Leadership in the age of AI agents

Two Thoughtworks consultants argue that managers will audit decisions shared with AI agents, and that organisations must record what each agent may decide.

Kaushik Sarkar and Muralikrishnan Puthanveedu write on the blog of Thoughtworks, a software consultancy. They open with a sales manager whose AI agent drafted plans, emailed customers and flagged a pricing error overnight without being asked. They argue that leadership models built for all-human teams need new rules for roles, structure, decision rights and accountability.

• Middle managers would move from supervising people's output to auditing decisions that people and AI agents make together.
• The authors propose treating agents as a separate class of asset, with a register of decision rights, escalation routes and audit trails.
• Citing research from MIT, they split decisions by ambiguity and risk: people frame problems, agents act on low-risk tasks, and learning from outcomes is shared.
• Leaders must define what an agent may recommend, decide and carry out, and when authority returns to a person.

A team that runs an AI agent can write down its decision limits before launch and name the person who answers for its outcomes. The piece is an essay without client evidence, and the page does not state the authors' roles.

Derived by Working Surface from the article; more in the Playbook. Source line: Organizations that simply shrink the layer without redefining its function will lose the oversight capacity they didn't know they needed until an agent twin makes a costly autonomous call nobody was watching for.

10 October 2026: as AI agents start work without being asked, the rules that limit them need to be written down and owned: by a person, a team and the machine itself.