> For the complete documentation index, see [llms.txt](https://docs.rainbird.ai/rainbird/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.rainbird.ai/rainbird/getting-started/how-does-rainbird-work.md).

# How does Rae work?

## Overview

You can create completely bespoke AI models in Rae, but a Rae decision model works very differently to the more widely known AI models based on Machine Learning (ML).

We use a unique blend of technologies to enable businesses to replicate their expert's decision-making, which is created by them and can provide complete transparency for every decision made.

We will break these down, but first it is useful to introduce the field of Knowledge Representation & Reasoning (KR\&R):

> Knowledge Representation & Reasoning is a field in artificial intelligence that deals with how knowledge about the world can be represented in a computer-friendly way, and how computers can use this information to solve complex problems.
>
> Knowledge Representation (KR) involves the creation of formal structures that represent knowledge about the world, objects, events, or situations. These structures can include logic, rules, ontologies, and semantic networks.
>
> Reasoning (R) is the process of deriving logical conclusions from a set of facts. In AI, this often involves algorithms that can infer new knowledge from the knowledge that has been encoded.
>
> KR\&R is fundamental in AI because it **enables machines to mimic human-like understanding and reasoning**, allowing them to interact more naturally with users and **handle complex, real-world scenarios**.

## Knowledge Representation

In Rae, knowledge is represented as a **decision model**, (what might occasionally be called a 'Knowledge Map'.)

A **decision model** is a way of storing and organising information that allows for a more intuitive understanding and retrieval of data. It represents a network of real-world entities — such as objects, events, situations, or concepts — and the relationships between them.

In Rae, a model can be created that represents the data an expert would need to make decisions, and the relationships between these data. Below shows an example of a decision model in the knowledge engineers Studio environment.

<figure><img src="https://2183467111-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FehfnPLAHSx4uHugNiyth%2Fuploads%2FuoU9HLwRrN5HUxuvD72K%2FGraph%20in%20Studio.png?alt=media&amp;token=a6d2a5c3-6a3b-4d08-8675-70796fd9871c" alt=""><figcaption><p>Knowledge and expertise is turned into a decision model</p></figcaption></figure>

These models start as an ontology which is then **extended** with additional layers of information and functionality, including:

* **Enhanced reasoning capabilities:** rules are created on the edges of the graph to infer new data based on other data, just as experts can read a set of data, infer new data and arrive at a judgement.
* **Dynamic and real-time updates:** the model can be connected to real-time data to ensure decisions reflect the latest information.
* **Personalisation and context-awareness:** models can adapt to individual user contexts, so personalised responses can be provided.

Although Rae uses LLMs to support the build phase, turning regulations and rule books into a decision model; these models are written in plain English (or Mandarin, or Swahili, whatever is meaningful to you).

Therefore these expert-made models, unlike Machine-Learnt models, can be easily read, interpreted, tested and modified, providing the trust and confidence that they will provide high-quality and consistent decisions.

Importantly, once built, tested and signed off, no LLMs are involved in the next stage...reasoning.

## Reasoning

> Reasoning is the process of deriving a logical conclusion from a set of facts.

This is where a user can ask for a decision from Rae (a 'query') and the reasoning engine will use the knowledge and logic encoded in the decision model to acquire dynamic/real-time data and infer new data in order to arrive at a decision.

Unlike decision trees, the path to making this decision does not need to be explicitly defined for every scenario, which can become unmanageable. At runtime, the reasoning engine will dynamically figure out the best path to return a decision, including the data it needs to make it, and where to obtain it from. It will also try multiple different paths if data is missing and there are alternative paths to reach a decision. If more information is required make a decision, the reasoning engine will ask for it. It will never guess.

This decision will also be in context to the entity you are making a decision about (e.g. a company, an account, a person etc.) as it will have collated and considered data unique to them. It is therefore highly personalised, and because this knowledge representation was built and tested by your experts, it will provide a decision the same way they would.

## Explainability

> Explainability in AI is about making the internal decision-making processes of AI systems transparent and understandable to humans. This is key for building trust, ensuring ethical use, and complying with legal standards where AI is involved in making significant decisions.

We've talked about how the decision model represents an experts knowledge, that they are written in a language you understand, that they are interpretable, and that the reasoning engine uses this model to collate data, reason over it and arrive at a decision.

This means we are also able to explain the full chain of reasoning used to arrive at a decision, which is transparent and understandable to humans.

We call this the evidence tree and it provides an audit trail that is explainable, traceable and enables Rainbird decisions to be used in regulated industries, or where explainability is important.

<figure><img src="https://2183467111-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FehfnPLAHSx4uHugNiyth%2Fuploads%2F1IrBa39SNSsq9QCNKfKD%2FScreenshot%202026-11-04%20at%2021.44.11.png?alt=media&amp;token=11da61c5-2a9a-4bef-a446-582208c72dde" alt=""><figcaption></figcaption></figure>

## What makes it unique?

The Rae platform is able to provide a unique set of capabilities that are challenging to find together with other AI or non-AI technologies:

<details>

<summary><strong>AI models created and managed by you</strong></summary>

You have complete control over your models and can easily refine these as situations change.

</details>

<details>

<summary><strong>Easy to interpret but scalable</strong></summary>

Decision Trees are often used for easy interpretability, but they are hard to scale to more complex use cases.

ML-models are easy to scale, but hard to interpret.

Rae decision models provide a sweet-spot between interpretability and scalability.

</details>

<details>

<summary>Trustworthy</summary>

With the model being created by experts and the ability to explain every decision made, we are able to provide AI decisioning that does not hallucinate and can be trusted.

</details>

## Why is it valuable?

<details>

<summary><strong>Automate complex, human-intensive, decisions</strong></summary>

We can automate away complexity, freeing up valuable resources and providing improved customer experiences.

</details>

<details>

<summary><strong>Situational decision-making</strong></summary>

Every decision is made in the context of the unique situation. This means a Rae model can provide highly personalised decisions or recommendations the way your experts would.

</details>

<details>

<summary><strong>AI decisioning for regulated industries</strong></summary>

We can enable you to adopt AI decisioning whilst remaining compliant with your industries regulations.

</details>

<details>

<summary><strong>Consistent and high-quality decisions</strong></summary>

These models replicate your experts, but they don’t suffer from noise and bias like humans do and don’t hallucinate like other technologies can.

</details>

<details>

<summary><strong>Get more decisions, faster</strong></summary>

Deploying an automated decision model means you can accelerate the time to a decision, whilst simultaneously making this available to more people at once.

</details>

Keen to understand how this has can used for real? Check out our [example use cases](https://docs.rainbird.ai/rainbird/getting-started/use-cases.md) for some practical implementations using the Rainbird platform.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.rainbird.ai/rainbird/getting-started/how-does-rainbird-work.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
