Algorithms are routine processes or sequences of instructions for analysis data, solving problems and performing tasks. Algorithms process data models and produce output decision, which is independent of algorithm design. They are widely used by individuals and organisations across various market sectors. Automated trading and execution of trades, consumer credit scoring and detection of unusual activity are only some examples of where algorithms can be applied.
Why is algorithm assurance important
The increased prevalence of algorithms has not come without risks. Faults in the design and operation of algorithms can have various implications including financial losses, disruption to operations and regulatory violations. It is for this reason that stakeholders are asking for a greater degree of assurance that algorithms are working as intended and risks are properly managed.
Algorithm Assurance gives stakeholders confidence around the operating effectiveness of these complex set of calculations. Assurance can be provided not only on the performance and design of the algorithms but also on whether risks and being properly identified and mitigated. The scope of assurance covers the risk management framework and controls that are in place.
New regulations are making some form of algorithm assurance mandatory. For example, MiFID II, Regulatory Technical Standard 6 (RTS6), prescribes organisation requirements for firms that perform algorithmic trading including the undertaking of annual self-assessments. Also, the proposed EU AI Act will introduce mandatory audits of high-risk algorithms.
How we can help
Controls Services: We can support you to build a strong controls framework by performing a detailed review of the internal governance structure and existing framework, for the design, implementation and testing of algorithms
- Governance & Regulation Compliance: We will provide you with assurance over your governance framework and internal controls relating to the implementation and operation of your algorithms
- Testing and deployment: We can provide assurance or help you build processes around development, training, testing and deployment of algorithms
- Documentation: Our team will ensure your documentation is transparent and reflects the models and decisions taken by the algorithm
- Monitoring: We can provide assurance over your monitoring processes to ensure that they are correctly designed to identify bias or incorrect decisions
Algorithm Validation: We can help you optimise and ensure resilience of your algorithms by measuring their performance under different circumstances and flagging hidden risks
- Code Assessment
- Model Validation/Risk Management