top of page
1-modra.jpg
ClsoeIT Logo

The Automation of Model Governance

What is automation?

Automation often represents the process of simulating human activity using computers or machines. It is a current discussion topic in many fields, and model governance is no exception. Our team has successfully implemented several process automations, including automation within the field of model risk governance

 

What can be automated?

The easiest processes to automate are those which are relatively simple. Processes which are typically automated in the field of model risk governance include the following: 


  • data preparation for the development or validation of risk models,

  • data quality checks or execution of various backtesting and monitoring procedures,

  • collection of results,

  • preparation of reports and inputs for further validations.  

How far can model risk governance be automated?

Beyond the above tasks, expert help usually becomes necessary due to the intellectually-demanding nature of the work.  Experts must collect the outputs of the above-mentioned steps, however the preparation steps do have the potential to be highly automated.


More advanced organisations can go even further than the above-mentioned simple steps and consider partial automation of the further decision process. 


Automated tiering: The how and why. 

Particular types of risk models are able to be automatically scored by defined a scoring algorithm and then assigned to a risk tier representing the respective risk of the model usage. This scoring algorithm would need to have access to all the model's qualitative and quantitative model data. 


These automatically calculated data are then represented in a portfolio overview map for a dedicated team. That team can then effectively prioritise their activities leading to higher productivity - a no-brainer for any company.


It’s good to note that, in order to achieve the above mentioned example, the software framework maintaining the automation of the model governance processes needs to have easy access to model data, its dependencies, and validation processes.


Automation of Model Governance workflow diagram

Don’t forget: Automation is a software.

Another important aspect to remember is the fact that automation scripts represent software and need to be treated as such. 


The standards applied in software development need to be applied here - using a code repository, automated testing, and other components. The design should keep in mind a certain level of reusability and of abstraction. 


The execution environment needs to behave deterministically in order to be able to repeat the procedure with all the same inputs in the future. All process executions should also be logged for audit purposes. 


The deployment of automation scripts to production additionally must follow the software development industry standards.


Portfolio status visualization - Model Governance

Virtualisation

The use of virtualisation tools, usual for the world of cloud solutions, has the potential to bring significant cost savings to the IT infrastructure. 


The right architectural design can help achieve the maximum server capacity while still keeping the full independence of automation scripts. This is a clear asset when dealing with large data sets.


The long term benefits of automation

One of the greatest benefits of automation is that it creates continuity. As people leave the firm, taking their knowledge and experience with them, many processes are proved less than sustainable.


I believe that the above mentioned steps - among others -  represent a way to create truly sustainable internal processes. The effort will pay-off significantly, especially in the long term.


Let us know your approach to automation in the field of model governance. Do you want to learn more about our viewpoint on automation? Let's get in touch!

Related Posts

See All

MGS integration with antivirus

One of the MGS features is to manage model-related files and documents. Of course, common and other model non-related files can be uploaded also to the public folder, users' private folder, or shared

MGS to Databricks connector

Analytics teams in banks and insurance companies use platforms like Databricks for data science, ML/AI model development, model testing, and more. These same companies also have teams such as Senior R

Comentarios


bottom of page