What is a workflow?
The Harvard Chan AI Glossary
In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.
A workflow is the sequence of work, information exchanges, decisions and handoffs through which people accomplish a task. It includes both what software does and what people must do around it. Waiting, searching for information and correcting mistakes are part of the workflow even when they are not shown in a formal process diagram.
Consider arranging a meeting. Someone identifies the participants, checks availability, resolves conflicts, chooses a time and confirms the invitation. The visible calendar entry is only one output. Much of the work concerns negotiation and uncertainty before that entry exists.
A prescribed workflow describes how the process is supposed to operate. Actual practice may differ. People may keep private spreadsheets, bypass a slow step or ask a colleague for help. Under pressure, the process may change again. Look at prescribed work, ordinary practice, work under strain and recovery after failure.
This matters when introducing AI. Automating the preparation of an invitation may save little time if the real difficulty is deciding whose availability takes priority. Conversely, helping people assemble accurate information may be valuable even when a person still makes the final decision.
An agentic workflow is a particular arrangement in which an AI agent can choose or adapt some steps. An agent is software that pursues a goal through actions and feedback. Not every workflow needs an agent. A fixed rule or a simpler change in responsibility may solve the problem with less complexity.
A user journey focuses on one person’s experience from need to resolution. A workflow can span several roles and systems. For example, an expense process involves the employee, approver and finance team. Improving one participant’s step may transfer work to another, so the complete process needs attention.
In a specialty clinic, referral intake may involve requesting records, checking administrative prerequisites, arranging clinical review and coordinating a first appointment. The formal diagram may omit the phone calls and informal reminders that keep cases moving. An agent that summarizes documents faster will not necessarily resolve an authorization delay or an unclear handoff. Map who owns each next step, what evidence permits it and how completion is confirmed. Then decide which parts need fixed rules, agent assistance or human judgment.
If you would like to find out more about AI and improve your leadership skills, consider the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.
AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.
For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.
Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy