Voiceflow's AI Evolution: Native Integration & Prompt CMS

Updated on Aug 30,2025

Voiceflow has revolutionized its AI integration, shifting from isolated AI steps to a native, comprehensive system centered around a dedicated Prompt CMS. This update provides greater control, easier testing, and more seamless integration with other functionalities. Learn how to migrate and leverage these powerful new AI features for enhanced conversational agent design.

Key Points

Voiceflow is phasing out the 'Set AI' and 'Response AI' steps by June 3rd, 2025.

AI is now natively embedded throughout the Voiceflow platform, offering a more integrated experience.

The new Prompt CMS provides a dedicated home for prompt design, testing, and sharing.

The 'Prompt' step allows for more granular control over AI responses and seamless integration with other steps like Logic.

The Knowledge Base integration is significantly enhanced with the ability to inject conversation history and customize prompts for more context-aware responses.

The 'Set' step is being used in place of the legacy 'Set AI' step, leveraging prompts to set variables dynamically.

A prompts library is available for inspiration and quick access to pre-built prompts.

The Evolution of AI in Voiceflow

Moving Beyond Isolated AI Steps

Voiceflow has fundamentally changed how AI is integrated into its platform.

Instead of treating AI as a separate component with designated 'AI Steps,' the company has opted for a fully integrated approach where AI capabilities are woven into the very fabric of the design environment. This means that AI is no longer an isolated task but rather a readily accessible tool that empowers designers to build more sophisticated and context-aware conversational agents. This marks a shift from a step-based process to an environment where AI is an inherent part of the design process. This approach addresses the feedback Voiceflow received regarding the limitations of the original AI Steps. Users wanted more control over prompts, easier ways to test them, and a more seamless way to share them. They also wanted to integrate prompts natively with other steps, like the Logic step. Now, with the AI natively throughout Voiceflow, all that’s possible. Voiceflow is really leveling up the playing field in the conversational AI space, empowering designers with native AI integrations and powerful prompt management tools. The move towards native AI embedding directly addresses the limitations of the previous step-based approach, offering a more fluid and intuitive design experience. With this transition, Voiceflow is setting a new standard for how AI should be integrated into conversational AI platforms.

Introducing the Prompt CMS: A Central Hub for AI Prompts

At the heart of this AI revolution lies the Prompt CMS, a dedicated environment for crafting, testing, and managing AI prompts.

This new hub provides a centralized location for designers to iterate on their prompts, ensuring consistency and reusability across different agents and projects. The Prompt CMS streamlines the AI design process by providing a robust set of tools for prompt engineering. This central hub empowers designers to create, test, and manage prompts effectively, leading to more sophisticated and accurate AI-driven interactions. This ensures that designers can iterate effectively, maintain consistency, and leverage prompts across multiple projects. Designers gain access to tools such as conversation pair testing, message history injection, and latency checking, allowing them to fine-tune prompts for optimal performance and user experience. The Prompt CMS serves as a Knowledge Base, promoting knowledge sharing and collaboration among designers. The new Prompt CMS also offers enhanced testing features, including the ability to add conversation pairs to simulate interactions and inject message history at any point. This allows designers to create more realistic testing scenarios, enabling them to fine-tune their prompts for optimal performance. This approach also allows for latency checking, to keep the experience seamless.

The Sunset of AI Steps: What You Need to Know

As part of this transition, Voiceflow is phasing out the 'Set AI' and 'Response AI' steps. These legacy steps will officially stop working on June 3rd, 2025. While this change might require some adjustments, the new AI-native design and Prompt CMS offer a far more powerful and flexible approach to AI integration. Migration is key to ensure continued functionality and access to the latest features and improvements. Designers are encouraged to migrate their existing projects to take full advantage of the enhanced AI capabilities and avoid any disruption to their agents' performance. Voiceflow has introduced new tools that make migration straightforward. For designers familiar with 'Set AI' and 'Response AI' steps, there might be a slight learning curve adapting to the new workflow. However, the investment in learning the new system will be well worth it, as the AI native design and Prompt CMS provide a far more powerful and flexible approach to AI integration. Voiceflow ensures a smooth transition by providing comprehensive documentation, tutorials, and support to assist users during the migration process.

Voiceflow's Prompt Library

Accessing a Wealth of Pre-Built Prompts

Voiceflow provides an online Prompt Library, for inspiration and quick access to pre-built prompts. The Prompts Library is found at link.voiceflow.com/prompts-library. You are able to download the prompts to use anywhere in your project.

Migrating to the New AI Features: A Step-by-Step Guide

Replacing 'Response AI' with the 'Prompt' Step

If you've been using the 'Response AI' step, the 'Prompt' step will now be your go-to for generating AI-driven responses.

The Prompt step gives you much more granular control over your agent's replies, allows you to customize prompt instructions, and integrates seamlessly with the Knowledge Base. In the workflow, you can add a Prompt step. Now, in the prompt step you can select an existing prompt or create a new prompt. You can design a prompt, test it, and check latency. You can add conversation pairs and inject message history at a certain point. This ensures a more robust, tested, and integrated AI Tool. The Prompt CMS is the place where you can do all of this. Now, instead of simply typing something and hoping it works, you can craft precisely worded messages, simulate full conversation pairs, and inject conversation history as desired.

Leveraging the Knowledge Base with Enhanced Prompt Control

The Knowledge Base remains a powerful asset, but with the new AI integrations, you now have greater control over how your agent uses it.

You can inject the last utterance or other variables into your question to give it a frame of reference. Rather than relying on the older Response AI step for Knowledge Base lookups, Voiceflow now encourages using the new KB search under the Develop Section. Instead of the AI Model / Knowledge Base selection previously available, Voiceflow now is using a KB Search Tool under the Develop menu. This KB search tool is to be used with the Prompt CMS, offering even more creative power. This will get your question just like previously. After the question is answered, you'll have a variable that contains the KB results, which are now in chunks. You are then encouraged to use the Prompt step, the step hidden under the Talk menu, which is so incredibly powerful. Now you can either select an existing prompt, or build an entirely new prompt. The system provides some instructions at the top of the prompt to guide you. You are really empowering the user to get more from their knowledge base.

Replacing 'Set AI' with the Versatile 'Set' Step

The 'Set AI' step is being replaced by the 'Set' step, a tool you're likely already familiar with. Now, whatever it is you are trying to set, you can use the set step. Variables can be set using everything that you have always done, but now you can also use a prompt to set a variable. And that’s pretty cool, too. With a prompt selected, you can apply the output to a variable, even creating an entirely new variable as you see fit. This enables you to use this exact same pattern with the condition step. This lets you run a prompt, detect what’s coming out of it, and do an action based on that.

Voiceflow Pricing

Understanding Voiceflow's Tiered Plans

Voiceflow offers a tiered pricing structure to accommodate various project sizes and needs. While specific pricing details are subject to change, understanding the general structure can help you choose the right plan:

  • Free Plan: Ideal for beginners and small projects, offering limited features and usage.
  • Pro Plan: Designed for growing projects, providing expanded features and higher usage limits.
  • Business Plan: Tailored for larger organizations, offering advanced features, dedicated support, and custom solutions.
  • Enterprise Plan: Aimed at very large companies with custom needs.

Pros and Cons of Voiceflow's AI Integration

👍 Pros

Natively integrated AI provides seamless workflow and enhanced design experience.

Prompt CMS offers centralized prompt management, testing, and sharing.

Increased control over AI prompts allows for more customized and accurate responses.

Enhanced Knowledge Base integration enables context-aware and personalized interactions.

Transition to 'Set' step provides versatility and familiarity for setting variables dynamically.

The Prompt CMS offers enhanced testing features, including the ability to add conversation pairs to simulate interactions and inject message history at any point. This allows designers to create more realistic testing scenarios, enabling them to fine-tune their prompts for optimal performance.

Comprehensive documentation and support ensure a smooth migration process.

👎 Cons

Legacy AI steps are being phased out, requiring designers to adapt to the new system.

Some users may experience a learning curve adapting to the Prompt CMS and AI-native design.

Specific pricing details and usage limits may vary depending on the chosen plan.

Migration from the old steps is a required action.

Voiceflow's Core Features: AI-Powered Conversational Design

Unlocking the Potential of Voiceflow's AI Integration

Voiceflow is a platform that allows you to build conversational experiences, like chatbots and Voice Assistants, without any coding. Voiceflow acts as the brain of your conversational agent, handling the logic and storing the user’s conversation state. Voiceflow is all about extensibility and customization. Voiceflow has some out-of-the-box integrations and Developer Tools and APIs to connect Voiceflow to your existing tools and workflows.

Here is a table of Core Features:

Feature Description
Visual Canvas Design and build conversational flows with a drag-and-drop interface.
NLU Integration Integrate with leading NLU engines like Dialogflow, Lex, and Rasa.
Code-Free Logic Implement complex logic and conditional branching without writing code.
Integrations Connect to various APIs, databases, and third-party services.
Voice & Chat Build experiences for both voice and chat platforms.
Testing & Analytics Test your agent in real-time and track key metrics to optimize performance.
Prompt CMS A dedicated environment for crafting, testing, and managing AI prompts, ensuring consistency and reusability across different agents.
Prompt Step Generates AI-driven responses based on user inputs. Customize prompt instructions for optimal performance.

Use Cases for Voiceflow's Enhanced AI Capabilities

Real-World Applications of AI-Powered Conversational Agents

The new Voiceflow integrations provides great power in a huge variety of use cases. The system is very flexible, allowing the system to provide value across a wide variety of industries.

  • Customer Support: Automate responses to frequently asked questions, provide personalized assistance, and resolve customer issues efficiently.
  • Lead Generation: Engage with potential customers, collect valuable information, and qualify leads for sales teams.
  • E-commerce: Guide customers through product selection, process orders, and provide shipping updates.
  • Education: Deliver personalized learning experiences, answer student questions, and provide feedback on assignments.
  • Healthcare: Schedule appointments, provide medication reminders, and offer remote patient monitoring.

Frequently Asked Questions

When will the AI Steps stop working?
The 'Set AI' and 'Response AI' steps will officially stop working on June 3rd, 2025. You should migrate your projects to the new AI-native design and Prompt CMS before this date to avoid any disruption to your agents' performance.
Where can I find inspiration for new prompts?
You can find inspiration and pre-built prompts in the Prompt Library. The Prompt Library is downloadable and useable anywhere in your project.
How do I test the Prompt CMS?
You can add conversation pairs and inject message history at any point to test and simulate the full user experience.
Is migration required?
Migration from the old steps is a required action if you want to continue building agents.

Related Questions

What is Prompt Engineering and why is it important?
Prompt engineering is the process of crafting effective prompts to elicit desired responses from AI models. It involves carefully designing the wording, structure, and context of the prompt to guide the AI model toward generating relevant, accurate, and creative outputs. Effective prompt engineering is crucial for maximizing the potential of AI models in various applications. The role of prompt engineering in conversational AI is paramount. Well-engineered prompts ensure that AI agents provide accurate, relevant, and engaging responses, leading to more natural and satisfying user interactions. Ineffective prompts, on the other hand, can result in confusing, irrelevant, or even nonsensical outputs, undermining the user experience. The quality of the prompts directly impacts the performance and reliability of conversational AI systems, making prompt engineering a critical skill for designers and developers. It is the key to unlocking the potential of AI and delivering truly valuable conversational experiences. Best Practices for Writing Effective Prompts: Be Clear and Specific: Define the desired outcome and provide sufficient context for the AI model to understand the task. Use Keywords Strategically: Incorporate relevant keywords to guide the AI model toward specific topics or concepts. Provide Examples: Include example inputs and outputs to demonstrate the expected behavior. Iterate and Test: Experiment with different prompt variations and evaluate the results to optimize performance. Consider the Target Audience: Tailor the language and tone of the prompt to the intended users. Inject Message History When Necessary Use previous prompts to give the model context and help it determine an adequate response.

Most people like