[ad_1]
Applying the model in finance
Risk management
Starting with risk management, there was no shortage of lessons from the global financial crisis.
One of the greatest privileges of my career was to be seconded from Morgan Stanley to work with U.S. Treasury Secretary Hank Paulson and many of you at the Federal Reserve during that painful period. I will never forget: after working with the U.S. Treasury and the Federal Reserve on Fannie Mae and Freddie Mac, I went home on the Sunday after the Lehman bankruptcy — only to get a call to come back down to the NY Fed. We learned that AIG would be out of liquidity by Wednesday due to counterparty risk in a derivative subsidiary in the UK.
That led me to the very important lesson that I have been repeating ever since: You cannot drive a car with mud on the windshield. You cannot run any institution with mud on the windshield. You need to clear it away — which is what data and risk analytics provide.
AI and ML can be a powerful assist in “clearing the windshield.”
Consider anti-money laundering. Instead of a single flagged transaction triggering a manual investigation which consumes hours or days to analyze, AI sees patterns across the entire network in real time. With AI, you can spot a newly created business account with dozens of small deposits from across the country — each just under the $10,000 reporting threshold. It then cross-checks the business director’s name against global sanctions and media reports, flagging a hidden link to a high-risk entity. Then alerts the security team before the funds are wired offshore. One of Google’s banking clients saw three times more financial crime risk detected, 60% fewer false positives, and a 50% faster path from detection to action.
Another critical example of early detection and risk management is around cybersecurity. The bad news is that attacks on financial institutions — already a primary target — are on the rise. The good news is the tools exist to counteract these threats.
At Google, cybersecurity is sacrosanct — embedded deep in our culture. Google was built from the ground up with an intense paranoia about anyone breaching personal data, and with a maniacal focus around every element of fortification, including what’s called “zero trust,” mandating strict identity verification and authorization for every access request, regardless of whether the user or device is inside or outside the network. That is what we use with our enterprise partners — public and private sector.
Yet we know that we have to keep upping our game, because the bad guys do. That is why we continue to invest massively in cybersecurity. That is why we acquired Mandiant. Mandiant’s insights underscore the need for a layered security approach — robust multi-factor authentication, constant patching, and better internal detection capabilities. Right now, the global median “dwell time” — meaning the time to detect an incident — is 11 days. The faster an incident is detected, the less damage an attacker can inflict.
That is also why we are leveraging Google DeepMind — Google’s AI research laboratory. Recently we had a breakthrough in applying AI to security threats that shows what is possible: Google DeepMind introduced an AI agent called Big Sleep. That agent was developed to proactively hunt for unknown software vulnerabilities. We were pleased to see it find its first real-world flaw in November, proving AI can plug security holes before they are exploited.
Since then, it has uncovered multiple critical vulnerabilities, including a recent one that was known only to threat actors. Using intel from Google Threat Intelligence, Big Sleep predicted the flaw was about to be weaponized — and shut it down. This is believed to be the first time AI has directly stopped a live exploit in the wild. Google is now applying this breakthrough beyond Google, helping secure open-source software across the internet.
Google will continue to push the frontier, because this threat is growing, and profound — both financial and reputational. I firmly believe that AI must be applied to keep up with the growing threat horizon.
Operational effectiveness and efficiency
In operational effectiveness and efficiency, I will call out three examples:
First, AI applied to customer support is increasingly seen as the gift that keeps giving. It provides operating leverage, clearing away the most basic questions, allowing customer support professionals to focus on the more complicated issues. More important, historically, when a call came into customer support, it did not capture the nature of the queries and analyze the root cause. Here, anyone implementing an AI contact center is actually able to harvest the insights from the customer center. I was recently with one CEO who said that once the company rolled out AI in customer support, employees indicated they wanted more of it. They essentially said, “This has eliminated the drudgery. Our teams have more time to think.”
Second, one of the most powerful new tools is something Google built called NotebookLM. With NotebookLM, you can load reports, articles, videos, audio files into it — and it will ingest and analyze. You can ask it to spot trends, or pluck out details. You can have it synthesize results and deliver them to you in the form of a podcast, which you can interrupt with questions.
Third, organizations are seeing an extraordinary uplift in developer efficiency and productivity with AI and tools like coding assist, which is something Google and many others are offering. We hear this repeatedly from banks as a priority.
Innovation and growth
The final category: AI can support and accelerate growth opportunities — to add alpha.
That innovation can be in the tech stack itself, by which I mean chips, models, and applications. We are talking to a number of banks and other financial services companies about the application of TPUs within their system to enhance their trading operations.
In the applications, it can come in the solutions that allow organizations to get closer to customers — especially with agentic AI. One example my colleagues are excited about is the opportunity to address so-called HENRYs: High Earners, Not Rich Yet. Top-tier financial institutions employ tens of thousands of financial advisors who dedicate their time to knowing, understanding, and serving high-net-worth clients. However, that white glove model cannot scale to address the fastest growing wealth segment — the mass affluent and “HENRYs.” That means there is an “advice gap” and a missed opportunity to build relationships with the next generation of top clients.
This is a huge opportunity to deploy agentic AI — systems that combine the intelligence of advanced AI models with access to tools, so they can hoover in information and understand, reason, and act across complex workflows — and take actions on your behalf and under your control. This enables advisors to assemble hyper-personalized recommendations for clients, reducing human preparation time by over 60%, which, again, frees advisors to focus on building relationships.
[ad_2]
Source link
