There are many steps before a company can begin to implement AI in a cohesive way, from integration to data handling to security. Asia Insurance Review spoke to Boomi’s Mr David Irecki about the various concerns an insurer might have to deal with before actually using AI.
As an integration and automation platform, Boomi helps its clients connect various tools to create a cohesive environment, which also includes automation and real-time decision-making. Boomi chief technology officer for APJ David Irecki spoke to Asia Insurance Review about the use cases that he has seen in the insurance industry.
“When we look at the integration challenges and the automation challenges for insurers, they are trying to integrate purpose-built systems with applications that are available on the market to solve specific business problems. With so many systems in place, it is essential for them to integrate those systems together and automate those use cases,” he said.
Some clients he has worked with also needed to be able to scale and expand their business, trying to achieve more high-volume, low-complexity underwriting delivered in sub-second timeframes, while also returning multiple options to its customers. All of this required complex software, typically from multiple service providers, that had to be integrated and connected into a seamless whole.
AI and ChatGPT
Mr Irecki also noted that many clients had already adopted ChatGPT internally, which prompted the platform to design its own version of GPT. “Many customers, especially as a lot of these technologies have been democratised across an organisation, want to be able to build these automations and integrations themselves through a conversational interface, which is what our GPT allows,” he said.
“If you go back five or six years, there was a lot of discussion around APIs and the API economy and API sprawl. We believe the same thing will happen with AI. There are going to be AI agents that do very specific tasks for very specific things,” he said. “In the insurance industry, the simplest agents might come in the form of a chatbot or a virtual assistant to provide that first customer interaction, but we are already seeing agents for underwriting, for predicting future trends, for risk analysis.”
He called it an ‘agent garden’, essentially a registry or catalogue of AI agents that users could look through and discover for themselves the tools they would need for their business. “Many of the companies we talk don’t know where to start. While there are many companies playing and exploring AI, only about 4%-6% of businesses are actually in production and live with AI day today.”
First steps
According to Mr Irecki, there are a few steps before companies can even begin to implement AI, and it begins with understanding their data. “We ran a survey within our own customer base, which is about 20,000 customers globally, and we found out that about 60% of an organisation’s data is dark,” he said.
Data becomes even more vital when it comes to large language models such as ChatGPT, especially when the AI begins to ‘hallucinate’. “We also had news about proprietary information being uploaded for public distribution on ChatGPT, and there are a slew of data culture and data literacy concerns surrounding it,” he said.
He emphasised that understanding data is necessary to solving these issues. “You also need to understand the security concerns around that data. Especially for insurers, who have large volumes of sensitive information and a regulatory need to safeguarding that data. The integrity of that data is crucial because any breaches or mishandling of that data can lead to significant financial and reputational damage,” he said.
Practical AI
Businesses are typically bimodal – the systems of engagement at the front end and the systems of record in the backend. But Mr Irecki pointed out a third part to a company’s architecture, which is the systems of analysis.
“These have been data lakes, various machine-learning capabilities and deep learning capabilities that have been around for a very long time. But you’ve always needed a team of data analysts or data scientists to understand that data and report back to the organisation,” he said. “But now with AI, those systems of analysis have turned into what we term systems of intelligence. AI is able to go through that data in real time and through integration and automation, be connected back to those systems of engagement and systems of record so that they can act upon that insight in real time.”
Underneath all of that, a business would need integration to connect all those systems together, data governance capabilities such as data synchronisation and master data to ensure its data is correct and API management to secure the data that is going in and out of those large language models.
“All of those traditional tools that people have used over the years to connect systems of engagement to systems of record, can now be used to connect systems of intelligence as well. And that is the practical AI story, because if you get your systems connected together and your and you get your data correct, then you’re going to have a much better outcome,” he said.
Upskilling
The introduction of these new agents also necessitates the introduction of new skills, and many governments have set aside budgets specifically for upskilling and retraining. While there is the understandable fear that AI would replace humans in the workplace, Mr Irecki has noticed over the years that humans typically have very specialised business knowledge that cannot be replicated by AI.
“It’s not their data entry skills. It’s understanding how the business uses that data and works. And so being retrained to provide higher level outcomes for the business is typically what happens,” he said. “There was a lot of talk around requiring specialist prompt engineers who would have to craft the right answer to ask a large language model to get the right outcome. But what we’ve been seeing over the last six to nine months is those skills are being democratised across the organisation. Everybody is getting access to a ChatGPT-style of tool and learning how to ask the questions in the right way, provide the right context, which is relevant to their organisation.”
He added that a lot of these new AI tools and technologies can be picked up very easily and that there is no centralised control for them. Businesses now want capabilities and tools that allow a user in any department to be able to use AI, but still allow IT governance around what data is provided to them, and how the AI tools are used. “It’s going to be a little bit of a mix, but definitely there’s training to be had across the board,” he said.
He said that organisations must be ready to embrace responsible AI from beginning to end, as AI holds tremendous potential. “We want business leaders to understand the risks and opportunities of AI better before, but whatever technology you as a business are going to use to explore AI and build new AI-powered business solutions, you need to do them in a way that doesn’t compromise your organisation’s mission or ethical standing.” A