@ the IERP® Global Conference, August 2023
Speaker Michael Hoo, Country Head of Customer Engineering, Google Cloud Malaysia, opened his session with a barrage of questions: Will generative AI grow business? Will it save costs? Will it differentiate or create a moat? Will it launch faster? Will it make employees happier? What was important, he said, was that the organisation was ready for generative AI adoption. This depended on how it managed several factors including data governance and privacy; how it adhered to regulations; what kind of security and compliance support it had; its reliability and sustainability; and the issues of safety and responsibility,
There were many reasons for an organisation to adopt generative AI, the primary one being the ever-increasing rate of digital consumption. Currently, millennials and Gen Z users spend more than four hours daily on their electronic devices – but their attention spans are declining. Generative AI can be applied towards better personalisation or targeting of consumer choices, particularly when the swipe-left/swipe-right function can deliver instant gratification. “Imagine what GenAI could do to (improve) your catalogue,” Hoo said. “Imagine the creativity it could unleash for your marketing and content design. Every customer could get an individually personalised ad.”
That is not all GenAI can do. Generative AI not only responds to input but creates new content. “It creates summarisation,” Hoo explained, adding that about 10% of enterprises have already moved AI into their mainstream production processes, applying AI in some form or other – in predictive analytics, and making sense of data, for instance. “At the top of the mind of CEOs is always the thought of increasing revenue, reducing costs and risks, and building competitiveness.” But tech adoption also requires consideration of other factors, such as the availability of skills in the market that will allow the company to make things work.
Many customers ask if they can develop GenAI themselves, Hoo remarked, explaining that whenever Google ‘trains’ an AI model, the carbon footprint generated is like sending a rocket to the moon and back. “That’s how much computing is required for generative AI,” he said. “It’s not going to be possible for a while because it requires so much computing power.” It makes sense, therefore, to build on what generative AI companies have already done. When applying GenAI to business or enterprise, the main areas to consider are maintaining control of data; complying with regulations; the reliability and sustainability of AI results generated; and whether the app will produce racist or biased output.
Urging users to understand the potential of AI, he said that in today’s environment, if a customer cannot be ‘nudged’ to decide within ten seconds, the odds are that you have already lost that customer. Apps are becoming more interactive, engaging and user-friendly; they are designed this way because of the way companies are now marketing products and services. “What is my customer? What is my customer segment?” he said. “Try to predict what they want to buy. The real potential of GenAI lies in how things will change. There will be assisted user experience and product downloads, more experiences will be generated, and there will be (added and more relevant) context.”
What this means is that instead of several (customer) segments, a ‘segment of one’ will be created – individual, very targeted segments that have specialised or customised catalogues for every individual. Although the complexities of this type of targeting (‘mass intimacy’) are not yet completely resolved, this is something AI can do for individuals as well as businesses. It may even be able to help individuals with their finances and budgeting in a meaningful way. In today’s Internet economy, the value happens at the apps stage, he said, but there were billions of apps available now, so where could value accrue, with the application of GenAI?
“If you do not know where this happens, you will apply it where you will not make any value for the business, and that becomes a risk,” he stated. Foundation models like ChatGPT for instance, are highly accessible and will decrease in price as they are increasingly utilised; foundation models, therefore, are not where value will accrue. Citing music service Spotify as an example of how AI may be applied, he said that Spotify starts predicting what kind of songs a user likes, after a short period. “With generative AI, the app will even be able to start identifying the types of songs you like during certain periods of the day, and put together a customised playlist based on this,” he explained.
The experience is completely personalised, and curated by the app, which can do it for many users, in a shorter period. While Spotify profits from the money it makes through licensing agreements with music companies, it also derives specialised or proprietary data. Once integrated, this data is enhanced when generative AI is applied. “What you want is for the AI to give you ‘nudges’ which matter to you,” Hoo said. If applied towards helping to manage one’s finances, for instance, it can identify income and expenditure, and compare this against that of one’s peers, showing what one can or cannot afford, given one’s current portfolio. For the producer, another opportunity to sell is created.
The constant human fear of missing out – “if you do not have this, you will miss this opportunity” – is thus created, by the app being able to present the right opportunity at the right time. “The ten-second window has happened,” he said. “This is an example of how you get so much data when all you asked was your cash flow. Use generative AI when you have proprietary data, embed it in the product on offer, and create a competitive difference.” There are many layers to AI; its effectiveness depends on how it is used – but he predicted “an explosion” of how to use industry-specific AI soon, He urged a closer look at proprietary data which can be utilised by generative AI, for future leverage.
