Voiceflow

How do you create job requirements in an industry that’s constantly changing? That’s just one of the topics we’ll tackle next week in our AI Strategy to Blueprint session: building the skills for an AI-focused team. Ahead of the event, we asked leaders in Customer Experience, Product, Engineering and IT to share their greatest challenges in building an AI-focused team. Some of their answers: ➕ Understanding technical requirements and distribution of labor ➕ Prioritizing steps on how to become AI-focused team. What comes first, second, third and what's the level of lift and ROI for each? ➕ Data privacy considerations, especially GDPR ➕ Recruiting new talent vs developing talent in-house ➕ How to attract the right people and compensate them appropriately Join us next Thursday, July 31 as we tackle these and other questions on building an AI-focused team. RSVP at voiceflow . com/events!

The most successful companies using AI agents don’t stop when their agent goes into production. 🛑 The final boss is iteration and scale — if you're already on this phase, congratulations, it's hard to get here. For this phase, measuring success revolves around two key themes. The first is being able to make quick additions to your app, and the second is achieving return on investment with a low total cost of ownership. Here’s how we suggest structuring your KPIs to understand how you’re doing and strive to improve: 🎯 time to make simple updates 🎯 time to launch subsequent version 🎯 time for next team to launch product 🎯 cost per additional project 🎯 total cost of ownership Choosing the right platform is key to success in iteration and scale. 👀 If a new LLM comes out, can you easily incorporate it into an existing agent? 👀 If a new business unit wants to build an agent for a different use case, is it easy to add on and take advantage of pre-existing templates and workflows? 👀 Can you easily make simple updates, or is it a long back-and-forth with development? Many teams think only as far as their first steps when it comes to choosing an AI agent platform. The most successful have a long term vision for turning AI automation into a core product in their company, understanding that AI agents will be as important (if not moreso) than websites and apps in the future. #aiagents #aiautomation #enterpriseai #genai

So, you’ve built your first proof of concept. Now what? We often see teams stuck in what Denys Linkov calls “Proof of Concept Purgatory” - they’ve launched some tests, had some early positive signals, but never get the agent into users’ hands. In some cases, these teams feel daunted by the steps to get to production. That’s why we recommend breaking the process into manageable chunks, aligned to 6 KPIs: 🎯 time to connect production data 🎯 time to move between product environments 🎯 time to build in an evaluation suite 🎯 time to finalize ui 🎯 time to complete the risk & security assessment 🎯 time to launch feature At this stage, collaboration across roles and quick iteration are both to ensure alignment and keep things moving forward. #aiagents #aiautomation #enterpriseai #aiforteams

As management legend Peter Drucker put it, “If you can't measure it, you can't improve it.” Building and launching Gen AI experiences is a team sport, so we should be measuring how well our teams are formed and how they perform with KPIs. At the earliest stages of your AI project, we recommend focusing on 4 KPIs: 🎯 time to standup a cross functional team 🎯 time to clear problem statement 🎯 time for user interviews 🎯 time to first demo Starting out with these KPIs and shared expectations gives you a benchmark for future improvement, and guidelines to understand if your project is on the right track.

Tomorrow: The next episode of Making Bots is an AI travel concierge for Hotels.com! In Making Bots, Senior Conversation Design Advocate @Peter Isaacs walks through the agent build and business impact of the automation. This time, it’s an AI travel concierge that can recall user booking history and bring dream vacations to life. And this use case is not just for hotels - think of the AI concierge as a framework that could be used for product recommendations in e-commerce, personalized account management in SaaS, a summer camp suggestion engine for elementary school students… the possibilities are endless. We’ll be on with Pete starting at 4pm ET tomorrow. See you there?

🛠️ Introducing: The full family of Anthropic Claude models in the Voiceflow platform What does this mean for you? ✅ full access to the full Claude model family, including the latest Sonnet 3.5 ✅ Joint model and build expertise, combining the knowledge of Anthropic’s cutting-edge AI research with Voiceflow’s agent builder ✅ Use case-specific guidance: Our teams will be working closely to create reusable templates, guides and comparisons for customer support, revenue generation, internal automation and other core use cases. The Claude model family combines the best of speed and performance and is ideal for reasoning, vision analysis, content generation and multilingual processing. Teams building and deploying agents need the control and flexibility to use the right models for specific tasks and actions. Controlling costs and performance is only possible with powerful model options. We’re thrilled to collaborate with Anthropic to help teams automate complex interactions and get to an impactful POC that’s ready for production. #aiagents #aiautomation #anthropic #llms #conversationalai #aidevelopment

What does it *actually* take to get an AI agent into production? At UNPARSED last month, our Head of Machine Learning Denys Linkov sat on a panel around this topic, and the video just dropped on YouTube. Before you get started with any AI project, the panelists agreed you’ll need to meet these 3 requirements: 📈 get aligned on your KPIs - what are you measuring and how will you measure it? 💎 assemble the right team - “conversational AI is a team sport” ⚡ decide if you’re building toward perfection or minimum viable product (MVP) Head over to YouTube to watch the full discussion and learn more about what the most successful teams do to launch their AI agents into production and maintain their experiences.

Does your team have the skills to build AI products? Without a conscious effort and well-articulated game plan to upskill your team, your organization could be left behind - or eclipsed by competitors that do focus on AI. Building an AI-focused team requires a combination of mindset and practical resources. Over years of working with teams building and deploying AI products, we’ve seen first hand what makes teams successful and where others could learn from them. On July 31, Denys Linkov and Sam Burns host the next AI Blueprint event: Essential Skills to Build an AI-Focused Team. This live one-hour session will be packed with the information and strategies leaders need to build AI-focused teams. Join us to learn: ➡ How to build a productive team to get to success ➡ ROI on courses, webinars, conferences and more ➡ What to keep in house vs outsourcing externally ➡ How to craft job requirements ➡ Live review of sample AI engineering resume With practical examples, real world case studies, and follow-up resources, you’ll come away with a road map to upskill your team and ensure your organization stays at the forefront of AI automation. See you there! Link to register is in the comments, or head to voiceflow .com/events.

⚡ Introducing: Gemini Pro 1.5 in Voiceflow At Voiceflow, we’ve prioritized your ability to work with any LLM to develop the complex agent that best suits your use case. We are excited to announce the addition of Gemini Pro 1.5 to our platform. This is the first Google model we support, and brings new capabilities and improvements to enhance your agent performance. With Gemini Pro 1.5 you can expect: ✅ better accuracy with natural language processing tasks ✅ faster response times ✅ support for more languages, providing a better experience for multilingual users. Try out Gemini Pro 1.5 in Voiceflow today! #aiagents #googlegemini #conversationalai #aiautomation

AI Agents are ideal for recommendations and guiding users through a product 🛫 In the next Making Bots, join Peter to build an AI travel concierge for Hotels.com. We’ll work through API integrations, prompt optimizations and per user personalization to create an agent that can recall user booking history and bring dream vacations to life. Join this Making Bots for a peek into: ➕ GPT-4 and Claude Haiku LLMs ➕ Hallucination checking & disambiguation ➕ Per user memory with API integrations and transactional databases ➕ Vector Database and Retrieval Augmented Generation (RAG) for content search See you there (link in the comments)! ✈️