Product Development: Where is AI making the greatest Impact?

I have to share a secret with you regarding this panel. This panel was supposed to look a bit different, as you can tell from the agenda. Steve told me about two three hours ago that I was offered a promotion to become moderator instead of panelists, and I was thinking about it for a minute and then I thought

its a great idea because it will allow me to ask many questions and I will not be forced to give smart answers to complicated questions and this is exactly what the job of a professor is about so sorry for you guys youll have to give the smart answers Ill just ask four questions I would like to start with all of you introducing yourself.

telling our audience a bit about what you do what your company does and maybe you add one sentence explaining what interests you in AI and insurance unless they do you want to start sure I can start so my name is Alicia Braga my background Im an engineer I also studied AI one time ago when I was quite boring and frankly kind of useful useless the same models that were developed 20 years ago or existing now we just now we have more computing power so right now at Kwon Do Kwon Do is a.

restaurant reservation platform if you’re trying to figure out whats the connection to insurance there’s none but Im in charge of building products and we are on the path of building smart products so Im also in charge of business intelligence and data science at kwon do and my previous companies that I worked for our very customer centric I was an engineering leader for.

a company called Sugar CRM and I spent some time at Google at Google on division helping other companies to build their journey to AI machine learning so yeah that’s me hello Im Sten Im founder and CEO of dziga we are london-based company I started my very first business when I was 17 back in high school making paper notepads and then on numerous other things since then but before ziggo I was one of the early employees at delivery and theyre essentially the problem occurred for for what now todays ego we offer essentially a flexible per-minute.

insurance for delivery drivers for new mobility and uber drivers and so forth so there was a big problem onboarding drivers for delivery and others and we just developed a full-stack solution in three and a half months got regulated launched almost three years ago and today we ensure a third of the food delivery market in the UK we operate in five countries now we just two days ago launched an insurance product per minute.

for electric scooters you see all around and we focus on the kind of commercial insurance and mobility well to really enable the movements or beat items or people and do it flexibly because this insurance is very kind of old-school in a sense that its just only annual policies whereas we do it per minute and yeah we and I think well that was last month we.

announced another large funding around 42 million for our Series B as well so that’s that’s us yes my name is flora millar Im a partner with Bain & Company were a strategy consultancy Im heading our advanced analytics practice enemy and my focus is insurance and that’s that’s why Im here today were working with a with incumbents in the industry and were trying to help them to tackle.

all of the challenges that come out of new technologies and really drive value and results that come out of data and analytics so whether its around building the strategy building up an operating model building out the capabilities and driving distinctive use cases and what what loves me and what excites me is how do you really get business results out of AI or machine.

learning because its a lot of buzz and a lot of hype around it and everybodys trying it but really turning it into business results this is what at least incumbents are really still struggling with and this is what excites me most and how to turn it into business results hello everyone so Im a rod Oprah norm Im a machine.

intelligence consultant at uipath so I you have a technical background in data science and after graduating actually Im a fresh graduate I joined the iPad so I you iPad is actually a unicorn company founded in Bucharest Romania after we just secured the series D funding this year we have more than a billion dollars investment in two and a half years we are the fastest growing software company in the world and we do our PA so that’s robotic process automation just to be clear we dont do any physical robots our P is just a.

piece of software that sits on top of our existing systems and its great automating repetitive tasks so we believe that we humans shouldnt be worried with you know we should be worried more with our creative stuff and anything that’s boring and repetitive that can be automated by the robots and what interests me most in in AI is what was said earlier how we can actually bring AI in enterprise and in in.

companies because everyone recognizes AI as being the next big thing but actually having AI at the production level is a is actually not so and not such a great adoption so we are supposed to discuss the question product development where is AI making the greatest impact and we were discussing it shortly over lunch and we thought that we dont want to let us be limited to the product aspect because we pretty pretty soon came to discuss process issues efficiency issues customer experience issues sales and distribution issues so I would be interested what do you think is the area.

where AI might make the biggest difference from a customer perspective or from an insurance company perspective you want to start Cardew yeah so I think the most important is how we tackle document understanding so that’s something that agnostic to any industries any organizations have documents that they want to process and actually understanding and extracting data from those documents is a big challenge so weve tackled maybe structured documents semi structured a little bit but we still have a long way to actually data extraction tackling data extraction from.

unstructured documents so I think this is an area where I can make a great impact and with the use of our PA to actually have a full circle automation in this respect the way we ingest information and populate different legacy systems or you know how you we use that data that’s already extracted with AI that can be automated with the use of we use of our PA yeah I totally share your view that on the entire process automation piece there’s a big impact of AI but if I take a look more.

broadly on the value chain then obviously claims is is a very big lever just taking a look at the size of expenses that are associated with it and just a little impact on the on the other quality of fraud detection claims handling has a huge impact on the P&L of any insurer obviously underwriting and pricing is a big area a topic where many insurers are still shying away out of regulatory reasons actuaries not Im not being too fond of being at least.

partially replaced by machine learning I think where I see big opportunities is around additional services and further personalization and customer experience because as you’re listening to all of the new entrants into the insurance industry there’s obviously always the risk of disintermediation and insurers are struggling and staying in touch with their customers and really providing a personalized experience and how can you leverage the data that you have to provide additional services to your customers whether its around mobility ecosystems providing based on the data.

and information that you all have more personalized advice on how to stay healthy I think its a big theme which also allows insurers to get more in touch with their customers which is one of the big struggles that incumbents currently have because if you only have a claim once a year or and if you have a life insurance policy you might never interact with your customer I think this is where big impact lies you want to add something no I was just gonna add from.

outside which i think is very interesting is pricing there’s not much insurance companies necessarily on this they understand it but they dont want to necessarily take the leap and try pricing based on machines they want to be able to control that but this is where the huge advantage for this technology is I feel ok we have heard three answers from more or less insurance experts here you are from a completely different industry what do you think where I could have the biggest impact across industries or if you have an opinion on the insurance industry I.

think in general its for me as a user and also for someone who builds products its important to tackle the decision making process and decisions are built and built on maybe some sort of biases from humans and when we talk about AI and how AI can help in decision making process speed and insurance regarding handling claims or for end-users to solve what kind of product am i picking is it the right product for me personalized content there’s this one component of it how can we help insurers and end users to make decisions that are.

data-driven and they can also trust those decisions so the second part of it how can we build AI that everyone can trust the aspect Im Im spending most time thinking about when it comes to insurance and digitization III is sales and distribution er there in the German industry probably across well wide industry there’s two schools are thought the one school is when you well insurance is not bought its sold we need insurance salespeople we need agents and brokers they will never go away and you can just forget to try to sell insurance online unless its for.

products where the customer has to buy the insurance like motor liability so when schools telling you nothing will change the other school is telling you it is completely impossible that the next generation will still buy from old-fashioned agent or brokers their insurance forget it all just build digital sales channels the problem I have with both arguments is that they are telling you whats not happening they are telling you its not gonna happen that its sold online and they are telling you its not gonna happen that the agents stay the argument why the other solution is superior is a lot.

weaker so I ask myself how will insurance be sold in the future and what should insurance companies do should they invest into empowering their existing agents and brokers with technology who have artificial intelligence or should they use artificial intelligence new technology to replace agents and brokers and we have some experts here Florian do we want to start with them yeah and and the.

truth sits exactly in the middle and we see this also from the customer studies that were doing the the futures hybrid customers want to do both and they want to be able to choose they want to choose whatever channel and they want to have a similar experience and they want to have a similar pricing and in similar products that theyre getting and this is obviously where where the challenge starts because we taking a look at the.

traditional incumbents we come from a legacy world with established sales forces we have Commission levels which vary across the distribution channels we have totally separated systems just creating in customer centric 360-degree view is a big challenge but you will need both in the future so you need to enable your sales force you need to tap into the online and mobile channel it all needs to be seamless whether similar experience and this is what it makes it so difficult but it will not be.

either-or it will be balls in the future what stain you are selling not through agents or brokers right well weve got 40,000 customers moto customers whove all bought directly from us and and theyre very incredibly happy like previously they were buying it through aggregator websites or somewhere else but they’ve come and bought directly from us because weve very simply explained what the product is and made the journey incredibly simple so and so.

that’s a clear kind of value add where you necessarily dont need a broker at all on the other side when we talk about the scooter companies a b2b side there it really depends it depends on the customers understanding of insurance so we some weve worked directly with them and we are easily able to explain what theyre buying but if they are not fully comfortable with the insurance kind of a language themselves they sometimes want to involve a broker in there and for us that’s fine but fundamentally I think.

that you dont really need a broker for a simple motor product as the complexity increases with insurance product that’s where specific skills our experiences are needed where it becomes more relevant and that’s where brokers are adding value and I think fundamentally brokers the value add has been very low or just now nowadays with technology is diminishing and brokers have to find other ways how they are violet because whats the point of being in this supply.

chain if the value add is diminishing or there is non-existent so I like it needs to be that kind of a whats the easiest resistance way if customer can buy it and simple to understand they should buy it direct if they dont then if they get sync there’s so much more value at working with a broker then brokers relevant but I think that’s shift of a.

roller broke is going to drastically change so there will be changed for the insurance industry you could think that the insurance industry is in the best position to deal with a new digital revolution because now its all about data and insurance for hundreds of years has been proud that it has a business model which is based on data so insurance could think and Im afraid some do think that they’ve they are the masters of handling data while we are seeing business models like.

Amazon Netflix you name them Spotify around us where there’s a new dimension of handling data so I asked myself what could insurance companies learn from other industry when it comes to deal with data how do I make the most out of my data how do I make sure that I can can really tap the full potential of artificial intelligence Alissa can you help us with that I think the common.

mistakes that sort of bigger players make be it insurance or banking they do think like that I have data I have everything so now I can do AI and generally the first step of getting to building smart products in understanding the data you have and start doing data cleansing understanding how can you use this data and how what kind of use case you have to make product smarter and if.

you have an answer that is tied to your business model and you actually have a machine-learning use case then great then you can look at your data next and its great just to have the data but understanding what you have its the key to solve the problem okay I do your company is providing services for insurance companies but also for other industries yes what can the insurance industry learn from other industries or from new services built around data yeah so we.

have I think a hundred plus customers in insurance but RP in general is industry agnostic so any kind of repetitive task you may have that can be automated with the use of RP a so I think other industries and that have started this digital transformation and they understand that you know you have these repetitive tasks that can be automated the low-hanging fruits but these maybe have matter for like four or five.

percent of your processes that can be automated and with the use of I and our PA you can automate and digitize processes in the range of maybe 40% within your company so I think this understanding that you know you have to think automation first you have to somehow break the way you do things in order to make them them better is key in in you know human investment in in general and I think one point in exactly as you said on paper it looks as if all.

of the insurers are blessed with all of the amazing talent and its true all of the actuaries that are highly highly skilled and highly sophisticated but if youve worked in an industry for decades that is a long term industry and where you have is your defect mentality because you know if you do if you setup a wrong product then you’re fully tied.

in making this shift towards a test and learn approach that you need if you start coding and building up your machine learning models I think this is the biggest the biggest culture and mindset shift which is required and which is probably easier further for the startups to deal with that versus when you’re coming out of a culture of zero defect one objection I hear every once in a while is that the insurance.

industry is working with some sets of the most most sensitive personal data and that insurance companies operating in Europe have to deal with very strict EU regulations which make it more difficult to build date build business models on big data which which set restrictions which other markets like China or the u s might not have how could that impact the emergence of data-driven business models would you like Stan sure I think well if you take.

GDP are and all the other rules that have been put in place around data its important and if you think from your individual point of view you want to make sure your data is protected but from a when you build a company well I think what its important when you build a company you think about you cant build a skyscraper I build 30 floors and then realize oh I forgot to put in water.

pipes you have to think about it as you build it from the ground up and I think on our case for example a data is such an integral part we work with listed companies like just II do VII and others they send us a hundred page document how we need to kind of a comply with data and so forth so we have to make it part.

of our DNA how we think about data how we protect data and how we structure that within the company and of course if you’re a large company youve been operating for twenty thirty years with the infrastructure suddenly rules come in place and then you need to start putting your water pipes outside the windows and and its not going to look pretty and its not going to be scalable for sure and and that’s why I think startups now they’ve got this advantage to build it this way by just making sure.

its kind of a inherent in them like even customer data of customer wants the data we now press one button in like two seconds all of their data gets presented and we can share with the customer I am sure all of you have had some experience with gdpr and its a very very cumbersome process and Ive heard some horror stories what has been put in place in order to kind of it adhere to those rules so yeah yeah I think I think its a structural disadvantage that we have so its not holding us back on the.

individual use cases there are always ways around and how you build consent and how you work with animal anonymization of the data but if you take a look at in a broader scaling you were bringing up China and if I take a look at ping on with 1200 data scientists and building up data assets up on 450 million users on on good doctor for example and this is just a tremendous data asset that theyre.

building up which they can leverage to build on additional services and where we all need to see and how can we influence and help shape an environment which allows us to to stay competitive in advance on AI in data wellness is has invited us to this conference in the capital of Germany and Germans are famous for many things one thing Germans are famous for is the German angst Germans are afraid of many things and.

they are skeptical when it comes to new technologies so I try to imagine how a German customer might react to an insurance company which uses artificial intelligence for decision-making at least in some processes and whether the customer would like that and then I asked myself how transparent do we have to be about it do we have to tell the customer that the service agent on the.

phone is an artificial intelligence machine or do we have to tell the customer that certain aspects of a decision-making in claims processing fraud detection are supported by artificial intelligence how transparent does the insurance industry need to bow to be about applying this technology especially an environment where German angst is a phenomenon you’re not a German national but you have lived in Germany for quite a while what do you think as yeah I lived in Germany for ten.

years so that probably rubbed off a little bit the German Ernst I think personally that we should be transparent and tell users whats being used and there are multiple examples of this not working well so one of the examples we had for example at quando we released a chat bot which everyone wants to build a chat bot that’s kind of kind of a fashion and the best practice is actually for building chat BOTS is to.

announce that hey Im a chat bot I can do one two three because otherwise people are just gonna try to break it and ask fancy questions try trying to understand it its is it is it a human and so I think in order to make it successful and improve the technology you should tell that yes were using artificial intelligence and were constantly evolving and improving this I do understand also another side though in insurance can be very sensitive topic because when you did when you build neural network you cannot easily say how.

the decision was make if its overly complex so the questions from the customers will come how did you make this decision how did machine make this decision based on what what can I change to change this decision and you should be able to answer those questions so when I talk to insurance companies as a consultant or with with startups every once in a while about implementing artificial intelligence solutions I very often get the answer okay we already have a solution for that we have an algorithm taking care of it we have artificial intelligence in.

in-house and then we continue discussing it and it turns out that yes they have some kind of computer system there but mostly its its rules-based its not artificial intelligence and then I asked myself what is the best project to start with these guys should we address the a problems where the biggest potential and usually they they will have already some some solution in place which is is a rules-based engine or anything they have invested over the last five five six seven years and theyll tell you well come back later but we have just.

invested in this area we dont want to try anything new or do I start with the B problems where typically the potential is smaller but then there is no solution whatsoever in place and that means it is much easier to prove the benefit of artificial intelligence and flow and you’re working as a consultant with these companies day to day give me some advice how can we how can we crack this industry open that they start implementing these things yes so my advice is always pick something where.

you can be successful where you can create a positive light house and I was would always pick an area where you can generate positive business impact because this is how you get the biggest excitement and yes you’re totally right there are a lot of areas where you have roots based systems in place but you often wonder how fast machine learning is able to out beat the the rules-based system so I always pick an area where you have substantial business impact to.

create a positive light house but also where you have departments and people that are passionate about it and that will implement it because this is the biggest hurdle that we see is create a model which creates outstanding results is easy but implementing and embedding it into the business process this is where it gets really hard so select a use case with high impact in an area where you know that you will easily get it implemented in the business process to get success really fast because no topic is as easily burned as AI if you do it wrongly in the first.

steps because then everybody will say told you its not working and then its very hard to convince the rest of the organization let me let me ask a bit more exists from my understanding artificial intelligence is also about being a bit patient it needs some time to train the modules and if you start with with problems where you can have immediate success a positive business case early on these are prowling that.

not the biggest ones right I I really see that its difficult to convince and show us to say ok you’re starting with an entirely new technology you have to experiment it with it you have to experiment with relevant questions and not just with some toy projects but then they say ok but if that means that that we lose 10% efficiency for the first 6 12 18 months were not gonna do it yes so it takes time but what takes time is the data engineering and that’s.

typically the underestimated part and I think we heard it earlier before because this takes up 70% of your time typically and but you will you often wonder how big the impact really is so take we just did a project on claims management in health insurance which is an established area we have all of those rules based systems around for multiple decades but.

the system was able to out beat the rules early on even after the first training because if you have the proper training set in the proper data in place the starting point is quite similar to what the rules-based system is already able to do but then well fast increase as itself learning over time yes it will take more resources in the beginning because you will you will surface more cases to take a look at and that will it drain your personal capacity that you have but but the business impact is.

actually quite high early on so what you guys cannot see in the audience is that we have a problem with a screen here its blinking and red but it might just mean that we are running out of time so I have one final question for my dear panelists and that is forget about your expert status and put yourself in the shoes of an insurance customer so you as a customer thinking about your.

experience with your insurer in which area would you like your insurer to improve in the next two or three years by implementing a I based solution so for me I dont know it was a millennia maybe I wanted completely automated and tailored for my needs as simple as that very similar imagine me leaving now here from katana electric scooter then get in an uber then get on a plane and then on the same on the other way I dont want.

to think about insurance I always want to be insured on every step of my journey and it needs to be seamless I would be very excited about that yeah for me it needs to be more personalized more personalized tailored to my specific needs and I want more than just the plane policy I want more services rounded yeah I was gonna say.

the same I think now in the age of so much data laying around about our customers is it possible with the use of machine learning to have more personalized products tailored to individual needs so that would be the top priority for for me well thank you everybody we are minus 13 seconds on the clock so thank you everybody and well give way on the stage yes very much give it up for a big one applause here.

thank you guys.