AI Innovation HomeAgentic Delivery ModelAWS Agentic PlatformConsulting POCsVendor & Contract SupportAI TrainingEngage With Us
Operating model
An enterprise model for agentic software delivery
AI coding agents are fast, but speed without structure creates risk. Our model puts clear specs, defined roles and human approval gates around every agent-written change.
OpenSpec is an open-source, spec-driven development framework for AI coding assistants. Before an agent writes code, the change is captured as a proposal with tasks and spec updates that people review and agree on.
The spec-driven delivery loop
- ProposeCapture the change as a proposal, a task list and spec updates, drafted with the help of an AI assistant.
- ReviewProduct owners, architects and engineers review and agree the spec before any code is written.
- ImplementAI coding agents work through the agreed tasks, with engineers steering and pairing.
- ValidateAutomated tests, code review, security scanning and spec checks confirm the change does what was agreed.
- ArchiveThe change is merged and its spec updates become the new source of truth for the system.
Four pillars of the operating model
People
- Clear roles for spec authors, reviewers and AI-enabled engineers
- Skills uplift and AI champions in each team
- Change management and adoption support
Process
- Spec-driven lifecycle with OpenSpec
- Definition of ready and done for agent work
- Fits with Agile, Scrum and existing SDLC controls
Platform
- Secure AWS environment for models and agents
- Integrated with your repos, CI/CD and ticketing
- Usage, cost and quality dashboards
Governance
- Human approval gates on specs and merges
- Guardrails, audit logging and data controls
- Aligned with Australia's AI Ethics Principles
Roles and responsibilities
| Role | Responsibility |
|---|---|
| Product owner | Sets priorities, writes acceptance criteria and approves proposals. |
| Spec lead / architect | Shapes proposals and specs, keeps the architecture consistent across changes. |
| AI-enabled engineer | Directs coding agents through tasks, reviews output and handles the hard problems. |
| Reviewer / QA | Checks code and tests against the agreed spec before merge. |
| Platform team | Runs the AWS platform, agent tooling, access and cost controls. |
| AI governance lead | Owns policy, risk, guardrails and reporting to leadership. |
Humans stay in control. Agents never merge their own work: every spec and every merge to a protected branch needs a person's approval.
Ready to explore agentic AI?
Book a discovery conversation with our team and we’ll help you find the right first step.