Our AI Assistant for Photobooth AI

For some time now at Premier Octet, we’ve been building conversational agents tailored for our clients. With this expertise, we decided to create a dedicated agent for Photobooth: our AI-powered photo generation app.

In this article, I’ll share how we developed this conversational agent, and why we chose this approach to handle part of our customer relations.

Context

Photobooth AI is our app that lets users generate personalized photos using artificial intelligence. Like any SaaS application, it requires effective customer support to guide users as they discover and use the platform.

Until recently, we used Crisp as our traditional chat support solution. While this worked, it had some limitations: restricted availability during business hours, repeated answering of common questions, and a heavy support load for our team for very similar requests.

We also noticed that some users tried to use the support chat directly to generate their images—uploading photos and writing prompts, expecting to get their creations through this channel. While this reflected their desire for an easy way to interact with our service, it led to dead ends and frustration.

Recognizing this, and leveraging our expertise in building conversational agents, we decided to develop a specialized AI assistant for Photobooth AI. Our goal? To optimize customer support and improve the overall user experience.

Automated and Personalized Support: Our Strategy

The main goal of this integration is to deliver automated, personalized support tailored for our platform. Unlike generic solutions, our AI assistant is configured with:

• Our full FAQ content: All frequent questions and their official answers
• The site map: To accurately guide users to the right sections
• A specialized system prompt: Crafted specifically for Photobooth AI’s challenges
• An integrated feedback tool: Automatically forwarding complex issues to human support

This approach lets us maintain high-quality support while automating much of the customer interaction. In short, it’s a system that deeply understands our product and can efficiently guide our users.

The Benefits of This Approach

Let’s dig into the concrete benefits of this implementation.

1. 24/7 Availability

The first benefit is obvious, but essential: continuous availability. Our assistant responds instantly to user queries, whether it’s 2 a.m. or a weekend. This 24/7 availability dramatically enhances the user experience and has a positive impact on conversions.

This international dimension is especially important for Photobooth AI, since our users come from around the world. Time zone differences made it difficult to offer traditional support during French business hours.

This automation greatly reduces the burden on our team, who can now focus on handling the complex issues that require human intervention. In other words: less time spent on repeat questions, more energy for real improvements and problem-solving.

2. Built-in, Always-Up-to-Date Product Expertise

Our assistant has deep, always up-to-date knowledge of everything related to Photobooth AI: credit system, event creation, style management, pricing, and more. This ensures consistent and accurate responses, always based on our official documentation.

3. A Better User Experience

Our goal for the user experience is clear:

• Instant answers: No more waiting for basic information
• Guided navigation: The assistant smartly directs users to the right areas of the app
• Multilingual support: The model we use responds in most of the languages spoken by our users

4. Operational Cost Optimization

Automation means a significant reduction in support costs by handling recurring questions automatically. Our team can focus on what delivers real value: solving complex issues and improving the product.

5. Smart Feedback Collection and Processing

Our assistant does more than answer questions. It includes an intelligent feedback management system that forwards specific needs directly to human support when necessary.

• Automatic escalation: Tickets are auto-created for issues needing human intervention
• Automatic translation: International feedback is translated to French for our team

6. Continuous Improvement Driven by Data

Every conversation is saved and analyzed, allowing us to:

• Identify our users’ recurring needs
• Optimize our product based on the most common questions
• Continuously improve the assistant’s answers using user feedback

This ongoing feedback loop helps steer Photobooth AI’s development in the right direction—a real asset for our product roadmap.

Under the Hood: Technical Architecture

Now for the technical details. Our assistant runs on a modern architecture, leveraging the latest in conversational AI:

Technology Stack

• AI Model: OpenAI’s gpt-4.1-mini, optimized for cost while maintaining high quality
• AI SDK: Vercel AI SDK for integrating models and tools
• AI Elements: For building the conversational interface
• Database: PostgreSQL with Prisma for persistence and analytics

Dynamic Knowledge System

The assistant builds its context in real-time by combining:

• A 25-line, Photobooth-AI-specific system prompt
• Our complete FAQ (203 lines), automatically updated
• The navigation sitemap, adapted to the user’s status (logged in/anonymous)
• A smart caching system with hourly revalidation

In Conclusion

Switching from Crisp to a custom AI assistant has completely transformed how we do support. Now, instead of waiting for user questions, we can anticipate them and provide guidance proactively.

For us, this is also about harnessing our expertise with conversational agents. It’s a concrete example of how well-integrated AI can have a positive impact on both the user experience and business operations.

If you’re interested in this approach or have any questions about setting up an AI assistant for your own product, feel free to get in touch—we’d be happy to share our experience and talk about your project! 🚀