How to balance risk, innovation and opportunity

Once relegated to the stuff of science fiction, it is fair to say that in 2023, with the advent of tools such as ChatGPT, artificial intelligence has gone mainstream. While it feels very new, the reality is AI has been around for a long time.

You interact with ‘traditional’ AI frequently – for example, natural language speech recognition apps such as Siri, or the smart algorithms in Spotify. These apps are built around rule sets and algorithms (like, say, how to play chess) that ‘learn’ from user behaviour (like how to specifically defeat you at chess), but don’t create anything new.

AI models have become supercharged and way more accessible. This has been enabled by incredible improvements in computing and processing power in ICT, the advent of cloud and hybrid cloud- processing platforms providing cost- effective horsepower on tap, improved intelligence on how to train models, access to vast data sets, and improved underlying data quality. What used to take many years to reach the level of human proficiency and speech recognition – especially for the New Zealand lexicon! – can now take weeks.

Now an even bigger change has come with generative AI – that is its ability to create something new. Generative AI models are trained on enormous libraries of data with billions of data points, images, code, interactions and relationships. From this, the models generate completely new outputs and data using machine-learning techniques that roughly replicate the neural patterns of the brain.

Spark New Zealand has been working on a scale AI model coined BRAIN – ‘Build Robust AI for Next-best-action’ – which has been deployed in sales and marketing to boost conversion and spend efficiency. In the telco market, only one per cent of the population may be in the market for a mobile or broadband plan at any given time. This means mass marketing campaigns, while necessary for consideration and long-term brand building, are a highly inefficient acquisition tool.

This is where BRAIN has changed the game. By bringing together existing internal data sets on customer preferences, with publicly available data sets that add even more insight, BRAIN can identify in-market customers and deliver personalised offers that fit a customer’s needs, interests and budgets.

This can be as straightforward as recommending Spotify to a customer with a high affinity for music, to something as complex as identifying when a customer is likely to be moving house based on real estate sales and rental data, and offering them a suitable broadband plan. The proof of the pudding is in the eating. Over the past three years Spark has achieved 17 per cent annual improvements in conversion and nine per cent efficiency gains.

With both forms of AI, there are tremendous opportunities to automate processes, create outstanding customer experiences (and therefore preference for your company or product), improve customer satisfaction through more accurate and timely interactions, and to develop, code, productise and create launch plans for completely new products.

AI is even more powerful when deployed as part of a converged solution that brings together a broader range of technologies. For example, Spark subsidiary Qrious developed a converged solution for EnviroNZ to help solve a health and safety issue – visitors to their waste centre jumping into areas where machinery was operating to retrieve items and putting themselves at risk.

By bringing together AI, computer vision and 5G-powered IoT cameras, Qrious was able to create an automated hazard detection system – identifying and tracking people and excavators within a specified detection zone, and calculating factors such as speed and distances between them.

These interactions are analysed in EnviroNZ’s business intelligence environment, where the images from multiple cameras are processed and logged in real time, with hazards flagged, alarms activated and teams alerted in real time, so they can respond.

The use of AI in business is not without significant risk and starting your journey should be a mindful process – not an ‘all or nothing’ decision. But with AI developing at rapid speed and your competitors all over it, there is an equal risk in doing nothing.

Start with the business problem you are trying to solve, not the technology. Where is there duplication, manual or inefficient processes, or wasted efforts? What are the critical insights you need to understand to succeed in market? What are your key business processes and the high-impact interactions and decision points between you, your customers and most important channels to market?

From here you can start with a small number of high priority use cases from which to develop, test and learn to determine how to launch something useful to market with minimum viable data. Consider how you can scale up your model quickly while tuning (and improving) for a superior outcome, while building your interaction records and data sets so you can automate more, improve interaction success, and add new services quickly.

At scale, it is the quality of your oversight platform that will determine your success with AI, particularly in the B2C environment. There are huge benefits that can come from this quickly evolving technology – for individuals, businesses and Aotearoa. But for AI to be successful it needs to be trusted, and with that comes great responsibility for businesses who use it. Embedding an ethical approach to AI from the outset is critical.

Ensure strong data privacy controls, appropriate data retention and storage policies, and clear controls to manage fairness and actively manage bias in AI models. Engage experts in data science and cyber security to establish effective data governance policies.

Many organisations use the European General Data Protection Regulation (GDPR) as a guide, which is a good place to start. If you are putting critical insight generation and process delivery into an AI environment, you want to make sure you have a robust, world-class cyber protection and defence framework in place, and have it externally assured periodically. 

To remain internationally competitive and to thrive in a rapidly digitising world, New Zealand businesses need to be exploring the opportunities AI will bring, while ensuring the appropriate governance is in place to effectively manage the risks. It is about starting an informed journey and developing the risk and opportunity management frameworks that will drive market advantage while overseeing robust, dynamic risk controls.

For directors, get in a huddle with your executive team and get a sense of where they see the primary risks and priorities. Commit to upskilling. Gain knowledge from trusted experts who can point to a track record in AI and have a local ‘A Team’ that can support you.

For companies, start small, find the use cases that matter, take counsel from experts who have walked the path and can help establish your control and oversight framework, engage your board, and make sure you set out with an absolute, system-wide commitment to operate within your own values, ethics and risk appetite framework. 

The opportunities are huge, but prudent trials, thoughtful policy and control development, and good advice will help you balance risk, innovation and opportunity. Doing nothing is not an option. 

Sheridan Broadbent CMInstD is a non- executive director with an executive and governance career spanning telecommunications, ICT, infrastructure and energy. She is an independent director on the boards of Spark NZ, Manawa Energy and Downer Group and is deputy chair of the NZ Business Leaders Health and Safety Forum.