One of the world’s largest risk advisors and insurance brokers launched a digital transformation five years ago to better enable its clients to navigate the political, social, and economic waves rising in the digital information age.
Paul Beswick, CIO of Marsh McLennan, served as a general strategy consultant for most of his 23 years at the firm but was tapped in 2019 to relaunch the risk, insurance, and consulting services powerhouse’s global digital practice.
His first order of business was to create a singular technology organization called MMTech to unify the IT orgs of the company’s four business lines. As part of its multifaceted manifest, MMTech, which Beswick leads from Boston and employs roughly 5,000 today, undertook a wholesale organizational transformation to better align all four businesses and establish a common technology platform for its digital future.
“The process essentially was to merge all of these teams into one and start to build a culture of one team looking for opportunities to share, collaborate, reuse, align on standards, and break down some of the frictions that have existed between those organizations historically, which had slowed down the rate at which we were able to get things done right,” says Beswick, who in management consulting had led about 500 partners with a budget of about $500 million. He initially turned down the CIO job but was persuaded to take it up by the prospects of leading Marsh McLennan on this digital journey.
The professional services arm of Marsh McLennan advises clients on the risks, shifts, and challenges facing the modern enterprise, most poignantly the vital role technology now plays in business and on the world stage.
One of the firm’s recent reports, “Political Risks of 2024,” for instance, highlights AI’s capacity for misinformation and disinformation in electoral politics, something every client must weather to navigate their business through uncertainty, especially given the possibility of “electoral violence.”
“The link between elections and civil unrest is well understood; given the number of elections in 2024, there is a correspondingly elevated risk for electoral violence,” the report states. Compounding this risk is a new and poorly understood factor: the potential for AI to amplify political misinformation and disinformation. “The US government has already accused the governments of China, Russia, and Iran of attempting to weaponize AI for those purposes.”
Re-platforming to reduce friction
Marsh McLennan had been running several strategic data centers globally, with some workloads on the cloud that had sprung up organically. As part of MMTech’s unifying strategy, Beswick chose to retire the data centers and form an “enterprisewide architecture organization” with a set of standards and base layers to develop applications and workloads that would run on the cloud, with AWS as the firm’s primary cloud provider.
But the CIO had several key objectives to meet before launching the transformation.
First, the misalignment of technical strategies of the central infrastructure organization and the individual business units was not only inefficient but created internal friction and unhealthy behaviors, the CIO says.
Marsh McLennan comprises four business units: Marsh, the firm’s insurance broker; Guy Carpenter, a reinsurance broker that offers insurance to insurers; Mercer, the firm’s second biggest business, handling health, benefits, wealth management, career development, consulting, HR, and training for enterprises globally; and Oliver Wyman, the firm’s management consulting arm, where Beswick began working as a summer job in 1995.
To address the misalignment of those business units, MMTech developed a core platform with built-in governance and robust security services on which to build and run applications quickly.
The idea, Beswick says, was to enable the creation of an application in days — which set a. high bar.
“It’s a full-fledged platform … pre-engineered with the governance we needed, and cost-optimized. At the current stage, if you are setting up a new application, we have a simple launch site and [after] entering in the details, you can have something up and running with a code repository and secret store connected to multifactor authentication running on our cluster in 20 minutes,” Beswick says.
The MMTech team continues to enhance and use the core development platform with all critical services built in, and while not all applications can be built in days, “you can genuinely stand up and get things through the process in under a week and that has been part of the engineering out all of the friction,” Beswick says.
Setting the standard for analytics and AI
As the core development platform was refined, Marsh McLennan continued moving workloads to AWS and Azure, as well as Oracle Cloud Infrastructure and Google Cloud Platform. Several co-location centers host the remainder of the firm’s workloads, and Marsh McLennans big data centers will go away once all the workloads are moved, Beswick says.
Simultaneously, major decisions were made to unify the company’s data and analytics platform. The firm had a “mishmash” of BI and analytics tools in use by more than 200 team members across the four business units, and again, Beswick sought a standard platform to deliver the best efficiencies. The team opted to build out its platform on Databricks for analytics, machine learning (ML), and AI, running it on both AWS and Azure.
Configured based on Marsh McLennan standards, the Databricks platform provides central control, while continuing to evolve as the firm’s crown jewel for both analytics and AI, with data cleaned and assembled for use, the CIO says.
“This is for people in the organization who have data and want to drive insights for the business and for their clients,” Beswick says. “I want to provide an easy and secure outlet that’s genuinely production-ready and scalable. The biggest challenge is data. It’s very fragmented, ownership is often unclear, quality is a variable, but we have teams really working on that and generating data faster than we can possibly catalog and clean up.”
Marsh McLennan has been using ML algorithms for several years for forecasting, anomaly detection, and image recognition in claims processing. With Databricks, the firm has also begun its journey into generative AI. The company started piloting a gen AI Assistant roughly 18 months ago that is now available to 90,000 employees globally, Beswick says, noting that the assistant now runs about 2 million requests per month.
Beswick is also preparing for extensive generative AI activity within the company based on Microsoft’s implementation of OpenAI, which offers security to his liking. The CIO is quick to point out that Marsh McLennan’s gen AI platform — like its development and analytics platforms — uses industry-standard products but its interface, tooling, core services, and enhanced capabilities, which go “beyond what the model can do on its own,” were built by MMTech at the company’s innovation center in Dublin, Ireland.
The platform include custom plug-ins to Word, Outlook, and PowerPoint.
“It’s our own version of Copilot, effectively,” Beswick says, explaining the other core advantage of the company’s standards approach. “This costs me about 1% of what it would cost” to license the technology through Microsoft.
Gen AI agenda
Beswick has an ambitious gen AI agenda but everything being developed and trained today is for internal use only to guard against hallucinations and data leakage. ML and generative AI, Beswick emphasizes, are “separate” and must be handled differently.
He estimates 40 generative AI production use cases currently, such as drafting and emailing documents, translation, document summarization, and research on clients. MMTech built out data schema extractors for different types of documents such as PDFs.
The company also built an avatar for its gen AI platform dubbed “Len Ai,” which allows employees to analyze data using natural language “so you can handle a few hundred thousands of rows of data, describe what you want, and the AI will figure out how to write queries that get that analysis done for you,” the CIO says, adding the tech team built a micro Len AI that serves as a scaled-up version of a Q&A chatbot for specific knowledge bases.
“Gen AI is quite different because the models are pre-trained,” Beswick explains. “It’s incredibly flexible, and accessing new capabilities just through prompting and prompt setup is massively easier and requires much less deep data science capabilities than classical machine learning models.”
Marsh McLennan created an AI Academy for training all employees. Within the past year, Beswick and his team have given tens of thousands of employees a range of gen AI tools to help them be more productive in a “diffuse” manner. And it’s been very valuable — saving about 1 million hours of labor or so, Beswick says, adding that value is hard to pin down.
Gen AI has been effective for a lot of “drudge” work, but replacing human expertise — especially in matters of advisory services — is still a ways off. The CIO is focused on creating building blocks that will enable the company to make the leap when it’s time, but given the company charter, he takes risk very seriously.
Still, generative AI, which is shifting to the “automation frontier,” will be worth millions in value as it evolves, he says. He continues to work with the C-suite, but AI work is driven out of MMTech, and he recognizes it is an amazing opportunity for IT to get on the ground floor with technology that all business units are demanding and know will be very valuable.
Beswick intends to lead the charge to a new era for the entire company. “It is possibly a once-in-a-career opportunity,” Beswick says. “Technology organizations get credit for delivering certain projects, and get credit when there is a crisis. But being able to pick up a new technology and drive out on the front foot in a corporation like ours is very rare and hugely valued by the company.”
Additional gen AI transformations:
Read More from This Article: Marsh McLennan IT reorg lays foundation for gen AI
Source: News