AI Action Assistant
AI Action Assistant is the operational AI layer inside FlexFiles. It understands natural language requests, resolves them against the current workspace, checks permissions and plan access, asks for clarification when needed, and runs allowed actions with confirmation for high-impact operations.
What AI Action Assistant is for
AI Action Assistant is designed for requests that should cause the product to do something, not just answer a question. It converts natural language into structured actions inside the active workspace and then runs only what the current user is allowed to run.
That makes it useful when the user already knows the desired outcome but does not want to navigate several screens manually. Instead of opening a folder, finding the correct control, and configuring the action step by step, the user can describe the intent directly.
Common jobs it can help with
- Create folders, rename items, move files, or prepare workspace structure.
- Create or manage guest upload links for a target folder.
- Run indexing or deindexing against the current workspace or a specific folder.
- Start file operations such as OCR, conversion, compression, flattening, or related processing where supported.
- Create or update workflows and explain why a plan, role, or limit blocks the requested action.
- Handle mixed requests that combine retrieval plus action when the underlying tools support that flow.
How it decides what to do
The assistant first interprets the request as intent plus entities rather than treating the whole sentence as one literal string. In practice that means it tries to separate the action type, the folder or file target, filters, follow-up references, quantities, and any modifiers such as password protection or date-based conditions.
It then resolves those names against the current workspace, applies permission and plan checks, and asks a short clarification question only when the target is genuinely missing or ambiguous.
- It prefers workspace names such as folder names over internal IDs.
- It uses recent assistant context for follow-ups such as another one, same folder, or do it again.
- It should not act outside the active workspace, even if the same resource name exists elsewhere.
Confirmation and safety model
Destructive, bulk, or permission-sensitive operations should not run silently. The assistant is expected to summarize what it understood, which resources will be affected, and what will happen before execution when the action has meaningful risk.
- Delete, revoke, role-change, and broad deindex requests should require explicit confirmation.
- If the request is low-confidence, the assistant should clarify instead of guessing.
- If the action is blocked by role or plan, the assistant should explain that constraint in plain product language.
It is most reliable when folder names, workflow boundaries, and permissions are already well-structured. Strong workspace hygiene improves assistant accuracy and reduces clarification loops.
Request patterns that work well
- Create a guest upload link in Candidate Resumes.
- Move all files from Inbox folder to Approved folder.
- Index the Contracts folder and alert me on new termination clauses.
- Create a new folder for current month invoices.
- Why can’t I deindex this folder?
- Create another guest upload link there with password protection.
Best practices
- Use the real folder or workflow name whenever you know it.
- Keep follow-up commands anchored to a recent successful action when possible.
- Ask for a dry-run style explanation first when the operation is broad or irreversible.
- Use AI Document Assistant instead when the goal is knowledge retrieval rather than workspace mutation.
Common AI Action Assistant questions
Does it act across every workspace I belong to?
No. It should stay scoped to the current active workspace and only use the permissions and plan access available in that context.
Why would it ask a clarification question instead of just doing the action?
Clarification is the correct behavior when a name resolves to multiple possible targets, when the target is missing, or when the phrasing is too ambiguous for a safe action.
When should I use AI Action Assistant instead of AI Document Assistant?
Use AI Action Assistant for operational commands such as create, move, link, index, run, update, or explain why an action is blocked. Use AI Document Assistant when the main job is asking questions against indexed content.
Keep exploring
Understand the indexed AI workspace model in FlexFiles and how to use it for retrieval, monitoring, and structured decision support.
Understand the file layer inside FlexFiles, including uploads, folders, metadata, and version-oriented workflows.
Create secure guest upload links to collect documents from clients, applicants, or partners without granting workspace access.
Understand how FlexFiles workflows automate multi-step document operations.