ASIC to conduct additional testing of licensing guidance chatbot
The Australian regulator has developed a proof-of-concept Technology-Assisted Guidance tool to help businesses navigate the credit and financial services licensing regulatory framework.
The Australian Securities & Investments Commission (ASIC) is planning to undertake internal (and potentially some external) testing of a prototype licensing guidance chatbot. This becomes clear from a report published earlier today summarising ASIC’s regtech initiatives and events conducted during FY2018–19.
The regulator has designed and developed a proof-of-concept TAG (technology-assisted guidance) tool to help businesses navigate the credit and financial services licensing regulatory framework. A proof-of-concept chatbot was developed. The purpose of the tool is to enable businesses to more easily enquire about financial service and credit licensing requirements, as well as related regulatory guidance.
In April 2019, ASIC approached the market seeking a technology company and legal firm to work together to develop a proof-of-concept for a licensing TAG tool. The intention is for the tool to provide informal guidance, not legal advice.
As a result of a tender process, ASIC selected Piper Alderman and Gronade (the supplier) to work on developing the TAG tool. The supplier developed a viable prototype that was demonstrated to ASIC in October 2019 and was publicly presented at ASIC’s 8th Regtech Liaison Forum on November 22, 2019.
The solution was designed and developed in an internal testing environment.
Due to the design of potential capacity for the TAG tool to be used by stakeholders searching the ASIC website, ASIC required the supplier to develop features not otherwise present in off-the-shelf chatbot solutions. The list of challenges included:
- finding a structured way to incorporate complex decision trees and translate these into a logic engine;
- anticipating user needs without having an existing dataset of previous interactions, conversations, or inquiries;
- guiding users based on legal and regulatory requirements in a prescriptive way, while ensuring they understood that chatbot licencing guidance did not qualify as legal advice.
These issues required the supplier to take several tailored approaches when designing the TAG tool.
The most significant design approach was that without a dataset of previous interactions, this chatbot was created as a data-centric solution. This meant that the chatbot had to be designed to anticipate the type of information users may require, and adapt multiple nuanced versions of communication in order to provide the most likely applicable answers, considering legal interpretation. The lack of an initial dataset limited the scope and development cycle of the chatbot.
Further, the complexity of the licensing regime made mapping the logic of the TAG tool rather challenging.
Finally, given how dynamic legislation and regulatory guidance is, there will be an ongoing need for continual updates to the TAG tool.
In 2020, the regulator plans to take the licensing guidance TAG tool project to Phase 2. This means that additional internal (and potentially some external) testing of the prototype licensing guidance chatbot will be undertaken. Phase 2 would inform future decision-making on whether to put such a tool into production.