Skip to main content

AI Assistant

Ask questions about your Workspace or build and test workflows in a chat panel that is always one click away.

The AI Assistant is a chat panel available on every page of the app. It answers questions about your Processes, Sessions, Deployments, and logs, and it can also make real changes for you: create a Process, edit a workflow definition, deploy a version, or run a test conversation against a deployed Environment. Open it with the robot icon in the top bar.

How it works​

The assistant is powered by a Large Language Model connected to the same Workspace data you see in the app. When you send a message, it decides which lookups or changes are needed, performs them on your behalf, and streams the answer back into the panel. It also knows which page you currently have open, so questions like "why did this Session fail?" work without pasting IDs.

Everything the assistant does runs with your own permissions it cannot see or change anything your Workspace role does not allow, and it never has access to Environment configuration values or credentials. If the assistant edits a Process you have open in the editor, the page refreshes automatically to show the change.

Each exchange is saved as a conversation, private to you within the current Workspace. You can revisit or continue any earlier conversation from the panel's history view.

What the assistant can do​

  • Build and edit Processes create a new Process (optionally from a template; new Processes get a Staging and a Production Environment by default), rename it, read and modify its workflow definition, save new versions with a commit message, and deploy to an Environment. It can also delete a Process if you ask.
  • Investigate Sessions list recent, failed, in-progress, or completed Sessions; filter by Process, Environment, or Channel; and drill into a Session's execution step by step.
  • Search logs filter error logs by severity, error code, Process, Environment, Session, or date range, and explain what went wrong.
  • Track Deployments and versions check Deployment status and history, and browse who committed which version and when.
  • Manage Environments list, create, rename, or delete a Process's Environments. It only ever sees Environment names, never their configuration.
  • Explore Modules browse the Modules available to your Workspace, inspect a Module's Tasks, and add a Module to a Process.
  • Summarize your Workspace Dashboard metrics, billing-period usage, collaborators, and integrations.
  • Test workflows start a test chat Session against a deployed Environment and exchange messages with the running workflow.

The first time you open the panel, a notice titled Before you start appears. It asks you to acknowledge, once, that:

  • Responses are AI-generated and may be inaccurate verify before acting on them.
  • Your prompts are sent to the LLM provider and stored on EW-FLOW's servers for up to 12 months.
  • You should not paste sensitive personal data, passwords, or special-category data (health, biometric, etc.) into prompts.
  • The assistant cannot exceed the permissions you already have in the Workspace.

Click I understand, continue to accept and start chatting, or I don't want to use this to decline declining closes the panel, and the notice reappears the next time you open it. Once accepted, the notice does not return unless its text is updated, in which case you are asked to accept the new version.

There is currently no in-app switch to withdraw consent after accepting. For consent or data questions, use the Privacy & Data link in the panel footer (which opens your privacy settings) or contact privacy@ew-flow.com.

Fields​

  • AI Assistant button the robot icon in the top bar of every page. Opens and closes the panel.
  • Usage badge a percentage next to the panel title showing the AI tokens used this period. Only shown when your Service Plan has an AI token allowance.
  • New conversation the plus icon. Clears the panel and starts a fresh conversation; the previous one stays in history.
  • History the clock icon. Lists your earlier conversations with their title and last-updated time; click one to reopen it.
  • Close the X icon. Closes the panel (clicking outside the panel also closes it).
  • Message input the Ask anything... box. Press Enter to send; press Shift+Enter for a new line.
  • Disclaimer the footer note "AI-generated. May be inaccurate." with a Privacy & Data link to your privacy settings.

Actions​

Ask your first question​

  1. Click the robot AI Assistant icon in the top bar.
  2. If the Before you start notice appears, read it and click I understand, continue.
  3. Type your question in the Ask anything... box and press Enter.
  4. The answer streams into the panel. The conversation is saved automatically.
tip

The assistant sees the page you have open. On a Session's detail page, "why did this fail?" is enough no need to paste the Session ID.

Continue an earlier conversation​

  1. Open the panel and click the clock icon to show your conversation history.
  2. Click a conversation to reload its messages.
  3. Type a new message it continues that same conversation.

Start a new conversation​

  1. Click the plus icon in the panel header.
  2. The panel clears; your previous conversation remains available in history.

Start a new conversation when you switch topics shorter conversations keep answers focused and use fewer AI tokens.

Example prompts​

  • "Show me the Sessions that failed today on the Customer Onboarding Process and explain what went wrong." the assistant finds the failed Sessions, reads the related error logs, and summarizes the cause.
  • "Create a new Process called Invoice Reminder from a template and deploy it to Staging." the assistant creates the Process, saves a version, and runs the Deployment, reporting each step.
  • "How many Sessions ran on the Production Environment this week, and how many had errors?" the assistant counts Sessions for the period and breaks down the failures.

AI token usage​

Every message consumes AI tokens the units the underlying language model uses to measure text it reads and writes. Your prompts, the assistant's replies, and the lookups it performs in between all count.

Usage is metered per Workspace against your Service Plan's AI token allowance and resets each billing period. Everyone in the Workspace draws from the same allowance it is the same counter shown on the AI Tokens card of the Dashboard, where you can see tokens used versus your limit and a forecast of when the allowance may run out. Inside the panel, the usage badge next to the title shows the same figure as a percentage.

warning

When the allowance is used up, the assistant replies with "AI token limit reached for this billing period. Please upgrade your plan or wait for the next billing cycle." and stops answering until then. See Billing & Plans for plan allowances.

Note that AI tokens are unrelated to the Token that moves through a Process during execution they only measure AI text processing.

Review AI changes before you rely on them​

The assistant's answers can be wrong, and its changes are real. Treat it like a fast but fallible collaborator:

  • When it edits a workflow, it saves a new version with a commit message open the Process in the editor and review the result before deploying it, especially to Production.
  • Verify factual answers (counts, error explanations) against the Sessions and Logs pages before making decisions on them.
  • Be explicit in prompts that change things: name the exact Process and Environment you mean.
danger

The assistant can delete Processes and Environments when you ask it to, with the same permissions you have. Deleting an Environment removes its configuration, and a deleted Environment can only be recovered by contacting customer support double-check names before requesting a deletion.

Best practices​

  1. One topic per conversation start a new conversation when you change subject; long mixed threads cost more AI tokens and blur the context.
  2. Name things exactly use the Process and Environment names as they appear in the app so the assistant acts on the right resources.
  3. Keep secrets out of prompts never paste passwords, API keys, or personal data; prompts are stored for up to 12 months.
  4. Review before deploying inspect any version the assistant saved before promoting it to a Production Environment.
  5. Watch the badge if the usage badge climbs quickly, check the Dashboard's AI Tokens card and plan an upgrade before the allowance runs out.
  • Dashboard the AI Tokens card that tracks your allowance for the billing period.
  • Billing & Plans what each Service Plan includes.
  • Processes versions and Deployments the assistant can create for you.
  • Sessions verify the Session details the assistant reports.
  • Logs the error logs the assistant searches when diagnosing failures.
  • MCP Access connect an external AI assistant such as claude.ai to your Workspace instead.