top of page

SMA Crown Confidential:
Our Data Methodology

an elegant and intelligent looking young man in glasses, data transfer,  SMA Crown Confidential logo, and AI word

SMA Crown Confidential Methodology Process

 

The SMA Crown Confidential working with client’s data, and Digital Confidantes creating and training process follows a systematic 7-step process.

Our 7-step process includes:

1. Defining the Main Focus

We start with defining the main focus of the created Digital Confidante or Personal Assistant, its purpose, use, application as well as required capabilities. Together with the client, we refine the characteristics and work out a ‘model’ for their Confidante . The model includes choosing the Assistant’s character, if different from the user’s, defining its role and responsibilities, as well as its preferred communication style.

 

2. Collection of Data

Once the ‘model’ for the required Confidante  is established, we move to identifying client’s available data sources which will be used for training of their Confidante , including property/business descriptions, lists, documents, manuals, summaries, prices, schedules, timings, and other relevant training data. We also consider private, technical, and regulatory restrictions that define the framework for data collection and analysis.​

3. Cleaning and Preparing the Data

Next, we focus on the preparation of the collected data, which involves structuring and organising it, removing possible errors, filling missing information, and formatting data correctly for AI understanding.

 

4. Building a Model

At this stage, together with the client we choose the LLM model which will form the base for their Confidante . The client can choose from the wide range of existing LLMs including Claude, Gemini, DeepSeek, Grok, GPT.

​ 

5. Testing of the Model

Once the Confidante  is trained, we perform rigorous testing and, if needed, adjust and fine-tune it.

​​

6. Deploying the Model

This stage involves implementation in relevant real-world applications be it client-facing apps, internal tools, or automated processes.

 

7. Monitoring and Improving

Once the Confidante  has been deployed, we keep an eye on its ongoing performance, including regular updates of sources with new data, introduction of improvements to the Confidante’s role, when and if requested by the client, as well as required technical upgrades.

Talk to Evelyne, Our AI Executive Assistant

CONTACT US

bottom of page