Skip to main content

​
Create Workflows or Procedures using plain English

Plain-English procedures power IrisAgent’s AI Agent and Agent Assist. The same procedures run for chat, voice, ticket auto-respond, and the copilot sidebar (Suggested Resolution and Dynamic Resolution). This is the fastest way to automate customer conversations and agent replies without building flowcharts or writing code.
Procedures take precedence over answers from your trained knowledge base. When a query matches a procedure’s scenario, the AI Agent and Agent Assist follow the procedure’s steps instead of generating a free-form answer from trained articles or past tickets. Use procedures whenever you want to enforce a specific flow (e.g. collect required fields, call an API, or hand off to a team).
You can configure standard operating procedures by describing them in plain English — also called SmartOps. Instead of designing a rigid decision tree, you simply describe the scenario and the steps the AI Agent should take, and IrisAgent trains a model to automatically recognize when the scenario applies and execute the workflow end-to-end. Common use cases:
  • Order status lookups: “When a customer asks about the status of their order, collect their order ID and look it up using the Order Status API.”
  • Refund and cancellation handling: “If a customer requests a refund, verify the order is within the 30-day window, confirm the reason, and create a refund ticket in Zendesk.”
  • Password reset and account unlocks: “When a user says they are locked out, verify their email, trigger a password reset link, and confirm once it has been sent.”
  • Subscription upgrades or downgrades: “If a customer wants to change their plan, confirm the new tier, check eligibility, and hand off to billing if manual approval is needed.”
  • Appointment scheduling: “When a customer asks to book a demo, collect their preferred date, time, and timezone, then create the calendar event.”
  • Troubleshooting known issues: “If the customer reports login errors on the mobile app, ask for their app version and OS, then walk them through the standard fix.”
To create a plain-English procedure:
  1. Go to Automate → Procedures on the IrisAgent dashboard and click Create New Procedure / Q&A.
  2. Enter the name, a short scenario description (when the AI Agent should trigger this procedure), and the actions the AI Agent should take (the steps, questions to ask, and any APIs or systems to call). Click on Save.
Write scenario descriptions the way you would brief a new agent — focus on the customer intent (“when a customer asks about…”) rather than exact phrasing. IrisAgent’s AI handles the wording variations, so you don’t need to list every possible way a customer might phrase the request.

​
Author the fields the AI should collect

The procedure editor has an Information to collect list. That is how you author the required fields the AI asks for, one at a time, before it calls an API or continues the steps.
  • Add, reorder, or remove fields. Labels become stable snake_case keys.
  • The list applies only to Instructions for AI Agent procedures (order ID, email, reason). A Direct Answer never interviews the customer, and the editor is hidden for that type.
Custom API placeholders like {{orderId}} are still collected automatically if you do not list them here. Authoring the list is how you control the questions and their order.

​
Scope a procedure to a product and a channel

On the same procedure:
  • Product ID (optional) pins the procedure to one or more products. Separate multiple product IDs with commas. When a case or chat has a product scope, IrisAgent serves procedures that share any of those product IDs, plus every unscoped procedure. A case or chat with no product runs every procedure, scoped or not. Leave the field empty to run for all products.
  • The Procedures table filters by channel: Chat, Email, and Voice, with per-channel counts. A procedure with no channel scope runs on every channel. Agent Assist uses the Email channel (the copilot sidebar is served as email). Unchecking Email to keep a procedure out of ticket auto-replies also removes it from the agent sidebar.
Ticket Agent can prefill the product. When you create a procedure from a Support Analyst recommendation, the Product ID field is filled from the source tickets so you do not retype it.

​
Create Zendesk tickets from a procedure

The built-in Create Zendesk ticket action creates a ticket in your connected Zendesk account using the information collected during the procedure. It is available automatically when Zendesk is your account’s active ticketing provider. You do not need to create a Custom API, configure an endpoint, or add separate credentials.

​
Add the ticket action

  1. Open or create a procedure under Automate → Procedures.
  2. In Availability, choose the channels that should use it. For inbound calls, select Voice and Inbound under Voice calls.
  3. Choose Instructions for AI Agent in the Action section. Do not use Direct Answer for a procedure that needs follow-up questions or action execution.
  4. Add the fields you need to Information to collect. For a callback request, for example, add Caller name, Caller email, Caller phone, and Issue summary. Add other business-specific fields as needed and put them in the order you want the AI to ask for them.
  5. Place the cursor at the appropriate step in your instructions. Under Actions → Connected actions, click Create Zendesk ticket. This inserts {{Create Zendesk ticket}} into the instructions.
  6. Describe when to run the action and what the AI should do after it succeeds. The action can be a step in the conversation; it is not limited to the end of a procedure.
  7. Set Status to Enabled and save the procedure before testing it.
The built-in action does not impose a fixed list of required fields. Your procedure defines what to collect. If the action is missing, ask an administrator to verify that Zendesk is the active ticketing provider for your account.

​
What goes into the ticket

Field labels are normalized to keys such as caller_email and issue_summary. Use one of the recognized labels above when you want a value to populate the requester or subject. Other collected values, such as account number or service address, are included in the description.
Add Caller email or another recognized email field if the ticket needs the customer’s email address. The action uses the email collected by the procedure; it does not substitute the connected administrator’s email when that value is missing.
Collected fields are not automatically mapped to Zendesk custom ticket fields, and the action does not automatically attach the full conversation transcript. If you need custom-field IDs or a different ticket payload, use a Custom API action instead. After a successful call, the action returns status: created and ticket_id. Later instructions can use that result to confirm creation and share the ticket number. Only tell the customer that a ticket was created after the action succeeds.

​
Example: collect a request for follow-up

Add Caller name, Caller email, Caller phone, and Issue summary to Information to collect, then use instructions such as:
To create a ticket only when a human is needed, put the action inside that condition in the instructions. If you also want a live handoff, run the ticket action first, then use Escalate to human. That button inserts the plain keyword escalate, without curly brackets.
Ticket creation and escalation are separate actions. Creating a ticket does not automatically transfer a call, and escalate does not automatically run Create Zendesk ticket. For phone handoff, also configure Voice AI call transfer.

​
Verify the ticket flow

Save the procedure with Status: Enabled, then start a conversation on a selected channel using the scenario you defined. Confirm that the AI collects the requested fields, runs the action at the intended step, and only reports success after creation. Open the resulting ticket in Zendesk and check its requester, subject, and description. A test that reaches this action creates a real ticket in the connected account; use appropriate test details.

​
Using Custom APIs in plain-English procedures

IrisAgent can call your own external APIs mid-conversation to look up live data — such as order status, account tier, or subscription details — and present the result to the customer in natural language, without any agent involvement. How it works:
  1. A customer asks a question that requires real-time data (e.g. “What is the status of my order?”).
  2. IrisAgent’s AI identifies the relevant workflow and determines what information it needs to collect (e.g. the order number).
  3. The AI asks the customer follow-up questions, one at a time, until it has all required fields.
  4. IrisAgent automatically calls your configured API with the collected data and the ticket’s context.
  5. The API response is handed back to the AI, which formats it into a natural-language answer for the customer.
Prerequisites: You must first configure the API under Custom APIs. During setup, any {{variableName}} placeholder you add to the URL or request body that is not a built-in ticket variable (e.g. {{orderId}}, {{email}}) is automatically treated as a field to collect from the customer. Example: A SmartOp for order status lookup might be described as:
“When a customer asks about the status of their order, collect their order ID and look it up using the Order Status API.”
IrisAgent maps “Order Status API” to the configured Custom API, collects {{orderId}} from the customer, calls the endpoint, and replies with the result — all without leaving the chat.

​
Create Workflows using drag and drop interface (only for cases)

This workflow builder applies to cases (tickets) only. For chat, voice, ticket auto-respond, and agent assist, use the plain-English procedures described above.
Agents and supervisors spend manual effort in triaging and resolving cases that may already be known or solved previously. IrisAgent solves this with automated case workflows built on AI-powered tagging, AI-based answers to known issues, and configurable actions for ticket assignment and notifications. Using the drag and drop trigger builder, you can combine any number of Conditions (such as AI category, AI sentiment, keywords in the subject or description, priority, channel, or customer attributes) with Actions (such as assigning to a team, adding tags, setting priority, posting an internal note, or sending a notification) — no code required. Common use cases:
  • Auto-tagging cases by AI category: Automatically apply tags like billing, refund, or outage when IrisAgent’s AI classifies an incoming case, so reports and views stay consistent without agent effort.
  • Routing based on AI categories or keywords: Send all cases tagged as billing to the Finance queue, or route any case mentioning “cancel my subscription” to a retention specialist.
  • Priority escalation: Automatically set priority to Urgent and notify a supervisor on Slack when a case is classified as an outage or contains churn-risk language.
  • Known-issue auto-response: When a case matches a known issue (e.g. a recurring bug), post IrisAgent’s AI-generated answer as a private note so the agent can reply in one click.
  • VIP handling: Route cases from enterprise or high-ARR customers to a dedicated team and add a “VIP” tag.
  • Sentiment-based escalation: Flag cases with negative sentiment for supervisor review or auto-assign them to a senior agent.
To create a new case workflow:
  1. Go to Automate → Case Triggers on the IrisAgent dashboard and click Create New Trigger.
  2. Drag and drop any number of Conditions (e.g. AI category equals billing, subject contains refund) and Actions (e.g. assign to team, add tag, set priority, send notification) into your trigger. Conditions are evaluated with AND logic, so the trigger fires only when all conditions are met.
  3. Give your trigger a name by clicking on the pencil icon.
  4. Click on the Add New Trigger button to save the trigger. The workflow will start running on new incoming cases immediately.
Start with a narrow condition set (for example, one AI category plus one keyword) and review the first few cases it fires on before broadening the rule. This helps you validate the behavior without mis-routing a large volume of cases.

​
Enable AI insights as a private note in your ticketing system


You can show IrisAgent’s AI insights as a private note on tickets so that it’s front and center for your agents.
  1. Go to Automate → Case Triggers on the IrisAgent dashboard.
  2. Enable or disable the toggle on the top for Private Note
  3. Click on the gear icon on the Private Note card to select/unselect the AI insights you’d like to show as a private note.
Feel free to email us if you encounter any issues or require assistance with product onboarding.