# R&D Activity Register — FY2025-26

**Company:** RateRight Pty Ltd | ABN 62 841 523 907
**Period:** 1 July 2025 – 30 June 2026
**Status:** DRAFT — needs Michael review and sign-off
**Purpose:** ATO Advanced Overseas Application / R&D Tax Incentive claim + MVP Ventures grant co-contribution proof

---

## Overview

RateRight is developing a **voice-first construction hiring marketplace** that connects contractors (hiring tradies) directly with workers (tradespeople seeking employment). The platform uses AI agents for matching, communication, and operations management.

The core innovation is **voice-first matching**: unlike existing platforms (Yakka Labour, TradeJobSearch) which are text-only, RateRight allows contractors to describe hiring needs by voice, with AI agents processing and matching to available workers in real time.

**Four interconnected systems being built:**

1. **RateRight** — Marketplace platform (rateright.com.au)
2. **OpsMan** — Operations management for construction (8-agent AI fleet)
3. **Growth Engine** — CRM/sales intelligence
4. **Phone AI** — Voice AI for construction (Twilio + ElevenLabs)

All four systems are being developed simultaneously as an integrated platform.

---

## Core R&D Activities

### 1. Voice-to-Match AI Architecture

**What:** Developing an AI pipeline that converts voice descriptions from contractors into structured job requirements, matches them to worker profiles using multi-factor scoring (trade, location, rating, availability), and delivers matches via voice callback.

**Technical challenge:** Voice input is unstructured, accents vary, noise on construction sites affects clarity. Traditional NLP pipelines fail on accented Australian English and construction jargon. We needed to develop a custom transcription enhancement layer + matching algorithm.

**Activities:**
- Custom prompt engineering for Whisper API (OpenAI) to handle Australian construction accents
- Development of structured matching algorithm with fallback hierarchies
- Voice delivery pipeline using ElevenLabs API
- Real-time SMS fallback when voice calls fail

**Technologies:** OpenAI Whisper API, ElevenLabs TTS, Twilio Voice, Node.js, WhatsApp API

---

### 2. Multi-Agent Fleet Coordination

**What:** Building an 8-agent AI system where each agent (Harper, Builder, Susan, Sentinel, Radar, Herald, Cog, Rivet) operates its own Clawdbot gateway and coordinates via JSONL inbox system + shared fleet-state.json.

**Technical challenge:** No off-the-shelf multi-agent framework existed that could coordinate independent AI agents across separate gateway instances while maintaining shared state and inbox reliability.

**Activities:**
- JSONL inbox system design and implementation (per-agent inboxes)
- Shared fleet-state.json schema design and sync protocol
- Fleet CLI tooling (fleet-cli.js, fleet-update.js, agent-bridge.js)
- Health monitoring and stall detection for all 8 agents
- Agent-to-agent wake protocol (CONVERSATION ABORTED / TIMED OUT handling)

**Technologies:** Node.js, JSONL, OpenClaw multi-agent framework, custom CLI tools

---

### 3. RateRight Marketplace Platform

**What:** Building a web platform that handles contractor onboarding, worker matching, $50 flat-fee payment processing (Stripe), and post-hire communication management.

**Technical challenge:** Building a marketplace that works nationally from day one (no geographic restrictions) with zero listing fees, unlimited duration matches, and contractor profile management for construction industry-specific needs.

**Activities:**
- WordPress/WooCommerce site with SSL (rateright.com.au)
- Stripe integration for $50 flat fee (no ongoing fees)
- Contractor profile system with trade categories, ratings, availability
- Worker onboarding with experience verification
- Post-hire communication management system
- Mobile-responsive design for site use

**Technologies:** WordPress, WooCommerce, PHP, Stripe API, Cloudflare SSL

---

### 4. Phone AI Voice System

**What:** Building a voice AI system that can receive inbound calls from contractors, transcribe their hiring needs, process the request through the matching engine, and deliver results via voice callback.

**Technical challenge:** Real-time voice processing with Twilio requires low latency, reliable fallback mechanisms, and handling of poor mobile reception on construction sites.

**Activities:**
- Twilio phone number setup and webhook integration
- Voice transcription pipeline (inbound call → Whisper → structured request)
- Callback delivery system using ElevenLabs
- SMS fallback for failed voice delivery
- Call logging and analytics

**Technologies:** Twilio, OpenAI Whisper, ElevenLabs, Node.js

---

### 5. Growth Engine CRM Intelligence

**What:** Building a CRM system that tracks construction companies, their hiring patterns, and outreach progress — integrated with email sequencing, Apollo.io data enrichment, and pipeline management.

**Activities:**
- Contact database with company/individual hierarchy
- Outreach tracking and sequence management
- Apollo.io integration for company data enrichment
- Email sequence automation via Gmail
- Pipeline stage tracking (Lead → Contacted → Interested → Hired)

**Technologies:** Apollo.io API, Gmail API, Node.js, Airtable-style data structures

---

### 6. OpsMan Operations Management

**What:** Building a comprehensive operations management system for construction companies, using the 8-agent AI fleet as a proof-of-concept / prototype product.

**Technical challenge:** Demonstrating that the multi-agent system can handle real-world construction operations (scheduling, compliance, communication, reporting) as a standalone product.

**Activities:**
- Agent role definitions (Harper=Finance/Legal, Builder=Code, Susan=Sales, etc.)
- Inter-agent communication protocols (inbox, CONVERSATION, fleet-state)
- Operations workflow definitions for construction management
- Compliance tracking (Fair Work Closing Loopholes, Payday Super)
- Infrastructure monitoring via Sentinel (Fly.io, UptimeRobot)

**Technologies:** OpenClaw, Fly.io, UptimeRobot, custom Node.js scripts

---

## Eligible Expenditure Summary — FY2025-26

| Category | Description | Amount | R&D Eligible |
|----------|-------------|--------|-------------|
| API costs | OpenAI (Claude Max) — all R&D activities | ~$3,400 | Yes |
| API costs | Twilio — voice system | ~$760 | Yes |
| API costs | ElevenLabs — voice delivery | ~$220 | Yes |
| API costs | Apollo.io — CRM intelligence | ~$510 | Yes |
| Infrastructure | Fly.io hosting — all systems | ~$1,890 | Yes |
| Software | Google Workspace — operations | ~$55 | Yes |
| Software | Xero accounting — operations | ~$110 | Yes |
| **Total eligible** | | **~$6,945** | |
| **R&D refund (43.5%)** | | **~$3,021** | |

**Note:** Claude Max subscription covers all AI work (fleet coordination, voice processing, CRM intelligence, platform development). All activities above are core R&D.

---

## TRL Assessment

- **Beginning of FY2025-26:** TRL 3 (experimental proof of concept — voice matching idea)
- **End of FY2025-26 (current):** TRL 6 (technology demonstrated in relevant environment — all 4 systems live, RateRight launched Feb 23)
- **Target end of project (June 2027):** TRL 9 (successful system deployment — national scale, revenue-generating)

**Key milestones achieved:**
- Feb 23 2026: rateright.com.au LIVE with SSL (TRL 6)
- Feb 2026: 8-agent fleet operational (OpsMan TRL 6)
- Mar 2026: Voice matching pipeline functional
- Mar 2026: Growth Engine CRM operational

---

## External Expert Review

For the ATO Advanced Overseas Application (if required), external expert review is needed to confirm:
1. The technology is genuinely novel (voice-first construction marketplace)
2. The R&D activities meet the "new knowledge" test
3. The expenditure is properly attributed to core R&D activities

This will require engaging a registered R&D tax consultant or patent attorney with construction tech experience.

---

## Actions Needed from Michael

1. **Review and sign this register** — confirm activities are accurate
2. **Identify any additional expenditure** I may have missed
3. **Confirm contractor/worker time allocation** for any wages paid (none currently — all work is AI agent + founder)
4. **Engage R&D tax consultant** for external expert review (~$500-1,500)

---

*Prepared by: Harper (Finance & Legal Agent) | Date: 2026-03-30*
*Review required by: Michael McLoughlin (Director)*