ASIC wants to talk a new language with NLP

The Australian Securities and Investments Commission (ASIC) wants to learn if Artificial Intelligence (AI) technology can assist it in regulating the corporate landscape. It has issued tender that focusses on a specific subset of AI, natural language processing (NLP), to help with tasks such as drafting correspondence, monitoring text translations of sales calls and reviewing Product Disclosure Statement documents.

According to CapTech, an IT management consulting company, growing numbers of businesses are adopting natural language processing (NLP), the technology that underlies virtual assistants such as Amazon's Alexa.

ASIC wants to undertake a series of trials in 2018 that will allow it to assess how NLP technology works and help it to evaluate its potential to lighten its workload in enforcing and regulating company and financial services laws.

Clamping down on poor financial advice is one of ASIC’s major priorities. One of the tasks it wants NLP technology to help with is the time-consuming and costly manual reviews of client files containing records of financial advice.

ASIC undertakes a significant number of file reviews each year, typically on a sampling basis. These files include multiple documents including: Product Disclosure Statements; application forms; a record of the circumstances of the client, file notes from discussions; and the Statement of Advice (SoA).

A selected sample of client files are examined manually by ASIC officers. The regulator wants to assess the ability of NLP technology to provide automated data review of a larger number of files, with key risk indicators identifying files for further investigation or to refine ASIC targeting of client file reviews.

Another onerous task that could be undertaken more efficiently via AI is ASIC’s reviewing of the advertising and promotion of financial products.

The job of reviewing advertising from radio, TV, print, billboards and online is currently undertaken manually.

“As social media advertising is often targeted to individuals based on their browsing history, ASIC is only getting an exposure to a small fraction of the promotions. This means that there is a potential gap that ASIC is not reviewing.”

When ASIC reviewers find an issue of concern, their first step is to draft an initial letter to the financial entity setting these out.

As part of the trials, ASIC wants to discover if NLP technology can not only help it monitor advertising across all media, including social media, but also automate as far as practical the drafting of initial correspondence.

Additional tasks that ASIC plans to address in the NLP trial include:

  • Speedier and more compressive review of voice to text translations of mandatory recordings of sales calls for general and life insurance products
  • Automated review of analysis of managed funds' product disclosure statements (PDSs), which currently require manual collation and reading and comprehension of text, a time-consuming and subjective process.
  • The ability to extract unstructured financial information in searchable-PDF financial reports and smartly map information into a more structured taxonomy.