Worken templates bundle tested prompts, greetings, and model settings for seven major industries. Use them to spin up a virt in minutes without reinventing onboarding messages or qualification scripts.
Before you start
Go to Virts and open the card you want to configure. This is where you will apply template instructions, greeting, model settings, and knowledge files.
Use Templates by industry as the source text, then transfer the selected blocks into this virt.
Open AI → Instruction and paste the role/tasks/rules block from the industry template. Keep the structure explicit so operators can review and update it quickly.
Replace placeholders like [COMPANY NAME] and [REP NAME] right away so the scenario is launch-ready.
Switch to Chats → Basic settings and paste the greeting from the template. The first message should clearly state what the virt can help with.
If you support multiple brands or regions, tune wording per channel to keep promises and tone consistent.
Open AI → Model and apply template recommendations for vendor/model, temperature, Top P, and token limits. These controls directly affect tone, precision, and cost.
For stricter support flows, reduce temperature and response length to keep answers predictable.
Open Knowledge Bases, verify Integration connected, then create a base for this template. Upload pricing, FAQ, policy docs, and delivery rules.
A maintained knowledge base reduces hallucinations and keeps replies grounded in your actual business data.
Send 2-3 representative customer prompts in preview chat and verify replies. The test should show that the virt answers delivery/payment questions clearly and suggests next steps.
When results are stable, publish to production channels and assign an owner for ongoing prompt/knowledge updates.
WARNING
Strip confidential data and update placeholder text before launch. Also keep pricing and policy docs current in the connected knowledge base.
Example
A regional retailer grabs the “E-commerce” template, swaps the brand mentions, and keeps the recommended gpt-4o-mini model with temperature 0.8. After uploading the catalog PDF, the admin runs a short test covering delivery, returns, and upsell prompts, then publishes the widget on the main storefront.