In the AI landscape there are models whose code and weights are publicly available (open) and models accessible only through the maker's paid service (proprietary). For a small firm weighing up how to bring AI in, that distinction has concrete practical implications, not merely philosophical ones.
Proprietary models: operational simplicity
A proprietary model, reached through a paid interface, needs no infrastructure to manage: the business pays for use and receives a ready service, with updates and maintenance handled by the supplier. It is the simpler choice for anyone without in-house technical skills devoted to AI infrastructure.
Open models: control, and a different cost
An open model can be run on your own infrastructure, offering greater control over data (which never leaves for a third party) but requiring the technical skill and computing resources to run it, update it and keep it performing — a cost often underestimated against the subscription for a proprietary service.
What suits most small firms
For a business with no in-house team dedicated to AI infrastructure, a proprietary model reached through a service generally offers the best balance of running cost against complexity managed, provided you check the chosen supplier's privacy and data retention policies carefully.