Software gets complex fast
Business rules, compliance needs, security issues, and customer expectations keep changing. Traditional documentation and code often drift apart.
Beginner one-pager
AI/works, explained simply
AI/works is Thoughtworks' way of using AI agents, engineering practices, and human oversight to help teams plan, build, modernize, and keep software healthy over time.
Business idea or old system
↓
Clear living specification
↓
AI-assisted build and tests
↓
Governed software delivery
In one sentence
Think of it as a software delivery platform where requirements, architecture, code, tests, governance, and AI agents work from the same source of truth, with experienced engineers guiding the process.
Why it exists
Legacy systems are hard to understand
Many organizations have old code that still runs critical work, but the original logic is buried inside the system.
AI needs guardrails
AI can speed up delivery, but teams still need architecture, quality checks, cost control, security, and human accountability.
How it works
1Understand
It can inspect an existing system and help reconstruct what the software actually does.
2Specify
It turns requirements into a dynamic specification covering workflows, architecture, security, data, and user experience.
3Build
AI agents help generate testable code from the specification, while engineers steer the choices.
4Operate
A control plane tracks agents, guardrails, cost, lineage, and delivery practices.
What a beginner should remember
- It is not just a code generator; it is positioned as an end-to-end delivery platform.
- The central idea is a living specification that stays connected to the software.
- It can help modernize old systems by recovering business logic from existing code.
- Human engineers remain important for architecture, judgment, validation, and accountability.
- The promise is faster delivery without losing control over quality, security, and governance.
Plain English version: AI/works tries to make software development less like handing scattered notes to many people, and more like keeping everyone, including AI agents, aligned around one reliable blueprint.
The delivery promise
Thoughtworks describes a "3-3-3" model: roughly 3 days to shape a concept, 3 weeks to create a prototype, and 3 months to reach a production MVP.