Marsh McLennan CIO Identifies A Fast Track To Drive GenAI Value

January 11, 2024
BY Peter High Founder and President of Metis Strategy
Icon Scrolling Bar

0%

Marsh McLennan CIO Paul Beswick and his team have made substantial progress in the adoption of generative AI tools and believes that he and his team are setting a sound foundation. but even higher levels of value will be achieved by building upon that foundation. “Very quickly after we had access to the [Microsoft OpenAI] APIs in a secure fashion, we created the chat interface on top of that, which is what we call LenAI.”

Paul Beswick’s pathway to become Chief Information Officer of Marsh McLennan is non-traditional to say the least. He joined the company via one of its operating companies, the strategy consultancy, Oliver Wyman. He rose through the ranks at that firm to become a partner and global head of Oliver Wyman Labs and global co-head of Oliver Wyman’s Digital Practice. “It all happened because I walked into the wrong meeting one day and got sucked into the project to design Marsh McLennan’s technology strategy with Scott Gilbert, who was my predecessor,” noted Beswick. “Then got sucked into trying to deliver it, which anyone who’s been a consultant should know, you should never do: you should never both write the strategy and take responsibility for delivering it.”

When he weighed the advantages of his pathway, he noted that he has been in executive committee meetings and boardrooms since he was in his 20s. He also acknowledged that having profit and loss responsibilities in various roles along his ascent at Oliver Wyman likely gives him a better appreciation of technology’s power to grow revenue for Marsh McLennan, not just where it can lead to cost savings.

Beswick’s current post has him overseeing technology for a conglomerate that includes Marsh, the world’s largest insurance broker, Mercer, leader in human resources, benefits, and investment consulting, Guy Carpenter which is in the business of reinsurance broking, along with Oliver Wyman. When he took over as chief information officer roughly three years ago, the company was in the throes of moving from a decentralized IT department to one that exerts much more influence from the center. “That’s a fairly new development in terms of how we’ve been organized. When I took on this role, we were starting the process of bringing what had been business unit-specific technology organizations together into one overall organization,” said Beswick. “Prior to that, we’d had different teams by business, but with a shared infrastructure and security organization in the middle. It’s been an interesting journey trying to forge one team out of what were quite independent teams before.”

Beswick sees a primary job of his as increasing the velocity of the business. “We do a lot of work to understand what slows us down, how we get tangled up in our own processes, where there’s bureaucracy that’s unnecessary, where we fail to engineer solutions to problems that we can engineer solutions to that can help things move significantly more quickly,” he underscored. “A huge chunk of where I spend my own time…is focused on trying to change the efficient frontier between speed, agility on the one hand, and security, compliance, robustness, and resilience, on the other.”

A primary pathway to this for Beswick and his team has been in building a platform strategy, building template projects and defining “patterns” that can be deployed readily, streamlining policy, compliance and nonfunctional aspects of every project that his organization undertakes. “One of the things I’ve learned as I’ve come into this job is how important understanding some of the organizational dynamics are and the points of inefficient but stable equilibrium that exist in organization structure that tend to lock you into patterns that are inefficient and thinking very deliberately about how you break through some of those things,” he said.

Beswick is excited about the amount of innovation driven by technology and his team’s ability to convey the art of the possible to the rest of the company. He thinks about technology in the spectrum of hard things to easy ones. “We are not in the game of doing really hard stuff,” he said. “That’s not the organization that we’re built for, but hard things get easier over time, and there’s this constant shift from more complicated and less accessible but powerful technology into things that are increasingly easy to get our hands on. At some point, there’s this tipping point where the hard becomes easy. If we can be there at the point where things become easy and we understand how to put them into action in a real business against our real processes and our real problems, that’s the area where I think we can create the most value. That requires you to be always playing around at the edge of that transition point and make sure you recognize when that transition has happened.”

A case in point is Marsh McLennan’s foray into generative artificial intelligence. It began by partnering with vendor partners, but that proved to be too expensive. However, when Microsoft made the OpenAI back-ends available in a secure fashion, Beswick and his team discovered that with a little bit of extra engineering, they could make that available to the broader company. The goal was to mirror the remarkable uptick in the use of ChatGPT in society. “I didn’t think we needed to spend a lot of time worrying about precisely what the use cases were,” Beswick admitted. “It felt like the use cases would be emergent. Very quickly after we had access to the [Microsoft OpenAI] APIs in a secure fashion, we created the chat interface on top of that, which is what we call LenAI.”

It took only a day and a half to deliver the first version of LenAI to a pilot group within the company. The focus on making IT a driver of velocity improvements were responsible for such a fast path. Soon a few hundred people had access to LenAI and within 28 days, the entire firm had access to it. “I think we’ve identified [roughly] 300 distinct use cases that people have been putting this to,” said Beswick. “Some are very specifically related to some small part of the business. Others are more generic. We’ve kept an eye on that, capturing that information, and we’re using that to then drive our build-out agenda for some of the things that are going to be more scalable implementations of this.”

Beswick believes his team has moved farther faster by turning the typical process on its head. Typically, people gather use cases, find a business co-sponsor, build a business case, assemble a project team, and then get started. Given Beswick’s need for speed, that was too slow. “By going the other way and driving something more generic out and flushing the use cases out, I think we’ve got further faster,” said Beswick with pride. As a result, “we added a couple of extra capabilities into the basics, [such as] internet search document upload. We do a lot of work with documents, so there’s lots of stuff people are doing with document summarization, with data extraction from documents and translation between languages, which these tools are good at. Email drafting, particularly for people for whom English is not the first language when we’re a business that largely operates in English [has been another powerful use case]. A lot of people are using it to tighten up their communications and streamline things.”

Code writing is another layer of value. Beswick noted with excitement that different parts of LenAI were written by LenAI. This will increasingly become the norm. Additional functionality that has been defined has included calculators, stock price lookups, weather lookups, database querying, and the ability to pull from a variety of news sources. “There are clearly some use cases where you can see transformation of various processes that we would run through today and would be fairly manual where we can really divert resources into much more high-value-added work,” said Beswick. “Those are starting to spin out. A lot of it’s around things like document ingestion, processing, and data extraction. Cross-mapping data from one data source to another, one data structure to another turns out to be a pretty tractable problem as well. I think we’re just scratching the surface as to what those sorts of things will be.”

Beswick and his team have made substantial progress in a short amount of time, living up to his goal of being a force multiplier. He believes he and his team are setting a sound foundation, but even higher levels of value will be achieved by building upon that foundation.

Interested in working together?

We’d love to hear from you.
contact us

Contact Us

    Icon

    Thank you for your submission

    We will get back to you as soon as possible. Back to site
    IT Support by SADOSSecure, Fast Hosting for WordPress