Prompt-driven content generation

Triggered by a request in a connected tool, the agent sends your approved prompt and the record's data to a Tess AI model, reads the generated text, and writes the draft back where your workflow expects it. It applies your rules to pick the template and tag the output. Drafts touching regulated claims, pricing, or customer-facing commitments are saved for review rather than published, so the agent notifies an owner and a person edits and approves the wording before anything generated by the model reaches an outside audience.

Workflow template runs

When an input arrives, the agent runs the matching Tess AI workflow template, feeding the record's fields as variables and reading each step's result. It applies your rules to store outputs on the source record and notify the requester when the run completes. Runs that return low-confidence results, hit a step error, or produce output conflicting with your validation rules are not accepted silently: the agent flags the run to a person with the details, so someone decides whether to retry, adjust the input, or handle it manually.

Model output storage

After a generation completes, the agent applies your rules to file the Tess AI output, whether a summary, reply, or classification, onto the correct record and related fields where connector permissions allow. It can attach the prompt used and a timestamp so the result is traceable later. Outputs meant for external delivery are held in a draft state and the owner is notified, since a person should confirm accuracy and tone before the content is treated as final and shared beyond the internal record with anyone outside.