Allianz intends to focus on AI in automating service and claims processes. At the same time, global infrastructure providers such as AWSshape new payment standards and open the way to the use of stablecoins as another “payment track”, parallel to instant transfers or cards.
The rest of the article below the video:
AI in finance: transforming operations instead of the “black box” of risk
Jan Malmendier, chief operating officer of Allianz in Germany, calls artificial intelligence a breakthrough force, but at the same time emphasizes that this does not automatically transfer responsibility for risk assessment to AI models.
“For us, artificial intelligence plays an extremely important role. From my perspective, it is the biggest and fastest technological change of the last 20-30 years,” Malmendier tells us during the Amazon re:Invent conference in Las Vegas.
However, he emphasizes that his position is prudent. “In classic risk assessment, we practically do not use AI. Ultimately, costs are a really sensitive topic. And we really want to make sure there's no bias there,” he explains.
As he explains, the classic pricing policy at Allianz is still largely based on rule-based algorithms rather than self-learning models, precisely because issues of cost and customer selection are extremely sensitive to potential “biases” in the data.
Instead, the company is focusing on areas where AI can improve the quality and speed of service without transferring full decision-making responsibility. “AI is great at understanding documents. It's also now very good at understanding customer intent on the voice side,” Malmendier describes in practical applications.
Allianz uses algorithms to analyze each incoming report, automating decisions in simple, low-value cases and leaving complex matters to employees. “In many cases we let AI handle cases that are not very complex and expensive, but for standard claims we are very good at deciding what we can pay now,” he says. This allows you to shorten your reaction time and maintain control over riskier decisions. The result is higher customer satisfaction: Allianz measures satisfaction from over 150,000 reviews and achieves a rating of 4.71 out of 5 stars – significantly higher than in the past.
Productivity, human resources and an aging Europe
From a management perspective, Malmendier associates the development of AI with two parallel trends: pressure to increase productivity in the European economy and demography. Allianz assumes that thanks to AI it will be possible to increase operational productivity by 20-30%.., and today it is about 10%. acceleration in the area of code creation and maintenance.
“We are targeting a 20 to 30 percent increase in productivity. Greater productivity through artificial intelligence. It's not just AWS, but also other big models and players that we use” – Malmendier explains the scale of the transformation. This translates into the ability to serve a larger number of customers with a decreasing number of employees.
Malmendier associates this with the inevitable shrinking of labor resources in Europe. “We have an aging population in Europe. We will serve more customers with fewer staff. And that is very clear,” the head of operations states bluntly. He also adds a macroeconomic dimension: “I think as an economy we need constant and strong productivity growth. And honestly, it hasn't been that high in the last 10-15 years in the European Union“.
At the same time, Malmendier stipulates that human contact will remain crucial where advice and trust are involved. “The human factor is important and will always remain. I am very convinced that we can create an outstanding service with AIwhich is not worse. In many cases this is better than what we can do today. But it's still good to have some human interaction there, especially when you're consulting on what appropriate insurance you have. It is based on trust,” he says. In this area, employees will still play a key role, while standard processes can be taken over by AI agents.
AI as the infrastructure of new customer service
John Kain, responsible for business development in the financial services sector at AWS, describes the transformation primarily through the prism of the specific impact on the customer experience. He has been at AWS for eight and three-quarter years and has watched the financial sector evolve from a technology perspective.
The Coinbase example shows how quickly the proportions between manual and digital support are changing. “Two years ago, only 20 percent of their customers could be served digitally. 80 percent required human assistance. By implementing generative artificial intelligence technology, they were able to achieve this for 65 percent of customers,” describes Kain. “They can serve customers through automated channels, and even for those who need to serve them, they are able to serve them 10% more on their first phone call.”
Rocket Mortgage, a mortgage lending platform, was up about 33%. an increase in the percentage of customers who complete the loan process. The Stripe payments platform, on the other hand, reduced the manual onboarding time for sellers by 70%.. “We found that their support staff could do three times more for customers. It's like a complete transformation,” says Kain.
AI therefore becomes an element of the operational infrastructure, allowing business scaling without a proportional increase in employment. “The ability to automate things like customer service was an area where we saw great results,” Kain points out. In the short and medium term, the biggest changes can be expected in the areas of: customer service, claims processing, onboarding, fraud detection and regulatory documentation management.
Stablecoins as the next payment track
One of the most interesting topics raised by Kain is the distinction between cryptocurrencies in the strict sense stablecoins, i.e. digital tokens linked to a specific currency or basket of assets. In his opinion, stablecoins should be treated primarily as an additional payment track, offering quick settlement and an alternative to bank transfers or card payments, with a different cost and risk profile.
“Stablecoins give customers a new payment avenue that they can choose if they want. Fast, instant settlement. Kind of like today's instant payments, which is different from using credit cards,” Kain explains. But he adds an important caveat: “I don't think generative AI will change the payments infrastructure itself. AI is the key to helping manage fraud, onboarding customers, effectively routing those transactions to the right place.”
Kain emphasizes that the role of AI is rather to support fraud risk management, customer onboarding and intelligent transaction routing between various channels, including potentially stablecoins. “We see the same thing in the payments industry as customers have choices from a payments perspective, and we expect that to continue,” he says.
However, when it comes to faster adoption of stablecoins, Kain points to regulatory barriers. “We are very excited about this potential. I think there needs to be more regulatory clarity on how to define the type of support for a stablecoin. But I think when this comes to the fore, our clients seem to be very well positioned to take advantage of it” – he explains.
Regulation and trust: a European approach
Both Allianz and AWS emphasize the importance of the regulatory framework – albeit from different perspectives. Malmendier assesses regulations on operational resilience and security (e.g. DORA) as burdensome, but beneficial in the long term. “Even if it's not us, but others, it will still damage trust if the industry does not adhere to high standards” – he says, pointing to the systemic dimension of the issue. “I think for us, trust is essential for our economy. “I think overall I would really value the topic of trust and that customers feel like they can trust financial institutions and how they deal with AI and technology is more important than maybe having a little more freedom in that way.”
AWS, in turn, invests in tools with built-in control and management mechanisms based on an agent-based approach to AI. Kain explains: “Our clients are very thoughtful about assessing the risks of generative AI solutions, especially when it comes to regulated sectors such as financial advice. This, what they do intelligently is both build a strong governance process around the entire generative AI infrastructure“.
He believes there are ways to safely implement AI even in regulated areas. “If you're giving investment advice, people are still in the loop to make sure it's accurate. But you can still use generative AI to determine intent. What is my client calling about? What are they most likely to be doing? And then I can serve them with very specific information,” he describes of the hybrid model.
Kain emphasizes that AWS is investing in reducing the so-called hallucinations – incorrect answers generated by AI without a factual basis. “We are examining investments in automated reasoning, the ability to actually accept documentstransforming them into mathematical code and ensuring that the agents' output is actually correct. And when you are able to do this, you can see 99 percent. reducing hallucinations,” he says.
Applications in credit analysis and market research
Tools based on generative AI today process large volumes of corporate documents, reports and market data, building a preliminary analysis of the impact of specific events on cash flows and company valuations. Kain points to customer experiences such as Yahoo Finance or Prudential.
“If I understand how a strategy works in US markets, shouldn't I try it in European markets? And shouldn't I see how it works in Asia?” – he asks rhetorically.
He also gives the example of Moody's, which built a system consisting of several dozen agents for preparing credit notes. “Moody's analyzes companies using 29 different agents. It looks at the company, its outstanding debt, its credit ratings, and then puts it all together. A process that took many weeks was shortened to just a few hours” – describes Kain.
