Harvard Data Science Review (HDSR) has published the Just Accepted version of a new article by Ulla Kruhse-Lehtonen and Dirk Hofmann of DAIN Studios: The Agent-Centric Enterprise: Why 2–10× Productivity Gains Demand Radical Workflow Redesign (published January 6, 2026). Just Accepted means the article has been peer-reviewed and accepted for its content, and it is now being copyedited to match HDSR’s final style and formatting. You can already read the accepted version here: Read the full article on HDSR.
The core argument is practical. Many organisations are currently adding GenAI to existing ways of working. That often improves individual tasks, but it rarely changes how work runs end-to-end. In the article, we argue that the larger productivity gains, in the 2–10× range, become possible when workflows are redesigned for agent execution, with humans supervising, handling exceptions, and improving the system over time. This is framed as potential when work is redesigned and measured, not as a promise that tools alone create outcomes.
To make the shift concrete, the article introduces Agent OS as a way to think about what has to change in the operating model when agents move from tools to active executors of work. It also presents the A.G.E.N.T. playbook, a delivery-oriented guide for redesigning one workflow into an agent-first version, with clear steps, controls, and measurable outcomes.
The article also includes a practical agent role taxonomy, assistant, analyst, tasker, orchestrator, and guardian, to help teams design agent-first workflows with clearer responsibilities and oversight. If you want the broader foundation behind that taxonomy and the A.G.E.N.T. playbook, click here.
The paper describes a focused two-month sprint approach: choose one workflow with real business value, redesign it for agent execution, implement with governance and human oversight, and track outcomes against clear measures. This is how teams turn agentic AI from experimentation into real operations: by delivering one redesigned workflow that works end-to-end, not a set of disconnected prompts and pilots.
This publication is closely connected to our ongoing collaboration with HDSR. Ulla and Dirk teach a module in the HDSR Agentic AI Intensive focused on applying agentic AI in organisations, with the same emphasis on workflow redesign, human supervision, and measurable impact.
We are pleased to see the article accepted and published. For us, it confirms that the practical focus we use in client work, redesigning workflows, setting clear human oversight, and measuring outcomes, is also what the wider HDSR practitioner audience values.