Can insurance trust AI?

A word from the CEO

AI is reaching a new milestone

At trade fairs, conferences and seminars alike, AI is now unavoidable. Behind this heavy media coverage lies a genuine technological shift: the emergence of generative AI alongside traditional predictive AI. Making full use of AI’s potential in the insurance sector requires understanding this shift.

• Predictive AI is a powerful tool, trained very specifically for one particular task. It is designed to make predictions about future events by analysing past data.
  For example, predicting claims rates by analysing historical customer claims data. Its development and integration are complex and take several months before it is operational.

• Generative AI brings new characteristics and, with them, new opportunities:

• Pre-trained models: Generative AI uses models already trained on large bodies of data, which allows it to handle a wide variety of requests without requiring specific training for each new task.

• Fast deployment: Using pre-trained models allows faster deployment, as they do not require the same amount of specific data and development time as predictive AI models.

• Adaptability: Generative AI can adapt to different contexts and questions, offering answers and solutions more flexibly.

Generative AI is therefore designed to create new data or content similar to what it was trained on. For example, writing an article on insurance trends in 2024 based on existing articles.

Everything you need to know about predictive and generative AI.

Artificial intelligence, real risks

Easy to integrate and highly effective, generative AI is presented as the solution to almost everything. It is indeed hard not to get excited about such a technological leap. Many discussions around AI currently focus on its ability to automate customer interactions. Some go further and entrust entire processes to AI. The idea is tempting, but not without risk.

The new applications of generative AI are exciting! However, not all these uses are equally mature, and they are not necessarily suited to every sector. Automatically summarising a video meeting, generating code, or advising on insurance products: can these tasks really be put on an equal footing? Do they carry the same level of risk if a spanner gets thrown in the works? Are we sometimes doing AI just “for the sake of AI”?

“ENTRUSTING ENTIRE PROCESSES TO AI IS TEMPTING, BUT NOT WITHOUT RISK”

Glue on pizza, when AI doesn’t stick to the facts

AI has only one mission: to provide an answer to a question. It does not really know what is right or wrong. Recently, a web user wanted to know the best way to stop cheese from sliding off his pizza. To get an answer, he asked Google’s AI, a tech giant from which we expect reliable, high-performing services.

On several occasions, the artificial intelligence advised him to use… glue. The advice makes us smile. Human common sense lets anyone realise that this advice should not be followed. Let us now transpose this problem to the insurance sector. What would happen if a completely far-fetched piece of advice or recommendation were given to a customer or prospect? Would we still trust our interlocutor just as much?

Saving policyholders’ trust

The real risk is giving AI too large a place in the value chain. Not supervising it, not checking its proposals and assigning it tasks that do not suit it.

Trust is at the heart of the relationship between policyholder and insurer. AI must be used in a way that does not jeopardise this essential trust. This is all the more true in a context where legislation tends to strengthen consumers’ freedom to switch insurance company. We will come back to this topic soon.

AI is a formidably powerful tool that offers us many new prospects for innovation. To make full use of it, it is important to regard AI as a tool serving people, not as a full-fledged player in insurance.

AI to make sure… you save time

At this stage, the main exploitable benefit of generative AI is improved operational efficiency, data analysis and personalisation of services. By using AI for these tasks, professionals become able to focus on higher-value tasks, such as strategic decision-making and human interaction.

Let us take a concrete example: a car insurance policy.

A major part of the work is to formalise a series of data so that operational chains can run without errors. The policyholder’s details, the make of the car, the model, the power, the claims history, the driver’s occupation and so on must all be encoded.

Currently, this data is collected through a series of online forms, and it is up to the customer relations officer to pick out the relevant information from the various documents (identity card, claims history, purchase invoice, customer email, previous contract, …). This is precisely the kind of task at which AI excels: “understanding” the information sought, finding it across multiple sources and then submitting it to the policyholder for validation.

Including this kind of process during a customer contact would let the customer service officer focus on the relationship and be freed from tedious administrative work.

What impact on sustainability?

Let us remember that not every innovation is necessarily progress. It is worthwhile, even essential, to put AI’s economic return into perspective with its impact on the carbon footprint.

The growing cost of these technologies and the polluting emissions linked to them are hard to reconcile with commitments to reduce carbon footprint or to sustainability.

Some models predict that AI will consume up to 25% of all the electricity used in the USA in 2030. Sam Altman (founder of OpenAI) himself says that a technological breakthrough in electricity supply will be essential to support the development of AI. Can we afford to be ultra-dependent on AI under these conditions?

Generative AI is an extremely interesting little revolution that opens up many new possibilities. It is up to us to use it in the most appropriate way!