Work as Code (WaC) is a business paradigm, created by Fourier Partners, that treats operational knowledge — tasks, roles, decisions, and outcomes — as a structured, executable asset, rather than undocumented practice spread across tools, documents, and individual memory.
Most organizations don’t have a documentation problem — they have an execution problem. Work as Code closes the gap between how work is described and how it actually runs, so the same definition governs whether the work is done by a person, an agent, or both.
- Explicit. All operational logic is written down — decisions, approvals, exceptions.
- Portable. Platform-agnostic. The description governs; any system can execute it.
- Owned. The organization controls its operational definitions permanently. No third party holds the keys to how the business runs.
Technology Implementation ≠ Workforce Transformation
AI projects fail because they are workforce transformations that are managed like technology implementations.
- Work today. Human workers do tasks, often inefficiently and expensively. The promise of AI is to streamline these human tasks with agents.
- Agentification. Agents are built, but the effort is managed like a technology project — not the workforce transformation it actually is.
Every workforce transformation of the past — outsourcing, offshoring — focused on codifying and standardizing the work, the processes, and the definition of success, and the savings followed. The first generation of AI projects focused on the technology while ignoring those other factors. Many have failed as a result.
Define, Scale, Deploy, Amplify
Define how your organization works first. Then scale it — with people, with AI, or both.
Workforce and organizational strategies often remain document-based rather than operationalized. As enterprises adopt AI agents and autonomous workflows, these strategies need to become executable, measurable, and continuously managed.
- Work Today — scattered everywhere, fully understood nowhere: diagrams, wikis, emails, trackers, and people’s heads.
- Documentation — the deliberate step to document, structure, and model how work is done, moving it toward an asset.
- Optimization — as you document, you naturally improve the work: small changes on paper, disproportionate value.
- Standardization — work becomes more effective when workers share standardized processes, templates, and tools.
- Orchestration — clear standards enable efficient routing between workers, and easier monitoring against SLAs.
- Agentification — agents enter once the process is defined and worth automating. They scale structure; they don’t create it.
FTDL: the Technology Framework
FTDL (Flexible Task Description Language) is an AI specification language created by Fourier Partners for describing tasks in a workforce where humans and AI work side by side — the operational equivalent of SQL for data. It is AI SDK agnostic by design.
FTDL is a structured schema designed to capture and formalize actual work processes. By integrating human and agent workforces, it facilitates the definition, visualization, and orchestration of complex, hybrid workflows — a durable, living blueprint that remains effective as systems evolve.
- Consolidate processes. Brings emails, tribal knowledge, and outdated SOPs into a single, version-controlled layer that survives any system or organizational change.
- Codify and own. Transforms implicit process detail into a permanent, structured corporate asset, fully owned by the enterprise.
- Anchor automation & AI. Establishes a canonical description of work so automation is transparent, reliable, and continuously aligned with business goals.
FTDL lets you define tasks and the shape of the team independent of execution agents. It sets clear expectations for both humans and AI — and it works with OpenAI, Anthropic Claude, Google Gemini, CrewAI, LangChain, and any other agent framework.
Where to start
You don’t codify the whole company on day one. Start with one high-volume process where the work is understood but undocumented, capture it as Work as Code, and put it into production with humans in the lead. Each process you encode becomes a reusable building block for the next — so you start every subsequent effort with a blueprint, not a blank page.