Glossary
Fallback model, explained
A fallback model is an alternative AI model set up to take over responses automatically when the primary model becomes unavailable, slow or returns an error.
Model providers have outages, rate limits and version changes. Without a fallback, a provider failure becomes a pause in customer service.
The fallback can come from another provider or be a different version of the same model, and it should be tested with the same instructions and evals as the primary.
Logging when the switch happens helps explain differences in behavior and measure the operation's real availability.
Conversations aren't support.They're infrastructure.
Step into the new era of the relationship between companies and customers.