What is deployment?
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.
Deployment is the act of putting a particular software version into an environment where it can run or be used. An environment is the collection of computers, settings and connected services supporting that version. Deployment can be for testing, demonstration or live use. The word alone does not establish that the product is ready for people to depend on it.
A developer may run an application on their own computer first. The team may then deploy it to a separate test environment so colleagues can try it. Later, a version may be deployed to production: the environment used for ongoing real work. These are different uses of the same general activity.
For an AI application, deployment can involve more than uploading code. The application may need access to a model service, stored documents and approved tools. The model is the trained component producing responses. Its configuration, instructions and information sources can affect behavior even when the visible interface has not changed.
Deployment also differs from merging code. A merge combines proposed changes into a shared line of development. That record may still need to be built, configured and placed into the target environment. Teams can automate those steps, but an automated connection between them does not erase the conceptual difference.
A useful deployment process identifies what version is being introduced, how it will be checked and what to do if it fails. A rollback restores an earlier version or configuration. Some changes, especially those affecting stored data or actions already taken, cannot be undone merely by restoring older software.
Readiness therefore includes more than seeing the homepage load. Can users sign in appropriately? Are permissions correct? Do important tasks complete? Who notices a failure? What happens to work already in progress? These questions become more consequential as real users and records are involved.
Putting a fictional-data dental dashboard on a website would deploy a teaching prototype. Connecting it to a clinic’s real patient records and making staff depend on its follow-up queue would be a different operating arrangement. That move needs verified connections, appropriate access, clear ownership and a way to continue work during failure. The word “deployed” does not tell us which arrangement exists. Ask where the software runs, who relies on it and what evidence supports that use.
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