Naked puts car insurance quotes inside ChatGPT, but the fine print still matters

The launch also appears to make Naked the first South African company to release a native app inside ChatGPT. But for consumers, the more important issue may be less about the novelty and more about who handles their data, who is responsible if the chatbot gets something wrong, and how far AI assistants should go in financial services.

Naked says the app connects directly to the same live underwriting and rating engine that powers quotes on its website and mobile app. This means the premium shown inside ChatGPT is not an estimate or lead-generation teaser, but a final quote that can be accepted by the customer.

For now, the ChatGPT app covers car insurance, including comprehensive and third-party-only cover. You can answer questions in normal conversational language and change aspects of the quote, such as the excess amount, to see the effect on the monthly premium.

Alex Thomson, co-founder of Naked Insurance, told Daily Maverick the launch was “a world first” in terms of a final car insurance quote inside ChatGPT, although he stressed that the company was being careful about the claim because other providers may offer indicative quotes or simpler financial products through AI interfaces.

“It’s worth saying at the same time that this is an exploratory step,” Thomson said. “It is showing what’s possible, rather than suddenly becoming a recommended way of getting quotes from us. We still think that our website and app are a better journey.”

How it works

You need to connect the Naked app inside ChatGPT before you can use it. On the left side of the screen, under “new chat” and “search chats” click on “…more” and then click on “apps”. Then enter “Naked” in the search bar.

Once connected, you can ask ChatGPT for a car insurance quote from Naked. ChatGPT then asks you for the details needed to produce the quote, such as the vehicle, overnight address and claims history.

You don’t need to answer in a rigid form-style format. Thomson says a user could provide several details upfront or respond to individual questions as they appear. ChatGPT then interprets the answers and passes the required information to Naked’s systems.

The back-end pricing and underwriting process is unchanged. What Naked had to build was the link between ChatGPT and its existing quoting engine, together with the checks around how information is gathered, confirmed and displayed.

When the quote is returned, it appears visually inside the ChatGPT interface, with controls that allow you to adjust aspects of cover. Thomson says, for example, that you could change the excess by clicking a drop-down menu or by typing an instruction to ChatGPT.

The ChatGPT route is not as fast as Naked’s own app or website. Naked says customers can get a final quote through its own channels in under 90 seconds. Thomson said the ChatGPT process took under five minutes.

The privacy question

The trickier part of this experiment is not whether ChatGPT can ask questions. It is what happens when those questions involve personal information.

Naked does not collect certain sensitive details, such as ID numbers and email addresses, through ChatGPT. Instead, the customer is directed to a secure Naked-controlled page to provide that information.

That data was stored by Naked and was not passed back to ChatGPT, Thomson said.

“We have to be very careful with the handling of personal information,” he said.

Every time information is sent to Naked, the customer must confirm that they want to send it. Before a quote is produced, Naked plays back the information it has received and asks the customer to confirm that it is correct.

“We actually can’t see what ChatGPT and the user are saying to each other,” Thomson said. “That is your private engagement with ChatGPT. ChatGPT sends us something, and then we play that back to you so that you can confirm that it is what you intended.”

International counterparts

There are at least two international insurers (both in the UK) that are using the same playbook, but neither gives customers quite the same offering as Naked Insurance.

Aviva, UK: Aviva announced a ChatGPT app for initial home insurance quotes. Customers still click through to Aviva’s website to choose extras, review documents and complete the purchase. So this is similar, but less complete than Naked’s claimed in-chat final quote.

MoneySuperMarket, UK: Existing customers can get an “instant full quote” for car insurance, while new customers can get quick estimates after five questions. It also covers broadband and a savings/current-account comparison. This is probably the closest to the Naked Insurance offering, but it is a comparison-service model rather than a single insurer’s live underwriting engine producing its own binding quote.

What if ChatGPT gets it wrong?

The accountability issue is where the launch becomes more complicated.

If ChatGPT misunderstands a customer’s answer, or if you supply vague or incorrect information, Naked’s position is that the responsibility sits with you, the customer, once the information is played back and confirmed.

Thomson gave the example of claims history. If you are asked how many claims you have had in the past 24 months and gives an unclear answer, ChatGPT could send Naked a structured response. Naked would then display that information back to you for confirmation. If you confirm incorrect information, the problem or liability sits with you.

“The responsibility really does sit with the client, because we are making it available,” Thomson said. “It is very important when using AI that people do it responsibly, cautiously.”

This is where the distinction between an AI assistant and a regulated financial adviser becomes important.

Thomson said ChatGPT was not acting as a broker or regulated intermediary. Rather, it was acting as a personal assistant with access to the Naked app. The app told ChatGPT what questions to ask and received the answers needed to produce the quote.

He cautioned that users may still ask ChatGPT broader questions about Naked’s cover, excesses or policy choices, but Naked was not validating those answers because it did not see them.

“If you ask it what excess should I choose, that’s financial advice,” Thomson said. “It should really not be answering that, but it may answer that.”

Insurance as AI admin

The bigger prize for insurers is not a quote inside ChatGPT. It is the possibility that AI assistants eventually become the consumer’s admin layer for financial products.

Naked’s release sketches a future in which a customer could set preferences once, after which an AI assistant could monitor their policy, track renewals, gather competing quotes, explain differences and help them switch if a better option met their needs.

That has obvious appeal for consumers who dislike forms, call centres and policy documents. It also raises a strategic risk for insurers. If AI assistants become the main interface, insurers could be pushed further into the background and compared mainly on price.

Thomson admits this is a concern across the industry, but he argues that AI could also help consumers assess the quality of cover, claims handling and customer reviews, rather than only ranking insurers by premium.

“What AI can also do is actually check that this company is legit, that they handle claims really well, that they are good at servicing,” he said.

For now, Naked is treating the launch as a learning exercise. It wants to see how customers use the tool, where the friction points appear and whether the technology could eventually become a serious distribution channel.

“We are not waiting for AI interfaces to go mainstream,” Thomson said in the company’s announcement. “We are making sure our technology is ready when our customers are.”

The experiment may be early, but the direction of travel is plain enough. Financial services companies are beginning to build for a future in which the customer’s journey may not always start with a website, an app or a broker. Sometimes it may start with a prompt. DM