)
  • Convert KW to Refrigeration Tonnes (RT)
• Formula-Based Pricing: Apply SABI's pre-configured pricing formulas per service type and system type to produce an estimated bid price
• Confirmation Workflow: System pauses at key decision points for estimator confirmation before proceeding
• Total Bid Compilation: Aggregate pricing across all services into a total bid amount
• Yardstick Comparison: Compare computed bid against area-based yardstick benchmarks; flag variances >15%
• Quotation Preparation: Generate a formatted quotation document (PDF/HTML) ready for dispatch
• Controlled Email Dispatch: Send quotation to client only after explicit estimator consent via dashboard
• Audit Trail: Full log of every email received, action taken, data extracted, and quotation sent
• Web Dashboard: Clean browser-based UI for managing bid pipeline, reviewing AI outputs, approving dispatch

2.2 Out-of-Scope

• Parsing of native ACAD (.dwg) files — only PDF drawings will be parsed in this phase
• Structural, Civil, or Architectural estimation — MEP services only
• Integration with any existing ERP, accounting, or tendering system
• Hardware procurement or on-premise server setup
• Training of custom ML models — Gemini API is used as-is
• Multi-company / multi-inbox deployment in this engagement
• Mobile native app (iOS / Android)
• Automated revision management for reissued drawings

3. System Architecture & Workflow

3.1 Email Intake & Classification Module

3.2 Attachment Processing Module

3.3 Thermal Load & Estimation Module (Air Conditioning)

3.4 Quotation & Dispatch Module

4. Technology Stack

Architecture Principles

• Human-in-the-loop: AI extracts and computes; humans approve at every critical step
• Configurable pricing: All pricing formulas stored in database — updatable by SABI without developer involvement
• Audit-first design: Every action, AI decision, and email sent is logged with timestamp and user
• Scalable: Architecture supports adding more services and drawing types in future phases
• Secure: OAuth 2.0 for Gmail; Row Level Security on all database tables; no email credentials stored in plaintext

Confirmation Checkpoints (Human-in-the-Loop)

5. Project Plan & Milestones

The project is structured across 2 Phases, 6 Sprints, 3 Months. Each sprint is 2 weeks. A confirmation checkpoint is built into each sprint before advancement.

Key Milestones

6. Commercials

6.1 One-Time Development Cost

6.2 Monthly Operational Costs (Post Go-Live)

6.3 Payment Schedule

Monthly operational costs (₹24,800 + Google API) commence from go-live date and are invoiced monthly in advance.

7. Client Responsibilities

• Gmail API Access: Provide Google OAuth 2.0 credentials for estimation@sabi.ae within Week 1. Ensure the account has appropriate permissions for read, send, and attachment download.
• Pricing Formulas: Provide SABI's pricing formula/rate card per service type (e.g., INR or AED per RT for VRF, per m² for plumbing, etc.) before Sprint 4 begins.
• Yardstick Values: Provide benchmark cost-per-m² values for each MEP service category before Sprint 5.
• Sample Drawings: Provide at least 5 real project drawing sets (PDF) for AI training validation and testing during development.
• Thermal Load Format: Confirm the standard format of SABI's thermal load schedule drawings so the parser can be calibrated accurately.
• UAT Participation: Nominate a designated estimator to participate in User Acceptance Testing (UAT) during Sprint 6. A minimum of 5 full bid cycles must be tested.
• Feedback Turnaround: Provide feedback on each sprint deliverable within 3 business days to avoid timeline delays.
• Branding Assets: Provide SABI logo, letterhead template, and corporate colours for the quotation document template.
• Point of Contact: Designate a single technical point of contact for access, credentials, and decision approvals throughout the project.

8. Important Points

• AI accuracy: The AI extracts data from PDF drawings. Accuracy is dependent on drawing quality and legibility. Handwritten annotations or low-resolution scans may not parse correctly.
• ACAD files: DWG files will be catalogued and listed but not parsed in Phase 1. PDF equivalents must be available for thermal load analysis.
• Pricing engine is formula-driven: The system applies formulas as configured. Pricing accuracy depends on the quality and completeness of SABI's input rate cards.
• Human oversight mandatory: The system is an automation aid, not a replacement for estimator judgement. All AI outputs must be reviewed before quotation is sent.
• Email sending control: No email is ever sent automatically. Every dispatch requires explicit estimator approval in the dashboard.
• Third-party API limits: Gmail API, Gemini API, and Supabase have rate limits. High-volume usage may require plan upgrades at additional cost.
• Data residency: Project data will be stored on Supabase (hosted on AWS). If specific data residency (UAE/India) is required, this must be communicated before development begins.
• Scope changes: Any feature not listed in Section 2.1 (In-Scope) will be treated as a change request and quoted separately.

9. Assumptions

• Thermal load schedules are provided as PDF drawings with tabular KW summation at the bottom of each column — consistent across projects.
• Enquiries will arrive primarily via email at estimation@sabi.ae with attachments in PDF or ZIP format.
• SABI will provide pricing formulas in a structured format (Excel or Google Sheet) that Bettroi can load into the system's formula engine.
• At least one estimator will be available for UAT during Week 11–12.
• Google API services (Gmail, Drive) will remain available under SABI's existing Google Workspace plan. No new Google Cloud billing account setup is in scope.
• SABI's drawing sets follow a consistent naming and folder convention so that service type can be reliably inferred from file names or sheet titles.
• The Gemini API key provided is production-grade with sufficient quota for the expected volume of requests.
• All feedback and approvals from SABI will be provided within 3 business days of each deliverable submission.
• This is a single-tenant deployment for SABI only. Multi-company requirements would require a separate engagement.

10. Change Control

Any modification to the agreed scope of work — including new features, integration with additional systems, support for new drawing formats, or changes to the confirmation workflow — will be managed through a formal Change Request (CR) process:

1. Client submits a written Change Request describing the required modification.
2. Bettroi assesses the CR and provides a written estimate of effort, cost, and timeline impact within 5 business days.
3. Client approves the CR in writing (email acceptable) before work commences.
4. CR costs are invoiced as an addendum to this proposal and are payable as per the standard payment terms.

Examples of Activities that Require a Change Request

• Adding support for native .dwg (AutoCAD) file parsing
• Adding a new MEP service type to the estimation engine beyond what is confirmed at kickoff
• Integrating with a tendering platform, ERP, or CRM system
• Deploying for additional email inboxes or additional companies
• Custom mobile application development
• Changes to quotation PDF format after template is approved

11. Warranty, IP & Terms

11.1 Warranty

• Warranty Period: 3 months from go-live date (end of Sprint 6).
• During the warranty period, Bettroi will fix any defects in the delivered system at no additional charge, provided the defect is reproducible and falls within the agreed scope.
• Warranty does not cover issues arising from: changes made to the codebase by third parties, changes in third-party APIs (Gemini, Gmail, Supabase) that alter expected behaviour, or data quality issues in drawings provided by SABI.
• Post-warranty support and enhancements are available under a separate Annual Maintenance Contract (AMC) to be agreed upon at go-live.

11.2 Intellectual Property

• Upon receipt of full and final payment, SABI will own the developed application code, database schema, and all deliverables produced under this engagement.
• Bettroi retains the right to reuse generic architectural patterns, utility functions, and non-proprietary components in future projects — but will not share or reuse any SABI-specific business logic, pricing formulas, or drawing analysis outputs.
• Third-party libraries, APIs, and frameworks (Next.js, Supabase, Gemini) remain subject to their respective licences.

11.3 Confidentiality

Bettroi agrees to treat all SABI pricing formulas, yardstick values, drawing data, and client information as strictly confidential. This information will not be shared with any third party and will be used solely for the purpose of this engagement. A separate Non-Disclosure Agreement (NDA) can be executed upon request.

11.4 Liability

Bettroi's liability is limited to the total development fee paid under this proposal (INR 3,00,000). Bettroi shall not be liable for any indirect, consequential, or commercial losses arising from the use of the system, including losses from quotations submitted based on AI-extracted data not reviewed by the estimator.

11.5 Governing Terms

• Proposal Validity: This proposal is valid for 60 days from 19 February 2026 (until 19 April 2026).
• Governing Law: This agreement is governed by the laws of the United Arab Emirates.
• Dispute Resolution: Disputes shall first be resolved through good-faith negotiation. Unresolved disputes shall be subject to arbitration under Dubai International Arbitration Centre (DIAC) rules.
• Force Majeure: Neither party shall be liable for delays caused by events beyond reasonable control (acts of God, government mandates, infrastructure outages).

11.6 Acceptance Criteria

• All 7 confirmation checkpoints function as specified in Section 3
• Email classification accuracy >85% on SABI's historical email sample set
• Thermal load KW extraction accuracy >90% on provided PDF drawing samples
• Quotation PDF generated correctly for at least 5 completed UAT bid cycles
• No critical bugs remaining after UAT sign-off
• Full audit log accessible in dashboard for all actions taken

12. Service Order Confirmation – Letter of Intent (LOI)

By signing below, SABI confirms acceptance of this Commercial Proposal (Ref: BETTROI K32 / SAL-QTN-2026-00002) and authorises Bettroi FZE to commence development of the SABI – AI-Powered MEP Email Processing & Estimation Automation System as described herein.

SABI confirms that all information provided in this LOI is accurate and that the authorised signatory has the authority to bind SABI to this agreement.