How does AI-powered revenue management work? From data to answers

Hardly any SaaS provider today gets by without AI features. Whether it's deeper insight into a hotel's own data, simpler workflows, or new automations, the promises all sound remarkably similar. What most people don't realize is that nearly all of these products rely on the same handful of foundation models, built by a small group of providers such as OpenAI, Anthropic, xAI, Google, or DeepSeek. Training a large language model (LLM) from scratch costs well over 100 million US dollars, so developing one in-house simply isn't worthwhile for most providers.

But if only a handful of models exist, what actually sets these products apart from one another?

Context makes the difference

Nowadays, when people talk about AI, they usually mean an LLM. To understand what really separates one AI solution from another, it helps to look at how such a system actually works.

The key factor is context: whatever information comes along with a given request. A question about the weather for the coming days, for instance, can only be answered reliably if the system knows the user's location. Without that context, any answer is little more than a guess.

The AI Revenue Manager: knowledge on demand

When an LLM is built into software like RateBoard, the goal is for the model to gather the relevant context on its own. That's exactly what a set of "tools" was built for. A tool pulls in additional information whenever it's needed, so the model's reasoning stays sound and easy to follow. The model itself decides which additional information it actually needs.

Right now, the AI Revenue Manager draws on the same core data that hotel teams already see within the RMS. That includes available room types and how many of each there are, along with key metrics such as occupancy, RevPAR, ADR, revenue, and pickup. Events, weather, and forecasts round out the picture. The number of these tools keeps growing, and connecting external data sources is already in planning.

If someone asks the AI Revenue Manager a question that the available tools can answer better than the model's built-in knowledge, it generates a call to the appropriate function, along with all the necessary parameters. The function's output is then parsed by the AI Revenue Manager and fed directly into the model's context. This way, the model gathers the knowledge it needs on its own, right when it's asked for.

Data quality over data quantity

AI is only ever as good as the data it can draw on. Reliable data quality, not sheer data volume, is what really separates success from mediocrity.

The AI reaches its full potential once benchmarking data comes into play. Only by comparing a hotel's pickup with the broader market, for instance, can you tell whether that performance is actually strong or weak. That's the only way to judge whether a rate adjustment makes sense, or whether the property is already outperforming the competition.

What makes LLMs ideal for revenue management

LLMs excel at pulling relevant information together from different sources and spotting patterns as well as connections within it. What would take a person considerable mental effort turns, with the AI Revenue Manager, into a simple conversation.

That's why the assistant's capabilities keep expanding. It's continuously fed with background knowledge specific to the leisure hotel sector, so that everyone in the hotel who deals with revenue management can get to the information they need as quickly as possible.