Building vs buying an AI solution: What should insurtechs consider?

While many insurtechs and insurers leveraging AI are considering whether to build or buy their solutions, an industry leader said the decision should ultimately be guided by how they intend to use the technology.

“I think insurtechs will be looking around and thinking [about] do we get to 100% coverage of AI supporting our business,” Erez Barak, CTO, Earnix, said.

While he believes a hybrid solution may be the best option, he suggested a general guideline for businesses to keep in mind.

“Map your business to where you have coverage and where you don’t. The more urgent, and I’d say more complex and deep areas, that’s where buy would probably trump the case. The more customized, more flexible, maybe not as complex but super important areas, maybe build is where you want to play,” he said.

Are Gen AI and AI the same?

One of the first considerations insurance businesses have to keep in mind is the distinction between AI and generative AI.

Artificial intelligence is a form of technology capable of interpreting human input and responding in kind, whether through actions or output.

Generative AI, on the other hand, wields creativity to create text, images and other types of content. It’s the specific form of AI being used by increasingly popular tools such as chatbots or deepfake software.

However, Barak suggested the distinction may not be as important as having both AI and GenAI work together to drive productivity.

“I think, broadly, you want intelligence everywhere. That’s where the world is heading, the ecosystem is heading and we’re heading as an industry. But I think different types of AI, GenAI versus others, will together serve the global needs and we won’t find ourselves comparing which one’s which,” he said.

Is building a solution feasible?

While the idea of building an AI solution may seem ideal, Leandro DalleMulle, VP and global head of insurance, Planck/Applied Systems, pointed out that it can come with downsides.

“In the earlier days of AI, building systems in-house was often a viable option… For industries that relied on bespoke solutions, building their own AI systems offered a competitive edge,” he noted.

However, the development of GenAI has “changed the game.”

“Training foundation models, such as large language models, is an entirely different level of complexity. These systems require enormous datasets… The cost of training a state-of-the-art GenAI model can run into hundreds of millions or billions of dollars, with ongoing expenses for fine-tuning and maintenance. For most organizations, this makes building GenAI systems from scratch financially and technically infeasible,” DalleMule said.

Simple vs complex AI needs

Despite the challenges, building an AI solution may be an option for companies who have specific needs and can function with simple models, according to Barak.

“In some other areas, building your own co-pilot, even using simple techniques or building some simple models that allow prediction, may make more sense with an internal data science team that can go faster, be agile in terms of changing needs and also provide a very customized solution,” he explained.

Businesses may be better off buying a solution for more complex and urgent needs, where there is less flexibility and more depth at play, Barak suggested.

Consider a hybrid option

Both experts noted that companies do not necessarily have to choose between exclusively building or buying, as they can opt for a solution that combines both.

“Companies are increasingly turning to ‘buy and customize’ strategies… Customization options, such as fine-tuning a model with proprietary data or integrating APIs, allow organizations to achieve tailored results without shouldering the astronomical costs of full-scale development,” DalleMule said.

“What you want to do is have this tapestry of areas, see where you want coverage first, look at the time dimension, the complexity dimension, the maturity dimension and build a portfolio of build versus buy,” Barak added.

He emphasized that there is no one correct answer for everyone, as some companies choose to buy everything while others choose to build everything. However, he said most of Earnix’s client base tends to take a mixed approach based on unique considerations and business priorities.

Assessment comes first

Although the build vs buy question is crucial, Barak said it’s secondary to determining how the company will leverage AI technology — which is increasingly becoming a question of “when” or “how” rather than “if.”

“My advice would be to work with the belief that 100% of your business is going to be supported by AI. You’re not going to be an AI business, but if you want to be in business, if you want to compete, if you want to innovate quickly, you want to have AI in every aspect of your business,” he said.

Earnix, founded in 2001, is a global insurance and banking operations provider based out of Israel.

Applied Systems is an insurance and technology software company founded in 1983 and based in Chicago. In July 2024, Applied Systems acquired Planck, an American insurance AI solutions provider.

© Entire contents copyright 2024 by InsuranceNewsNet.com Inc. All rights reserved. No part of this article may be reprinted without the expressed written consent from InsuranceNewsNet.com.

Rayne Morgan is a journalist, copywriter, and editor with over 10 years' combined experience in digital content and print media. You can reach her at [email protected].