82 EYEBROW “Companies are quickly progressing from using AI for individual productivity to employing agentic systems that support entire business processes,” says Bahnemann. “Two examples from Germany stand out. DOMCURA, a non-life underwriting agency, used Azure OpenAI and Azure Cognitive Services to build KIM – an award-winning AI-powered virtual employee that processes claims from submission to decision in around 15 minutes. KIM has now been commercialised as a white-label product for third parties, moving from an internal efficiency gain to a new business model. Meanwhile AURA, an AI-powered underwriting assistant built jointly by HDI Global, Microsoft and Reply, supports underwriters through document-heavy intake, policy reasoning and decision support – reducing manual effort in one of the most compliancesensitive processes in commercial insurance. “Together, these two cases illustrate what the partner ecosystem enables in practice: claims automation on one side, underwriting intelligence on the other, both built on Microsoft’s cloud and AI governance stack with domain-specialist partners.” Looking ahead, Bahnemann stresses the continued importance of cyber resilience, including important questions around issues like third-party dependencies, patch velocity and fallback planning. “In the coming months, many firms will remain focused on productivity and decision support, while cybersecurity enhancements, AI-supported vulnerability scanning and the automation of patching processes currently have the highest C-suite priority and regulatory attention,” he says. “In addition, organisations remain under pressure to modernise their data foundations because AI only works well if the data platform is secure, resilient and usable at scale. “If advanced models like Claude Mythos can compress the timeline from vulnerability discovery to exploit, then response speed, architecture discipline and visibility matter even more. That is why the compliance conversation should not stop at policy and contracts, but extend into security posture, operating model and resilience execution. Financial firms will need to defend at AI speed, not just build AI features faster. And this is exactly what our new Microsoft multi-model agentic scanning harness (codename MDASH) is built around. It is a multi-agent vulnerability discovery and remediation system, using more than 100 agents that are injected in the right combination with the right set of models in your software delivery workflow. This works as a structured pipeline that takes the code base and emits validated, proven findings to help organisations harden their own developed software products.” Keeping these considerations in play will enable the industry to continue its AI journey with ambition and confidence. “In the longer term, we are moving from AI as an assistant to AI as part of the operating model,” says Bahnemann. “Financial services organisations will not move all the way to autonomy overnight, and in many areas they should not. But AI will increasingly sit inside service operations, compliance workflows, cyber defence, software engineering, customer service and control functions. The firms that win will be the ones that combine productivity gains with trust, resilience and supervisor-readiness.” “ In financial services, confidentiality starts with identity, authorisation and trust boundaries – not with a generic promise that AI is secure” FEATURE
RkJQdWJsaXNoZXIy NzQ1NTk=