1. Product vision, roadmap & strategy

- Define and own the product vision, roadmap and execution strategy for Smartask 2.0 and Celia, keeping every decision anchored to the Output-vs-Outcome thesis.

- Translate the four-stage workflow and the Fiverr-style "Submitted - Needs Correction - Accepted" model into clear, low-friction experiences for clients, the CLOUDIT team, firms and management.

- Prioritise ruthlessly: ship a thin, working slice first, then expand through a phased rollout.

2. AI platform architecture (LLMs & RAG)

- Design and develop the AI platform architecture: Celia as an aggregator/orchestration layer that routes prompts across multiple LLMs (e.g. ChatGPT, Claude, Gemini, Grok, Perplexity) behind one interface.

- Architect the RAG framework - ingestion, embeddings, vector storage and retrieval - so prompts return accurate, business-specific answers.

- Define the prompt-routing logic, guardrails and fallback strategy to balance quality, latency and cost; avoid single-vendor lock-in.

- Establish token metering and an AI-hours model that feeds billing (token cost + margin + human hours).

3. Smartask workflow & human-validation layers

- Build and integrate the Smartask workflow with AI and human-validation layers across all four stages: Pre-Processing, CLOUDIT Review, Firm Review and Client Output.

- Implement the "Satisfied / Need Human Validation" branch, task-level chat (current + archived), the documents- outputs toggle, and the consultation-booking flow.

- Build time logging that captures AI-hours, human review hours and call hours at their own rates.

4. Client-level knowledge systems & integrations

- Develop per-client knowledge systems: normalise structured and unstructured data (ERP/accounting, POS, CRM, spreadsheets, meeting notes, owner/employee knowledge) into one unified data structure per client.

- Build and maintain integrations and connectors to common accounting/ERP and POS systems.

- Define how the knowledge base is kept current and continuously improved as new client data arrives.

5. Engineering leadership & delivery

- Lead, mentor and grow the engineering, data and AI teams; set technical standards and ways of working.

- Own delivery: run agile cadences, manage scope and dependencies, and ensure timely, high-quality releases.

- Make pragmatic build-vs-buy and architecture decisions, and own the technical risk register.

6. Cross-functional collaboration

- Work cross-functionally with Operations, partner firms and clients to ground the product in real workflows.

- Enable seamless multi-party collaboration: firm-branded email, in-app chat/notifications, reminders, and VOIP with call recording and transcripts available to firm owners.

- Partner with the CEO and management on pricing guardrails, go-to-market and rollout decisions.

7. Quality, performance & cost efficiency

- Monitor AI output accuracy and define how quality is measured, reviewed and improved over time.

- Own system performance, reliability and security - including the encrypted credential vault, role-based access control and a full audit log.

- Track and optimise AI token cost and infrastructure spend against revenue and margin targets.

8. MVP & pilot execution

- Drive MVP development to a live, demonstrable end-to-end task with one pilot firm and one client.

- Run the pilot, capture feedback, validate billing on a real sample invoice, and convert learnings into the rollout plan.