8217;s just compounding because every single provider that we use is thinking about AI and all of their providers are thinking about it. We’re all in this AI bubble and that’s why we see it move so quickly. It’s really hard to keep up – in a few months, the stuff we’re talking about is probably going to be different and change.” Kogan has been using machine learning and artificial intelligence to increase its operational efficiency for many years, but recently, it has been experimenting with generative AI to personalise in-situ product images, assist with coding and improve its chatbot.  But as Stefkovski noted, the landscape is changing rapidly and retailers can’t afford to get left behind. He shared seven examples of how businesses can use generative AI and his advice for those just getting started.  1. Warranty claims  One potential use case for generative AI in retail is assessing warranty claims. Stefkovski shared a prompt he created to test whether a generative AI tool could accurately assess a customer’s warranty claim, and while he was impressed by the results, he believes this application still requires human intervention.  “It demos well and is really useful, but how do you scale in reality?” he said. He also noted that claims involving products that don’t have obvious damage, like a shattered screen, will be difficult for a generative AI tool to assess.   “It may be great as a copilot system, but it’s probably not ready to just be your warranty assessor,” he said. 2. Review summaries  One application of generative AI that Kogan is currently working to implement is review summaries. Stefkovski noted that many SaaS products offer this capability. The benefits are clear, and the risks are minimal. “If you have too many reviews, customers can’t sit there trawling through them all,” he said.  “Even if they do sometimes miss something, [customers] can still scroll down and see the source data.”  3. Video summaries  Like many businesses, Kogan is using generative AI to summarise internal video calls. As Stefkovski noted, generative AI tools can identify who is speaking, summarise the call, assign action items and share the document with relevant parties.  “That’s pretty useful for note-taking, but the next step people in the industry are doing is they might record their initial workshops and then distill that information with another [generative AI] model – put it into dot points and slides and presentations or templates – and it saves them a ton of work,” Stefkovski shared.  “It’s super efficient and also, you’re always underdocumented, and so having the documentation there just means that you can search for it quickly and find it if you do need it with barely any human involvement.” 4. Training guides Similarly, Kogan is using a generative AI tool called Scribe to create training guides.  “It’s like a browser extension or desktop app, so you click ‘start recording’ and just start doing what you need to do. It means you can make a guide for something tricky or hard. Usually, you would have to take screenshots and then drag them into a Word or Google Doc, so it’s super efficient,” Stefkovski said.  5. Coding assistant  The next generation of coding assistants are like “auto-complete on steroids”, Stefkovski shared. They use generative AI to predict the next five or six lines of code based on what has come before.  “Coding is not as expansive a language set as English or another language, so it’s very good for predictive models,” Stefkovski said. “But what is really impressive is a tool that we use internally called Cursor AI. You go on the site and say, ‘I need to build a preference centre, it needs to speak to the API and allow people to pick their preferences,’ and it just starts throwing code in. Now, if you don’t know how to code, you’re not going to get far, but if you do know how to code, it makes you five to six times more productive.”  The next step in this space is the rise of no-build tools, where users simply need to describe their idea, such as, ‘I want a Twitter clone that does X’, and the tool will start coding it.  “It’s got constraints,” Stefkovski admitted. “I don’t know how well it scales, but it’s something I’m going to play with. Maybe it’s great for mini apps that help your team productivity.” 6. Deep research The next use case Stefkovski is looking to explore is what he calls deep research. For example, he recently asked a generative AI tool to summarise what consumers are looking for online when they’re in the market to buy a projector, such as brightness, resolution and portability, and assess how well Kogan’s product listings answer these questions.  “I want to know if we are addressing all their concerns,” he said. “Then, you can use generative AI to rewrite the descriptions to better work for this.” 7. No-code workflows The biggest opportunity for businesses to realise productivity gains from generative AI is by using no-code drag-and-drop tools, such as Make.com, to automate workflows.  Using the example above, businesses could ask an AI model to identify the key questions consumers have when searching for a particular product online, and score their product listings against those questions. Then, they could input this information into another AI prompt to rewrite their descriptions to address those questions. They could specify that the AI model should flag any descriptions it’s unsure about in a spreadsheet for humans to intervene, and then upload the new descriptions onto their website using Shopify and other SaaS integrations, or API integrations for custom-builds. “This is where that real 10x-ing of your day-to-day job hits,” Stefkovski said. “Now, the only danger is again, what the hell is it writing if you do this at scale? 
If you do it on 500 products, you are going to have to – at some point – check this, right? Because we’ve seen it hallucinate.” A potential solution to this would be to have another AI agent check the quality of the first AI agent’s work. Stefkovski’s advice The potential for generative AI to increase users’ personal productivity is significant, Stefkovski said. And while many businesses like Kogan are already experiencing efficiency gains, this is just the beginning.  “Overall, I think, get creative, that’s the only way with this stuff. Just play around with it and you’ll surprise yourself,” he said.
 Further reading: ‘Building momentum’: Kogan founder hails revamped company as back on track