> Markdown rendition of https://www.voiceflow.com/platform/rehearsal ("Rehearsal | Voiceflow"). Canonical page: https://www.voiceflow.com/platform/rehearsal · All pages: https://www.voiceflow.com/llms.txt Rehearsal # Test the change. Trust the release. Voiceflow’s automated agent testing product combines AI simulations, behavioral checks and regression tests to catch problems before customers do. [Get started](https://creator.voiceflow.com/start) [Talk to sales](https://www.voiceflow.com/demo) ## Better customer experiences, at scale. **Trusted by 3,000+** customers, with **1 billion+** agent messages processed. > “Voiceflow takes the pain out of building powerful conversational experiences in a group setting.” > “There’s only one real tool that gets the job done for me and that’s Voiceflow.” [“The engineering experience was smooth — we could quickly move from a workflow designed by Voiceflow to one that we owned and could refactor ourselves.”](https://www.voiceflow.com/stories/how-turo-built-chatbot-in-two-months) ** [45%**Offer-to-leadrate](https://www.voiceflow.com/stories/how-sanlam-studios-ai-coach-drives-leads-and-financial-literacy) > “Voiceflow is the A-Z of learning and executing on AI Automation.” ** [Under 12 weeks**From concept to live deployment“It allowed us to rapidly iterate on conversational flows and reduce dependency on engineers.”](https://www.voiceflow.com/stories/stubhub-internationals-90-day-journey) > “With Voiceflow, it’s much easier to navigate our designs and test conversations.” ** [70%**AI resolutionrate](https://www.voiceflow.com/stories/automating-60-of-customer-support-for-90-products-in-12-weeks-how-ai-automation-transformed-trilogy) > “With Voiceflow we're building interactive prototypes as we design. This is the dream.” > “Voiceflow takes the pain out of building powerful conversational experiences in a group setting.” > “There’s only one real tool that gets the job done for me and that’s Voiceflow.” [“The engineering experience was smooth — we could quickly move from a workflow designed by Voiceflow to one that we owned and could refactor ourselves.”](https://www.voiceflow.com/stories/how-turo-built-chatbot-in-two-months) ** [45%**Offer-to-leadrate](https://www.voiceflow.com/stories/how-sanlam-studios-ai-coach-drives-leads-and-financial-literacy) > “Voiceflow is the A-Z of learning and executing on AI Automation.” ** [Under 12 weeks**From concept to live deployment“It allowed us to rapidly iterate on conversational flows and reduce dependency on engineers.”](https://www.voiceflow.com/stories/stubhub-internationals-90-day-journey) > “With Voiceflow, it’s much easier to navigate our designs and test conversations.” ** [70%**AI resolutionrate](https://www.voiceflow.com/stories/automating-60-of-customer-support-for-90-products-in-12-weeks-how-ai-automation-transformed-trilogy) > “With Voiceflow we're building interactive prototypes as we design. This is the dream.” > “Voiceflow takes the pain out of building powerful conversational experiences in a group setting.” > “There’s only one real tool that gets the job done for me and that’s Voiceflow.” [“The engineering experience was smooth — we could quickly move from a workflow designed by Voiceflow to one that we owned and could refactor ourselves.”](https://www.voiceflow.com/stories/how-turo-built-chatbot-in-two-months) ** [45%**Offer-to-leadrate](https://www.voiceflow.com/stories/how-sanlam-studios-ai-coach-drives-leads-and-financial-literacy) > “Voiceflow is the A-Z of learning and executing on AI Automation.” ** [Under 12 weeks**From concept to live deployment“It allowed us to rapidly iterate on conversational flows and reduce dependency on engineers.”](https://www.voiceflow.com/stories/stubhub-internationals-90-day-journey) > “With Voiceflow, it’s much easier to navigate our designs and test conversations.” ** [70%**AI resolutionrate](https://www.voiceflow.com/stories/automating-60-of-customer-support-for-90-products-in-12-weeks-how-ai-automation-transformed-trilogy) > “With Voiceflow we're building interactive prototypes as we design. This is the dream.” ## Define the behavior. Then put it to the test. Simulate customer scenarios, check the outcomes and investigate failed runs. SimulationsBehavior checksPersonasPast runs ## Explore conversations you haven’t scripted. Let AI play the customer. Define the scenario and success criteria, then test how your agent responds. Describe the customer scenario Tell the simulated customer what they want and how to behave. Explore follow-ups and exceptions without scripting every message. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#creating-a-test) Define a successful outcome Set the criteria the conversation must meet, such as explaining a policy and offering a handoff without issuing a refund. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#creating-a-test) Set the conversation limit Choose how many turns the simulation can use to reach your success criteria. Keep each test focused and bounded. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#creating-a-test) Combine simulated and scripted turns Add an AI simulation alongside exact user messages and real agent replies to control some parts of a test and explore others. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#creating-a-test) ## Test more than the final answer. Check the agent’s routing and actions, including what must never happen. Verify the route Check that a turn reaches the intended playbook or workflow, such as the returns process for a refund request. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#adding-checks-to-an-agent-turn) Assert required tool calls Confirm that the agent calls the tools the task requires, including looking up an order before issuing a refund. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#adding-checks-to-an-agent-turn) Prevent unwanted actions Set a routing or tool check to No when an action must not happen, such as issuing a refund without approval. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#adding-checks-to-an-agent-turn) Choose how checks run Run checks independently or sequentially when they depend on shared context. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#adding-checks-to-an-agent-turn) ## Start from the customer’s context. Save starting variable values and reuse them across customer scenarios. Set starting variables Give the test concrete customer data, such as an order ID, so the agent starts from a known state. [Learn more](https://www.voiceflow.com/docs/documentation/build/data/personas#how-personas-work) Combine context values Add the variables your agent expects to model an account or customer scenario without rewriting the conversation. [Learn more](https://www.voiceflow.com/docs/documentation/build/data/personas#how-personas-work) Reuse a saved persona Apply the same customer context across tests and repeat runs without entering its values again. [Learn more](https://www.voiceflow.com/docs/documentation/build/data/personas#how-personas-work) Reset the starting context Remove the selected persona when you want to test the conversation without those saved starting values. [Learn more](https://www.voiceflow.com/docs/documentation/build/data/personas#how-personas-work) ## Find the failure. Understand the cause. Review saved results, inspect failed checks and trace the actions behind them. Compare saved runs See passed and failed runs in the history table, including who ran them and when. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#reviewing-past-runs) Find the failing check Open a saved result to see which check failed within the conversation. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#reviewing-past-runs) Read the failure reasoning Use the explanation attached to a failed check to distinguish a wrong route from a missing tool call. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#reviewing-past-runs) Inspect the execution logs Open the logs from the result to trace the actions behind a failure before changing the agent. [Learn more](https://www.voiceflow.com/docs/documentation/measure/tests#running-tests) Feature highlights ## Test the conversation and the actions behind it. Combine scripted conversations and AI simulations with checks on responses, routing and tool calls. Catch failures before customers experience them. ### **Conversation tests** cover customer scenarios Tests & simulations**Returns agent** **Eligible return**Issue refund Queued Running Passed **Outside return window**Explain policy Queued Running Passed **Damaged item**Escalate to team Queued Running Passed ### **Behavior checks** verify responses, routes and tools ![Routing checks in an existing Voiceflow documentation capture. Demonstration values are illustrative.](https://www.voiceflow.com/docs/images/Tests-add-check.png) Build together ## Make quality a shared responsibility. Bring customer scenarios, business rules and technical changes into the same testing process. SupportCX DesignProductEngineeringQALeadership ### **Support** brings the real-world edge cases. Use customer expertise to identify the questions, exceptions and handoffs your tests need to cover. Playbooks ### **CX Design** checks the experience before release. Refine instructions in Studio, then test them in Rehearsal. Use scripted turns and response checks to verify tone, clarity and the next step. Agent Studio ![Pine customer guide prompt with rich headings, voice principles, guardrails, and example dialogue in Voiceflow](https://www.voiceflow.com/images/team-workflow/cx-design-main.png) ### **Product** prioritizes what must keep working. Choose the customer journeys each change must protect. Ask Atlas to turn real conversations into Rehearsal tests, then review the results with the team. Atlas ### **Engineering** checks what changes affect. Test tool calls and routing alongside responses before merging an updated environment. Functions ### **QA** turns expectations into checks. Write scenarios, inspect failed runs and verify fixes. Review saved test conversations in Prism when deeper investigation is needed. Transcripts ![Voiceflow transcripts table with resolution, quality, policy compliance, and sentiment evaluations for illustrative customer support conversations](https://www.voiceflow.com/images/team-workflow/qa-main.png) ### **Leadership** sets the standard for release. Use service outcomes in Prism to identify business risks. Align the team on what Rehearsal must test, and review the results before expanding a rollout. Analytics ![Creator Leadership workspace with fictional Pine support data](https://www.voiceflow.com/images/team-workflow/leadership-main.png) Developer capabilities ## Test the integrations your agent depends on. Verify tool calls in a controlled conversation, then investigate the underlying request or function when a check fails. Onboard agent Set up Voiceflow by following https://www.voiceflow.com/agents/setup [Docs](https://www.voiceflow.com/docs/documentation/measure/tests#adding-checks-to-an-agent-turn) ### **API requests** can be tested before use ### **Function tools** belong in your behavior checks Returned result`{ "order_status": "shipped", "tracking_url": "…/track/1042" }` Customer stories ## Better customer experiences, at scale. See how teams build, refine and launch customer experiences with Voiceflow. ←→ [70% AI resolution rate.Read story →](https://www.voiceflow.com/stories/automating-60-of-customer-support-for-90-products-in-12-weeks-how-ai-automation-transformed-trilogy) [![eSnipe](https://www.voiceflow.com/images/esnipe-cover-logo.png)70% help center tickets automated.Read story →](https://www.voiceflow.com/stories/how-esnipe-automated-70-of-their-help-center-tickets-with-an-ai-powered-search-agent) [82% bot satisfaction.Read story →](https://www.voiceflow.com/stories/how-turo-built-chatbot-in-two-months) [45% offer-to-lead conversion.Read story →](https://www.voiceflow.com/stories/how-sanlam-studios-ai-coach-drives-leads-and-financial-literacy) [30 support hours saved per week.Read story →](https://www.voiceflow.com/stories/how-roam-saved-30-hours-a-week-in-customer-support-hours) [30% project time saved.Read story →](https://www.voiceflow.com/stories/flowcharts-to-voiceflow-how-wooliesx-reinvented-their-conversation-design-workflow) [![BMW](https://www.voiceflow.com/images/bmw-cover-logo-v2.svg)50% reduction in prototyping time.Read story →](https://www.voiceflow.com/stories/how-bmw-reduced-prototyping-time-of-their-in-car-voice-apps-by-50) [90 days from idea to launch.Read story →](https://www.voiceflow.com/stories/stubhub-internationals-90-day-journey) [How Rocket Mortgage Accelerated their Team’s Conversation Design WorkflowRead story →](https://www.voiceflow.com/stories/how-rocket-mortgage-accelerated-their-teams-conversation-design-workflow) [How Chase Uses Voiceflow to Run More Efficient, Inclusive Conversational AI Design ReviewsRead story →](https://www.voiceflow.com/stories/using-voiceflow-to-run-more-efficient-inclusive-ai-design-reviews) FAQ ## Frequently asked questions **What are AI simulations?** AI simulations let an AI play the customer through part of a test conversation. Define the scenario, success criteria and maximum turns. The simulation passes when it reaches those criteria within the turn limit. **Do tests run automatically when I edit an agent?** No. Start a single test or a batch yourself. Automated here refers to simulated conversations and checks; it does not mean every edit triggers a test run. **What can a test verify?** Check the response with exact matching or an LLM judge, verify routing to a playbook or workflow, and check that the expected tool was called. **Can I test different customer contexts?** Yes. Personas supply saved starting variable values, and a launch event can change how a test starts. Simulations let an AI play the user toward success criteria within a turn limit. **Do test runs use credits?** LLM-powered activity in tests consumes credits, as it does in a real conversation. ## Explore the platform ### [![Studio](https://www.voiceflow.com/og/products/studio.png)StudioBuild and edit AI agents together, with deep control over how they work.Explore Studio](https://www.voiceflow.com/platform/studio) ### [![Atlas](https://www.voiceflow.com/og/products/atlas.png)AtlasMeet Atlas, your copilot for building and improving AI agents.Explore Atlas](https://www.voiceflow.com/platform/atlas) ### [![Prism](https://www.voiceflow.com/og/products/prism.png)PrismUnderstand agent performance and investigate what happens in every conversation.Explore Prism](https://www.voiceflow.com/platform/prism)