Noupunt AI

Brand Assistant 24/7

A chatbot trained on a brand's own content — products, services, values, FAQs — integrated directly into the corporate website. Answers visitors' questions naturally, guides them through the site and provides relevant information 24 hours a day.

The challenge

Most corporate websites hold a lot of information but offer little help finding it. Users have to navigate, search and guess. A contact form takes days to get a reply, and a live chat with human operators depends on schedules and availability. The challenge is to provide an immediate, reliable point of contact that stays on-brand — without multiplying support costs.

The solution

We design a conversational assistant that integrates into any website, customised to each brand's tone and visual identity. The interface is simple by design, but the conversational engine behind it is capable of much more than answering questions:

• Autonomous navigation — takes users to specific sections of the website without them having to search
• Data capture — collects contact information, preferences or needs during the conversation
• Interface actions — fills in forms, opens sections or triggers page functionalities
• Persistent context — remembers the conversation thread and adapts responses accordingly
• Cross-session memory — stores previous conversation history and lets users pick up where they left off days later, even after closing the browser
• Grounded in real content — every response is generated from the brand's own documentation, avoiding hallucinations
• Connectable tools — the engine can execute real actions: send emails to the company, log analytics, check availability, generate quotes or any operation exposed via API
• Engagement buttons — users can move the conversation forward without typing: quick reply buttons that streamline the interaction and reduce friction

Each implementation activates the capabilities the project requires. This version shows the most direct use case: an informative assistant, always available.

Results

• Lightweight, embeddable widget compatible with any web stack
• Responses grounded in real content via RAG
• Uninterrupted support without relying on schedules or support teams
• Extensible conversational engine: navigation, forms, data capture and connectable tools as needed

An assistant that answers is useful. An assistant that acts is a different level.

AI Guide for Virtual Exhibitions

A conversational agent designed to accompany visitors inside a virtual tour. It's not a chatbot that talks about the space — it's a guide that controls it, navigates through it and generates dialogue from real content.

The challenge

A virtual tour without a guide is a passive experience: the user moves around, looks, and leaves. Static information panels cover the basics, but they don't generate conversation, don't go deeper and don't adapt to each visitor's interests. The challenge is to turn the tour into an active experience where someone accompanies you, explains and takes you where you want to go.

The solution

We design an AI agent integrated directly into the 360° tour with two layers of operation:

Conversation and navigation:

• Deep knowledge — the agent knows the exhibition content and can discuss it beyond the information on the panels: historical context, connections between works, curiosities
• Autonomous navigation — moves the user to specific points in the tour by changing views and scenes, without them having to search or touch anything
• Active engagement — doesn't wait for questions: suggests, proposes and guides the route based on what sparks interest
• Cross-session memory — remembers previous conversations with the visitor and can resume the tour days later, even after closing the browser
• Engagement buttons — visitors can move the conversation forward without typing: quick reply buttons like "Tell me more", "Take me there" or "Next work" that streamline the interaction

Integrated information points:

• Hotspots placed throughout the tour that, with a single click, show basic information about each work: title, author, technique
• A second click opens the chat with a real-time generated expansion that takes into account the conversation so far — it knows what has interested the visitor most and steers the discussion accordingly
• The same conversational engine powers both modes: quick response and deep exploration

Where it can go further

The same system adapts to any navigable environment: museums, trade fairs, real estate showrooms, heritage sites, product catalogues. Wherever there's a digital space and content to explain, a guide like this fits in.

Results

• Conversational agent with real navigation over the 360° tour
• Dual information layer: instant panel + contextualised generative expansion in chat
• Knowledge that goes beyond the application's static content
• Active experience: the guide proposes, moves and goes deeper

A virtual tour with its own guide isn't just visited — it's experienced.

Virtual Catalogue Advisor

A conversational agent with full knowledge of a product or space catalogue that talks to users to understand their profile, showing and loading content as the conversation progresses. It's not a search engine with filters — it's an advisor that listens, understands and proposes.

The challenge

When a company has dozens or hundreds of products, finding the right one for each person is a real problem. Traditional filters require the user to know what they're looking for. But often they don't — they want someone to guide them based on their needs, budget and priorities. The challenge is to replace passive browsing with a conversation that understands and decides.

The solution

We design an agent that integrates deep catalogue knowledge and acts on the interface while conversing:

• Conversational profiling — through natural questions, the agent identifies the user's preferences, budget, priorities and lifestyle
• Full catalogue knowledge — masters the entire database: technical specs, prices, locations, associated services, nearby areas, comparative advantages
• Dynamic loading — as the conversation progresses, it shows and loads relevant content on the interface: images, tours, data sheets, comparisons
• Autonomous navigation — takes users directly to the spaces or products that suit them, without them having to search
• Cross-session memory — remembers previous conversations and can resume advising days later, even after closing the browser
• Engagement buttons — users can move the conversation forward without typing: quick reply buttons like "I'm interested", "Show me", "Compare options" that streamline the interaction

The working example is integrated with Yugo, the leading brand in student accommodation. The agent starts by asking which city the student plans to study in. From there, it converses to understand the profile — whether they're looking for premium quality or affordable pricing, how much they value nightlife and social life in the area, transport, proximity to universities — and ends up proposing the most suitable residences and rooms, showing shared areas, gyms, study spaces, distances and included services.

Where it can go further

The same system works for any company with a broad catalogue: estate agents, hotel chains, retailers with multiple product lines, museums that want to guide visitors towards the works that interest them most. Wherever advising is needed, not just listing.

Results

• Agent that profiles and advises through natural conversation
• Dynamic content loading on the interface based on the conversation
• Autonomous navigation across the full catalogue
• Persistent memory to resume sessions
• Quick reply buttons to reduce friction

When you have a hundred options, the best filter is someone who listens.