Atomic Work Research

Research

Our research explores how AI and human judgement can work together to eliminate the coordination tax that costs organisations billions every year.

May 2026·9 min read

Why Atomic Workflows Outperform Linear Automation

Most automation tools treat business processes as simple trigger-action chains. This model breaks down at the edges where real work happens: conditional branches, multi-party approvals, retry logic, and compliance requirements. We examine why typed atomic workflows — where each step carries a semantic contract — outperform linear automation at scale.

Read paper
March 2026·8 min read

The $400B Coordination Tax: How Teams Lose Time

Coordination tax is the hidden cost of moving information and decisions between people, systems, and processes. It manifests as approval delays, status update meetings, manual data entry, and handoff failures. We quantify this cost, map where it accumulates, and show how structured workflow automation eliminates it without removing human judgement from the processes that require it.

Read paper
January 2026·10 min read

Human-in-the-Loop: The Missing Layer in Enterprise AI

Enterprise AI deployments have concentrated on replacing human tasks. The next wave requires something more nuanced: AI that knows precisely when to pause and involve a human, route to the right person, enforce a response deadline, and log the decision permanently. We introduce a framework for structuring human decision points within automated workflows — and argue that human-in-the-loop is not a compromise but a competitive advantage.

Read paper
November 2025·12 min read

36 Atomic Actions: A Complete Taxonomy of Business Operations

Every business process, regardless of industry or complexity, decomposes into a finite set of primitive operations. After analysing thousands of workflows across NDIS compliance, MSP operations, property management, and HR automation, we identified 36 distinct atomic action types that together cover the complete space of business operations. This paper presents the taxonomy, explains the design principles behind it, and demonstrates how the 36 atoms compose into arbitrarily complex real-world workflows.

Read paper