When AI Agents buy insurance, what happens to customer trust?

[2.26–16.40] Hi, so your conversations with Kane obviously. Kane’s going to introduce them in that camera.
[16.98–21.60] So we’ll start off by looking now and then when the conversation starts just talk to each other.
[21.96–27.24] If I move, ignore me, try not to follow me with your eyes. I might check that camera or do something.
[35.94–45.28] Hello, welcome back to VUX on the road here at CCW Europe in Amsterdam. I’m with Bogdan Tancic
[45.28–50.46] from Lemonade, Director of Customer Experience. Keen to get your thoughts on a bunch of things
[50.46–55.22] because I’ve been following Lemonade from afar for a long time and we kind of mentioned earlier
[55.22–62.64] on I’m sure that Lemonade was the brand that originally was pretty much only available to
[62.64–66.82] interact with via SMS or something like that. I don’t know how true that is. I’ll have to
[66.82–71.78] dig out the thing and find it out but essentially long history as a company of leveraging
[71.78–78.64] AI would you say? I would completely agree with that and if I may take you a step back there
[78.64–84.08] and ask you a question back. What’s your feeling about traditional insurance companies if you were
[84.80–89.86] comparing Lemonade with the others? Lemonade when you say is fairly new on the market
[89.86–94.70] comparing to the traditional insurance companies and it’s actually built on the premise of that
[94.70–99.50] insurance can be about trust and not the mistrust. There is this perception of people
[99.50–105.58] having insurance treating as a mistrust element with a lot of complexity and etc. Therefore
[105.58–111.02] Lemonade was actually built to disrupt that whole perception and making sure that technology
[111.02–117.14] can amplify that. So Lemonade was built I think more than 10 years ago to actually disrupt
[117.14–123.72] that original traditional insurance space around the technology. So it’s basically the AI
[123.72–130.68] native company which built the insurance around it. Interesting. So another question there
[130.68–137.16] that sort of stems from that is that today there is this kind of like almost like a tension
[137.16–142.54] when it comes to trust of AI because on the one hand you’ve got certain media outlets talking
[142.54–146.54] about it being a doomsday scenario a lot of the researchers from a lot of the frontier labs are
[146.54–151.04] talking about it could take over the world and annihilate humanity. At the same time those
[151.04–155.26] same people are also saying that it could be the best thing for humanity ever because we can
[155.26–160.46] cure all kind of diseases it’ll be really beneficial for us. So like the general public
[160.46–165.24] if I think about my mom for example are kind of in this weird position those that are not in
[165.24–168.78] the industry are in this kind of weird position where either AI is oh I don’t know about that
[168.78–174.56] or it’s kind of like positive. As a company that’s built on technology that’s infused with AI
[175.08–179.76] how do you handle that whole kind of trust question because the trust that you want
[179.76–184.86] to establish is built in those little tiny interactions that you have with customers over
[184.86–188.80] time. So is there anything you can share in terms of how do you go about building trust?
[189.38–193.12] Definitely and I think that’s going to become like even bigger topic in future and we can
[193.12–199.42] discuss more about that like trust is a big big factor of consumers that we have.
[200.36–206.34] Part of that is earning trust in a small points of interactions with customers right.
[206.82–212.34] One is whenever they need you so that you are there so that they can quickly and to the point
[212.34–217.86] resolve their issues that they have and I think you earn trust over time it’s not the
[217.86–222.28] thing that you get or lose basically you can lose it pretty pretty quickly and that’s a big
[222.28–227.18] thing right but like you earn it over time you’ll earn it at every point of interaction that you
[227.18–232.56] have from the moment you quote right eliminate we can talk about a little bit about how we leverage
[232.56–237.50] AI right from the moment you quote basically to get you the best pricing for the best coverage
[237.50–242.62] right that’s all powered by AI right and then the moment you’re on board right and then the
[242.62–247.42] most important part when you claim right when you’re actually there for the customer when
[247.42–251.74] they really need it that’s where you want to show that you can quickly resolve the problem for
[251.74–256.50] when they actually need you right and so those are the moments when you actually gain this trust
[256.50–262.48] and earn it with the customers that you have and AI definitely amplify amplifies that and gives
[262.48–268.04] you ability to be you know the most sincere the most accurate to to to the customers that they
[268.04–273.60] have nice can you share a little bit in terms of where you’re kind of leveraging AI and types
[273.60–277.36] of use cases that you’re that you have live and things like that yeah definitely I was
[277.36–282.02] mentioning about you know starting with the onboarding right we have AI walking you to
[282.02–286.32] the onboarding so that you kind of see like what are you getting into what are you going to be
[286.32–289.96] under written for what are you going to get covered for like what’s going to be the pricing
[289.96–294.66] so that we collect the right type of data right and that we can actually get you the right
[294.66–300.98] quotes and then once you get the policy then you get kind of the whole post-sales journey
[300.98–306.12] also powered by by AI what does it mean like on the background we have also the pricing
[306.12–310.76] which I was just mentioning we have the underwriting covered by AI we have also the the
[310.76–317.08] fraud detection right so we are able to be super accurate about giving you the best price and the
[317.08–323.84] best coverage ever interesting um curious to get your thoughts on this kind of emerging
[324.52–331.98] kind of trend if you like around personal agents you know grokbot dots muse these things
[331.98–337.30] have been kind of released fairly recently you know we had claw open claw before it where you
[337.30–343.32] know there is an emerging group of people that have their own personal agents and those agents
[343.32–347.52] are going to be the ones going out and getting renewing their insurance or making the
[347.52–352.06] claims on their behalf and stuff like that um obviously it’s early days for that but the
[352.06–356.34] evidence I’m seeing so far is that businesses are telling us that actually we get interactions
[356.34–361.86] like that how are you thinking about that in terms of as a business this emerging thing is
[362.50–367.76] what’s kind of going through your mind definitely it’s happening it’s starting to kind of evolve
[367.76–371.76] and we need to recognize that more and more people are leveraging their own personal AI
[371.76–379.20] agents I can personally argue that AI agent can buy things but only a person can have a feeling
[379.20–384.98] or a troubled time where they actually need to claim like that that difficult day in their
[384.98–388.42] life like that’s what they are going to express and that’s where you’re going to speaking
[388.42–392.46] about the trust that you said that’s where we really want to be there right so speaking
[392.46–398.08] about the agents I think the landscape is going to shift in what the companies need to cover
[398.08–402.56] in terms of their distribution and how this service right what does it mean like the agent
[402.56–407.68] in itself doesn’t matter about ux that you have right they don’t matter about the website polish
[407.68–414.36] website they have you need to have like a very accurate apis with the machine possible language
[414.36–419.30] and the data that’s exposed to those those agents so that’s kind of a completely I would say a new
[419.30–424.50] landscape how to service those agents so that they can make the most informed decisions would
[424.50–431.76] that be a purchase or explanation etc going back to the trust of AI in general as a concept right
[431.76–437.40] I would argue that even if you were to employ an agent right now today maybe over time it will
[437.40–442.22] change and evolve the trust that you put in your own agent you’ll still like to verify the
[442.22–446.66] very end so maybe right now the agent what they can do they can search they can search like a
[446.66–451.22] number of insurance companies out there and see like what’s the price comparing to the coverage
[451.22–456.60] and etc and give you kind of the final recommendation based on what they have I think
[456.60–461.60] that’s gonna be the distribution game for sure like so that we can be there with the right
[461.60–466.52] pricing with the right with the right coverage but then also very much important what happens
[466.52–471.36] after that right how do we treat those customers how do they get the level of service that
[471.36–475.64] they deserve so in the time they actually need us so that we are there so I think that’s still
[475.64–481.14] gonna be heavy on that element of earning trust while actually they are with us so that I think
[481.14–485.42] those are kind of maybe two distinct journeys that we are yet to see like one is agentic
[485.42–490.30] buying at the very beginning and that’s also very I would say still questionable in a way
[490.30–495.92] what are you allowing agent to do like on whose behalf is it regulated or not so I think
[495.92–500.94] regulators today even are a little bit lagging behind and learning together with us
[500.94–505.20] you know what’s allowed how are we going to regulate the space and etc so I think
[505.20–511.26] we still have a fairly long runway there to go and learn right but I think that’s definitely
[511.26–516.32] a field that we are exploring. It’s interesting isn’t it because you know hearing you speak
[516.32–522.74] there it’s like yeah there’s like layers to it one is kind of like getting a presence in those
[522.74–527.74] agentic platforms if I’m doing a comparison I use Claude to go out there and scour the internet
[527.74–532.72] for all the best insurance deals give me them all on my plate sort of thing you want to be
[532.72–540.92] there at the same time then I might have mews or what have you go to a comparison website or
[540.92–546.78] lemonade’s website directly do some research on that or even actually go ahead and start getting
[546.78–551.92] me a quote or something and all that lot and then the level beyond that is that actually
[551.92–558.48] the agent makes has a live interaction either with one of your agents live agents or with one
[558.48–563.10] of your AI agents sort of thing it seems to me there’s kind of like levels to it and what
[563.10–567.26] you’re saying there in terms of like how do you authenticate how do you know that this is
[567.26–571.92] Cain’s agent or this is Bogdan’s agent like it seems early days but I mean this thing could
[571.92–576.70] we don’t know what chat gpt took off in a month so this stuff could potentially happen
[576.70–580.84] pretty quickly I mean it’s interesting to to figure it out. I agree with that and one can
[580.84–586.56] argue like what matters in the in that space right does empathy matter that do agents matter
[586.56–590.36] about the empathy or how they are treated or they matter about the pricing or they matter
[590.36–595.18] about the coverage like what elements would actually go in a buying decision right now if
[595.18–599.74] that will be an agent right like we know humans actually are buying based on their feelings
[599.74–604.64] and emotions more often than like on a rationale right one can argue like what’s what happens
[604.64–609.68] with the agent right how do they gonna buy like how are you gonna instruct your agent agent
[609.68–615.66] Cain to say hey my agent please buy because based on this set of criterias right so that’s
[615.66–619.44] going to be very interesting to see what are the buying decisions and criterias for those agents
[619.44–625.34] to do and I might not know why I buy as well and then secondly is you know businesses have
[625.34–630.34] spent a lot of time and effort and a hell of a lot of money competing on the experience layer
[630.34–636.90] so where you know insurance companies they all insure people so fundamentally at a fundamental
[636.90–640.50] level there’s a lot of similarities between insurance companies but everything we’ve been
[640.50–644.28] talking about here so far has been trust exactly that’s a differentiator for lemonade
[644.28–649.68] is the trust that we have the interactions we provide what happens to the interaction
[650.48–656.96] if the interaction is happening between my agent and an AI agent that lemonade provide does the
[656.96–661.68] interaction matter because the agents will probably just figure it out anyway you know it’s
[661.68–667.64] interesting yeah I would say like there is always going to be a downstream kind of to a certain level
[667.64–672.00] human element of that right so maybe the transaction will be closed at the beginning
[672.00–675.98] based on the best price the best coverage and etc but it’s going to be very important what
[675.98–681.06] happens also downstream and that’s what I’m trying to to put a focus on right now is what
[681.06–686.50] actually happens after right when you actually need like insurance tech companies
[686.50–690.82] are not traditional software companies right you don’t need to be tied to an app right
[690.82–696.54] you actually it’s a service oriented business right so you actually need to interact let’s say
[696.54–700.94] with a company when you need to download your policy right when you need to know to add someone
[700.94–705.24] else to your policy right when you need to clean something right so AI has been like
[705.24–712.38] significant help in providing for instance that level of safety like peace of mind so that you
[712.38–716.86] know that you are covered and you are actually going to be treated well when you need it let’s
[716.86–722.60] say you have a you have a pet right at your home and you are the veterinary and unfortunately you
[722.60–726.60] need to pay for something and you’re a cover for that that’s not the moment when you want to have
[726.60–733.42] a friction right and that’s when you earn trust with your customers right so again not not saying
[733.42–737.24] that it’s not important like how you compete on the upfront in the in the top of the funnel
[737.24–742.14] right when you say hey how my agent’s gonna treat like I think it’s a fairly big distribution
[742.14–747.76] game over there like how are we gonna be compared to the others basically AI will help us be a little
[747.76–752.40] bit more not a little bit more like way more precise for we cover how we cover right and then
[752.40–758.44] to which price elements or price points right but then very important what happens maybe because
[758.44–762.30] it’s close to my heart and I’m living that world like a post-sales world like optimizing that
[762.30–769.64] journey very important how you retain trust and even kind of continue having your lifetime
[769.64–774.82] customer if you will yeah I think the challenging part of insurance is that if you’re one of the
[774.82–780.30] customers who’s fortunate fortunate enough to not have to claim you probably have a worse
[780.30–786.94] relationship because if I do have to claim you have an opportunity now to really build that trust
[786.94–790.84] if I have a claim and it goes through so swimmingly and there’s no problem and I’m
[790.84–796.08] treat really really well and all that kind of thing the the chances of me saying well regardless
[796.08–800.18] of the price stops being a problem then yeah because the renewal is when I like the company
[800.18–805.52] and I trust the company so maybe it’s when I don’t have a problem that’s where the
[805.52–809.72] issue is because I might be more fickle then I’m just looking for a cheaper deal sort of thing
[809.72–815.58] so it’s you know where in that post-sale relationship can we build our trust it’s an
[815.58–820.92] interesting question true and again like just because our ability to work with the latest
[820.92–826.38] models with our latest language models as well as the machine learning models allows us to be
[826.38–831.02] even more precise for you like when it comes to let’s say for the renewal time right
[831.02–835.24] to be more precise be competitive with the market and give you the best price ever
[835.78–840.28] for the coverage that you need yeah brilliant brilliant well well done appreciate your time
[840.28–843.34] thank you so much thank you really appreciate it nice one thanks very much