How Kleio Is Transforming Complex Journeys With AI Agents

Online sales often fall short where it matters most: high-value, complex products. Whether it's planning a multi-stop vacation, choosing the right real estate investment, or upgrading to a more energy-efficient home, many users end up frustrated. They visit a site, poke around, and leave without taking action. Filters and search bars aren’t enough - because most people don’t know exactly what they’re looking for. They need help defining it.

That’s the gap Kleio fills. Formerly known as Autodm AI, Kleio is building a next-generation conversational platform powered by modular AI agents designed for complex user journeys. In a recent episode of Datadriven 101, our co-founder and CTO, Louis Poirier, shared an inside look at the technology, business logic, and use cases powering our success.

👉 Watch the full interview on YouTube to learn more!

Why Traditional Web Experiences Fall Short

Most websites are built on a big assumption: that users know what they need. But when the stakes are high - booking a luxury trip, investing in a property, or selecting the right energy solution - this just isn’t true. People aren’t sure what to ask, what to compare, or how to even begin defining their needs.

Kleio changes this by replacing passive browsing with active, AI-powered conversations. Instead of leaving users to figure it out themselves, Kleio’s conversational agents guide them - step by step with tailored conversations - toward clarity and confident decisions.

Turning Conversations Into Conversions

At the core of Kleio’s platform is its multi-agent architecture, built to handle complex online journeys. Rather than relying on a single, generic chatbot, Kleio deploys a team of specialized agents, each designed for a specific part of the user experience.

As a comparison, Kleio’s agents could resemble the roles of real-world staff: one agent welcomes the user, another might focus on specific destinations, and a third is responsible for closing the sale. Other agents can handle diverse tasks like showing agency contact details or sharing flight booking links. Each agent is equipped with different tools, and they work together to support non-linear, dynamic conversations that resemble real human interactions more than scripted chat trees.

This orchestration enables Kleio to guide users through complex decisions  -  like choosing a vacation or comparing products  -  in a more natural and efficient way than traditional website navigation.

Why Multi-Agent > Monolithic AI

Instead of banking on a single massive language model, Kleio takes a modular, multi-agent approach. Louis explains this quite clearly - large, monolithic language models tend to hallucinate and are harder to test and control. Kleio’s solution is to break down the problem into smaller, more manageable agents, each performing a very specific task.

Each agent is lightweight, task-specific, and optimized for performance. This makes them easier to test, control, and trust - especially critical for enterprise clients where hallucinations and inconsistency aren’t an option.

This approach allows the team to limit what each agent can access, block or allow data sources, and test behaviors with precision. Kleio also leans on SLMs (small language models) instead of LLMs. They’re faster, cheaper, and often more effective for focused tasks like routing, extraction, query generation, product matching, or classification. In this case, smaller really is better.

Behind It All: Kleio’s Knowledge Graph

To reduce the risk of hallucinations  -  when an AI generates plausible but false information  -  Kleio minimizes each agent’s scope and grounds responses in structured company data. 

This brings us to a foundational element of Kleio’s architecture - its property-based knowledge graph. Products are represented as nodes connected to relevant attributes such as destination, pricing, availability, themes, and more.

As Louis describes, this allows Kleio to handle complex filtering and personalization that goes beyond basic SQL queries. For instance, the graph supports linking products through shared themes like “adventure” or “family,” even when those tags are not explicitly listed. The system can then recommend related alternatives  -  e.g., if a user has been to Greece, it might suggest products in Montenegro with similar thematic links.

Kleio also enriches product descriptions using language models, extracting details like Wi-Fi availability, the presence of water slides, or unique destinations  -  all of which feed into the graph to improve downstream recommendations.

Crucially, this structure supports both pre-processing and runtime operations: data is enriched ahead of time, but queries made during a live conversation traverse the graph to provide real-time answers.

How Havas Voyages Tripled Leads With Kleio

A standout example of Kleio in action is its deployment with Havas Voyages, a major French travel agency. In just a few weeks, Kleio helped triple their lead volume by guiding users through the complex travel planning process via personalized, conversational experiences.

As volume grew, so did the need for better lead qualification. Kleio responded by integrating with their system, tapping into a massive catalog of over tens of thousands of offers and millions of product references. A custom knowledge graph layered on top allowed agents to instantly match users to the best-fit travel options based on timing, preferences, and budget - before handing them off to sales.

The result: higher conversion rates and several less wasted sales hours on low-quality leads.

Designed for Enterprise, Built for Scale

Kleio isn’t a plug-and-play chatbot - it’s a fully configurable platform that integrates into any tech stack. From Salesforce to spreadsheets, CMSs to CRMs, Kleio adapts to your systems, tone of voice, and user experience - whether on desktop, mobile, or multilingual environments.

Integration is simple: one Javascript snippet for the web, with APIs for everything else. Clients can start with a single use case and scale across teams, departments, or markets - no replatforming required.

From POC to Production-Ready AI

A lot of companies get stuck at the GenAI proof-of-concept phase. Demos look impressive, but quality, governance, or scale becomes a blocker.

Kleio is built to move beyond that. With observability, versioning, and QA tools baked in, clients can test at scale, monitor results, and iterate safely. It’s not just AI - it’s enterprise-grade AI that performs in production.

In one deployment, 25% of users who spoke to Kleio’s agents submitted both a phone number and email - crushing the performance of static forms or generic bots.

Built by Enterprise AI Veterans

Kleio’s founding team - Louis Poirier, Adrien Mathieu, and Philippe Wellens - spent over a decade at C3 AI, leading complex AI deployments across manufacturing, defense, and energy. They’ve delivered multi-million-euro projects at global scale - and they’re now applying that expertise to transform how AI supports sales and marketing.

That experience shows in Kleio’s architecture, QA rigor, and ability to plug into real-world systems from day one.

Not Just for Travel: Designed for Any Complex Journey

While Kleio launched in travel, its use cases extend far beyond. Real estate, energy, cosmetics, automotive - any vertical with a complex buying journey benefits from Kleio’s human-like guidance.

Whether it’s helping someone choose the right skincare regimen, compare electric vehicle incentives, or decide where to buy a new home, Kleio makes the process intuitive, personalized, and immersive.

📺 Watch the Full Interview

Want to dive deeper into Kleio’s vision, technology, and real-world success stories (in French)? Watch the full Datadriven 101 episode with our CTO Louis Poirier on YouTube.

‍

‍