Generative AI in Practice is InsTech’s monthly newsletter dedicated to the use of generative AI in insurance.
The InsTech Perspective
Are you hiring or firing?
This month, the world of AI watched a dramatic firing and hiring episode. On Friday 17 November, OpenAI fired CEO Sam Altman. By Monday, he had been hired by OpenAI’s main investor, Microsoft. Then, nearly all OpenAI’s employees threatened to quit unless Altman was reinstated; Microsoft had offered to hire them all on their existing salaries. On Tuesday the former CEO was re-hired.
A leading consultant recently told us that one common question insurers ask about generative AI is “whom can I fire if it goes wrong?” If a human makes a mistake, it is clear who is responsible. When AI is involved, the lines of accountability blur. To use generative AI successfully, insurance companies need to build a team that can harness new technologies while maintaining accountability. This is no mean feat: the revolt at OpenAI and Microsoft’s bold job offers prove that the most AI-proficient workers can easily find another job.
Use case of the month
Your in-house ChatGPT
This newsletter tends to focus on how generative AI can change core insurance processes, such as submissions ingestion, claims handling or customer service. But for many people today, their use of generative AI starts and ends with ChatGPT. The chatbot can be a useful assistant for drafting emails and documents, summarising text, getting answers to simple questions and many other tasks.
Insurance companies can benefit from their employees using such a tool to become more productive at work. However, they are often wary of employees putting confidential information into this chatbot. Although OpenAI has certain safeguards to prevent confidential data from being used to train its models, companies prefer to have more control over their own and their clients’ data. Some companies have explicitly banned employees from using ChatGPT at work.
Several companies including AXA, Marsh McLennan and Nationwide have now launched privately hosted equivalents to ChatGPT for their employees. This involves hosting the AI model behind ChatGPT on Microsoft’s Azure cloud infrastructure, which many companies already use for computing and software. Instead of ChatGPT prompt messages being sent to OpenAI’s servers, the questions and answers in each interaction remain within the organisation’s Azure environment. Early adopters at Marsh McLennan say the chatbot helps them save eight hours per week. Allianz is planning to do the same and we expect more insurers to follow, working with Microsoft, Google or other providers.
What the insurance industry is doing with GenAI
Marsh McLennan has rolled out a generative AI tool, using a privately hosted version of OpenAI’s models, to its employees worldwide.
Allianz Commercial is working with Microsoft to explore generative AI use cases including chatbots that can answer questions about its risk appetite and summarise exposure data.
Aon says insurers are revising policy language related to AI, including addressing ‘silent AI coverage’ (where policies that do not explicitly mention AI may expose insurers to AI-related losses). Could insurers’ approaches to AI risks echo how they have dealt with cyber exposures in recent years?
Insurer Nationwide has provided all its employees with access to Microsoft’s Bing Chat Enterprise, a privately hosted generative AI tool connected to the internet.
Insurer Helvetia is now using generative AI in its self-service chatbot for Swiss customers, after a successful pilot.
Most insurers’ experiments with GenAI involve OpenAI’s models, but Luxembourgish insurer Lalux says it is also testing a French LLM.
News from the InsTech network
Eigen Technologies has launched an updated version of its no-code data extraction platform. Among the new features is a GenAI tool to help companies gather insights from documents.
Sprout.ai has raised £5.4 million GBP of investment. You can find out how the company uses generative AI to automate claims processes in our recent interview with the CEO.
Microsoft joins Indico Data’s podcast to discuss how GenAI can help insurers become more efficient without compromising security.
78% of insurers lack a clear strategy and the in-house capabilities to operate AI in production, according to Simplifai in a recent article.
Wenalyze, which provides information about SMEs to insurers, is now using generative AI to gather data from disparate sources such as regional property registers and classify businesses more accurately.
WTW writes about the risks associated with using AI in healthcare and how to manage them, including the insurance implications.
Other news and insights we’ve seen
We reported last month that AXA’s survey of experts ranked AI as the fourth biggest future risk. But it’s not as much on the radar of corporate risk managers. In Aon’s Global Risk Management Survey, risk managers rank AI risks 49th in 2023 and expect them to rise to 17th by 2026.
GenAI chatbots are changing how consumers research insurance, creating new challenges and opportunities for insurers’ marketing teams, according to SEO & media experts at New York Life.
Insurers UnitedHealth and Cigna are being sued for their use of AI in denying claims.
73% of UK insurance executives expect generative AI will take their jobs. Nearly half viewed GenAI technology as a ‘foe’, not a friend.
People who don’t have health insurance are more likely to ask generative AI bots for healthcare advice, according to Deloitte.
AM Best says AI is not the main cause of recent layoffs at large insurers, noting ‘the proliferation of AI has led to new job creation’.
Are your tech sales cycles changing? At least one AI vendor has seen insurers buying technology weeks after initial contact with vendors, when it used to take a year, according to Insurance Thought Leadership’s Paul Carroll.
Forrester predicts a major insurer will offer insurance specifically for AI hallucination in 2024. We wrote last month that offering performance guarantees for AI products could prove a profitable opportunity for insurers.
How InsTech has used AI this month
Humans in InsTech’s research team are behind our new report released on Tuesday, but generative AI helped us make the report’s presentation as compelling as its text.
First, report editor Ali Smedley used ChatGPT to generate suggestions for the title, settling on “Beyond the smoke: understanding and mitigating wildfire risk”. Then, our designer Charlotte Salisbury was tasked with designing the front cover image. Together with the title, the front cover forms people’s first impression of the report.
For this report, it was particularly challenging to find an appropriate image consistent with past report front covers (we use a wooden toy theme) and available for commercial use. Charlotte used Adobe Firefly, a generative-AI-powered image maker. Firefly is trained on Adobe stock images, openly licensed and public domain content. It generates images based on a text prompt, along with an optional image reference for colour and style.
The generated image itself was not perfect, but with some tweaking in Photoshop, we had a unique image ideally suited to the report. The AI tool saved Charlotte hours in manual editing time. We’re also pleased that image creators still benefit, as Adobe has created a compensation model when it uses creators’ content to train and retrain Firefly.
You can download the report here.
Find out what you’ve missed…
Issue 4: Catastrophic risks, insuring AI and pet insurance insights
Issue 3: Claims automation, billion-dollar spends and AI-generated business plans
Issue 2: Distribution at scale and GenAI use cases from Lloyd’s syndicate MIC Global
Generative AI in Practice is InsTech’s monthly newsletter dedicated to the use of generative AI in insurance – you can sign up for free here.