“There's so much hype around AI. Let's focus on simple implementations of traditional machine learning. We'll get more value that way. Do it with explainability. says Jamie Tuppack, founder of Xplainable.
Brisbane-based early-stage AI software and consulting startup is now targeting ASX200 retailers with services that bridge the “how and why” gap between machine learning predictions and non-technical users. It has four paid pilots under its name, including .
According to Xplainable founder Jamie Tapak, the software runs real-time scenarios, highlights a breakdown of the predictions, and outlines potential optimization points that can be applied within the business.
“We introduced this for ASX200-listed companies who were receiving too many inquiries through online contact forms,” Mr Tapak said at the winning University of Queensland (UQ) Ventures annual ilab Accelerator Pitch Night last month.
“We were able to prioritize leads and also highlight new insights, such as how the length of an inquiry or the use of certain terms directly predicts a customer's likelihood of purchasing. .
“This is not possible with traditional machine learning, and as a result we were able to achieve an incredible 80% increase in conversions, which equates to $24 million. .”
The problem today is that users who create machine learning models tend to exist in silos, often having years of deep technical expertise but struggling to communicate it to the business. He says it's true.
“On the other hand, business users don't know what ML means. Business users who do know, 'What's the added value? How long will it take to implement? And, most importantly,' How do I know? Can I trust you?'
“This is exactly what we do. Xplainable bridges the gap between the business and technical stories. Rather than just providing predictions, we explain how and why those predictions were obtained To do.”
tapak says business news australia The idea for this business came to me while working at a resource company. Therefore, it was necessary to predict the possibility of mechanical failure of large vehicles and equipment.
“We got a really, really accurate result, which is on a better scale than blind guessing, but the problem is it's back to numbers,” he explains.
“I gave the number to the maintenance guys and they didn’t know what to do with it.
“What we've built with Xplainable is that we can tell you the 'why' behind it, like 'high EGR (exhaust gas recirculation) is being attempted,' or 'bearings are worn.' It's a targeted heuristic that you can actually look at and solve problems, not just numbers.”
This is a service that can be applied to a wide range of areas where companies collect data in various forms, such as house price forecasting for real estate companies, compliance and risk checks for insurance companies, and maintenance for resource players.
“I know there are some companies that want AI right now because of all the hype around it, but there are so many facets of AI that it’s hard to know what kind of thing it is. “I don't know if that's the case,” Tuppack said. Say.
“They're typically very early in the data spectrum. Usually it's the whole lifecycle from the education part all the way to implementation and deployment, which is actually a very long process.”
The entrepreneur sees Xplainable likely to become more popular in the retail industry, where companies often generate large amounts of data, but also to address topical issues around bias and discrimination inherent in machine learning algorithms. I'm also optimistic about the role this business can play in this. .
“As regulations around AI and algorithmic processes tighten, we are perfectly positioned to capture that,” he says.
“This is a side effect of the explainability factor. If you have a column representing gender or demographics, you can actually highlight whether that person is more or less likely to achieve X, and by doing so Ask internal stakeholders if they are satisfied.
“I want to get in front of the government and let them know that there are Australian-based companies that are actually trying to solve this problem.”
He says financial loans are another area where the Xplainable model can be useful.
“This way, we can communicate both ways to say, 'In this case, your loan was denied, but if we increase your monthly payment by $200, or if we increase your payment by $200, we'll give you that amount.' 'Right now, the risk on the loan is at 52%, but if you do these things you can get that down to 48%,' he says.
“If you explain to people how things work, they're much more likely to adopt. This was brought to our attention a lot when we were dealing with companies and businesses. I think twice as many people would be happy with a solution that doesn't: just get a number, but explain why you got that number.”
