Fortra said Nov. 17, 2026, that selected customers are testing an AI module for GoAnywhere MFT and Globalscape EFT. Its announcement places general availability in early 2027, with migration capabilities planned for later that year. The company says administrators will review proposed changes before applying them. These are company-described plans and controls, not evidence of measured improvements in production.

Rapid News Brief: Existing projects

GoAnywhere’s Projects tutorial describes the existing work an administrator configures. Its example retrieves database records, writes a spreadsheet and sends the resulting file to a trading partner through SFTP. A project is therefore more than a file’s destination: it defines a sequence of tasks, along with decisions, loops and variables that determine execution.

The guide describes a visual outline where tasks can be rearranged and a separate pane where their attributes are configured. Task labels can be changed. Variables carry values such as file lists or database results between steps. It also describes testing with debug-level logging, then reducing that level once the project behaves as intended. After execution, administrators can inspect or download a job log. These are documented product mechanics, not newly announced AI capabilities.

Dependencies survive until they fail

Globalscape’s Advanced Workflows manual, marked version 10 with a 2022 date code, describes a separate dependency. An Event Rule can invoke a named workflow; removing that workflow does not stop the containing rule from running. Instead, the rule fails when it reaches the deleted workflow.

The manual also describes supplying custom parameters when selecting a workflow, with names and values entered separately. A value can refer to an EFT variable, while the parameter’s name cannot contain one. Administrators can add further actions before saving changes to the server. Workflow names can appear in reports and logs. This older documentation explains operational review requirements; it does not establish that the new AI module detects or repairs those dependencies.