Florence, 42, divorced, raising two children. She is a doctor, tenant of her apartment and owns a car, which she has finished paying off. According to the analysis of her profile, she has an 80% chance of agreeing to subscribe to a property tax exemption product and it is preferable to contact her by email rather than by telephone to offer her an appointment. Here is the type of report that generative artificial intelligence could provide in the blink of an eye to a private banking adviser, based on his database and the analysis of his client’s expenses and resources. With even a pre-written message to send to him. “This enormous time saving can be reinvested in the interpretation and development of high value-added advice, underlines Sébastien Lacroix, from the consulting firm McKinsey. We can also imagine that customers themselves solicit these virtual experts for more complex searches than with current chatbots, which only react to the keywords used in the request. 00 euros”.
Also for day-to-day requests, generative AI is taking robocalls to the next level. The Truffle Capital fund strongly believes in the potential of the conversations of Zaion, one of the 16 fintechs in its portfolio. This language processing company can interpret the tone of the customer who picked up his phone to call his bank: complaint in anger after an incident or simply need a document? The answer will be adapted accordingly. “Major establishments are already Zaion customers, such as La Banque Postale, others will become so, assures Bernard-Louis Roques, co-founder and pioneer fintech investor at Truffle Capital. Last year, the company handled 1 million calls per month, today four times more. With a phenomenal scale effect!
The banking sector will be among the big winners of the world’s shift to generative AI, according to McKinsey. In terms of risk control too, this technology will find its raison d’être: automation of the detection of money laundering and fraud, research of information on a customer applying for credit, in order to ensure that he does not have activity in tax havens, or simply to assess his solvency.
Other possible applications: the translation of computer code or the comparative analysis of documents. In the highly regulated banking world, will the teams of laborers who for weeks noted the main points of evolution between a 1,000-page pad and its freshly updated version be relieved or worried to learn that robots are capable of carrying out similar work in a few tens of seconds? “Employees are not going to be replaced by AI but by employees who use AI. The machine in reality is not very intelligent, it is just very efficient”, tries to reassure Laurent Daudet, co-founder of LightOn. This Parisian AI start-up is particularly attractive to banks, concerned about data confidentiality, but also to players in the health, defense and human resources sectors.
Large companies are reluctant to send their data to OpenAI servers: some have already scared themselves, like Samsung, a handful of whose employees used ChatGPT by submitting sensitive information. Never mind, “the LightOn platform works without any data going to an external server, argues Laurent Daudet, co-director general of this rapidly expanding box of 25 people. We install our models on company servers”. And for tricolor establishments, “the fact that our company is French is an asset”, he adds.
Will generative AI also spread its tentacles on the financial markets? That financial data specialist Bloomberg is training its own generative artificial intelligence language model leaves one wondering about its potential reach. “AI has been used in trading for a very long time. It will now be possible to cross-reference stock market and textual information, this will increase its strength”, says Laurent Daudet. But it is not without risk, points out Dominique Ceolin, boss of ABC Arbitrage, a connoisseur of the financial markets. “For the moment, ChatGPT is not very efficient in real time. In the long term, such tools could nevertheless accelerate reactions on the stock market. If everyone has them, all operators will receive the same response and will make the same decisions, which will increase volatility, he fears. Unless several systems coexist, and provide different answers. ” “Accidents” are not excluded if autonomous decisions are taken, without a human intervening party having time to check their consistency. Time is money. To go too fast is to risk losing a lot.