Microsolve Business IT Insights

The AI Opportunity for Australian Financial Services practices

Written by Dale Jenkins | 4 October 2024 10:12:44 AM

The integration of Artificial Intelligence (AI) in the financial services sector in Australia represents a significant shift in how financial institutions operate, offering numerous opportunities for efficiency, personalization, and enhanced risk management. However, as with most changes, it also presents challenges, particularly regarding privacy, data security, ethics and regulatory compliance.

Without a detailed understanding of what AI actually is, it can be a daunting exercise to re-imagine how your practice can make best advantage from this new technical wunderkind.

This article is not a recipe on what AI can do, but more an exploration of how you can identify the opportunities within your practice, engage your team on possible changes and then dive into the technology.

Myth-busting

Let me begin by addressing some widely held myths.

"AI" is NOT:

  • a single application that will revolutionise your practice.
  • a "plug and play" add-on to your favourite tool (despite what many vendors promote).
  • something that comes pre-trained for how you work and the exact services you provide.

AI IS a range of technologies that can be loosely classified as natural language models that allow a computer to understand the intent of a set of instructions based on the summation of the words that it is given.  Where tools such as Google search look only at the words in a search string and return those articles that include the words, an AI engine is able to provide the most PROBABLE answer based on its training data and the prompt you gave it.

The most important concept to understand in the above is that AI provides answers based on PROBABILITY.

The Financial Services AI Opportunity

The opportunity is NOT to change everything in your practice, nor to replace all of your people!

AI is good at interpreting natural language prompts and supplying a response based on the data available to the engine. Your practice has LOTS of data - financial records across all of your clients over many years - this is the ground that AI can mine to find the gold!

So where is the best place to mine for gold?

Let's face it, no two practices are identical, so rather than point at the areas to mine, here is a simple process to identify the most likely candidate areas in your practice:

  • Routine, Repetitive Tasks - while not strictly requiring AI, any routine task with a well-defined start, end and data requirement is fertile ground for automation - whether that includes the power of AI for decision making and communication or not.
  • Data Aggregation and Analysis - wading through mountains of data to find the proverbial nugget of gold is where well trained AI based analysis tools excel - we all have the capability to spot variance in P&L trends, but what about double errors that are cancelling each other - a quick prompt of "find all transactions that varying by more than 10% from baseline over the last 6 years in account X" would be a massive time saver for most practices.
  • Risk Assessment and Management - extending on the above data aggregation and analysis point, AI's ability to consume vast quantities of data in rapid time give it an unparalleled ability to provide real-time assessments of financial data looking for transactions that may be fraudulent, changes in purchasing habits, missed invoicing, risks from employee leave allocations, etc, etc.  Think of it as providing canary in the coal-mine services across all of your clients!
  • Communication Monitoring - keeping track of the happiness of clients with each interaction is not easy - most practices we work with have multiple communication channels in use at any one time with no practical way of knowing which clients are unhappy and at risk of leaving.  Again, this is an opportunity for an AI tool to provide real-time communication "sentiment analysis" and alert when an issue is detected - ideally, this will reduce client churn and provide plenty of opportunity to intervene only when needed.

The above is not an exhaustive list - have a look within your own practice at areas of "pain" - that is where to start digging your gold mine!

Challenges and Risks

Any technology, new or existing, comes with a set of unwanted side effects - commonly known as challenges and risks - AI is no different.

Keep in mind that we have all been here before.  The Internet, NBN, Smart Phones, IOT, Cloud Computing, social media - these are all technological advances that introduced (or renewed) sets of challenges and risks, some unique, some just a smidge different - but none, after review, that couldn't be handled in a clear and logical way - again, I reiterate, AI is no different.

Let's quickly address the four of the most common areas identified as needing specific attention:

Data privacy and Security - AI has a great ability to process substantial volumes of data - this strength requires careful control and governance to ensure that it does not become an avenue for disaster. Ensure that any AI tools are configured such that processed data is NOT used for training of models, quality assurance or stored outside of your organisations data environment.  Ideally, utilise a privately deployed AI model that does not have Internet access and allows for configurable control of how data is managed rather than a publicly accessible "app" such as Chat-GPT. Having an internal only AI tool provides far greater security control, however, care must be taken to ensure that enthusiastic employees don't mistake a public AI tool for the internal corporate tool.

Regulatory Compliance - This is an emerging area with potential for rapid change in many industries as tools mature and risk profiles are better understood. Avoiding bleeding edge deployments is probably a good idea until well defined regulations are available within the Financial Services industry - this should not be seen as a showstopper to innovation, more a be prepared to have to adapt if/when regulations are released by APRA, ASIC, etc.

Accuracy and BIAS - of all the areas requiring attention, this is the one to really have a specific focus on.  Early in this article I used the term PROBABLE to describe the outcome of an AI prompt - that is exactly what a Generative AI prompt outcome is - the MOST probable arrangement of information based on the training data and prompt.  Generative AI is not a deterministic response - often the most probable response will be close enough to a deterministic outcome, however, care needs to be taken to provide guardrails to AI responses so that clearly incorrect (ie: low accuracy) responses have some form of check step before being acted upon.  Many tools allow for configuration of the level of probability that is accepted in responses - some areas may need this to be set to 0.

Workplace acceptance and engagement - of all the challenge areas this is the one that gives the best indication on the health of your workplace culture and will either be a walk in the park or require the sort of workplace cultural gymnastics that makes you wonder why you bother getting out of bed each day - yep, it will be that stark!  The best tips for successful engagement are to start small (both in terms of expectations and team size), be open and transparent in communication with the team and ensure that failure is seen as an acceptable outcome.  It will take time for some members of the team to accept change and that is OK - ultimately this is not about revolution, but about evolution and improving both work outcomes as well as the process.

Not covered in the above topics, but absolutely needing attention is the message that is portrayed to clients - it needs to be respectful and take them on the journey as well.  Ultimately, it is their information that is being processed so they do have a right to understand and be involved.

Where to Start

A journey of a thousand miles...
Starts with a thorough understanding of where you currently are!

Let me be really clear, there is absolutely no point in spending hard earned dollars on testing and deploying an AI project if you don't already have a well-structured, understood and documented IT infrastructure in place - i.e.: you can't build a second story on your house without having sound foundations in place!

Honestly there is no one perfect process that will work for every practice - there are, however, 5 universal stages that we have found to provide a great basis to work from:

  1. Know where you are and who is on-board
  2. Establish an end-goal, check-steps and boundaries
  3. Form a steering Group and communications strategy
  4. Start small and pilot one or two ideas
  5. Review, Report & Recommend

What Next

Yes, as sure as the sun rises in the east and sets in the west, there will be some new technology on the horizon that will revolutionise the world and change business as we know it.  Embrace the journey, ask questions and be ready to pivot!  Oh, and we're here to help.