Rise of AI Conference 2024 · Berlin · Applied AI Stage

Unleashing GenAI for Industries

Feiyu Xu · Non-Executive Director of Airbus Group and Board Member at ZF

At Rise of AI 2024 in Berlin, Feiyu Xu, Non-Executive Director of Airbus Group and board member at ZF, argues that generative AI marks a shift from AI that classifies and recognizes to AI that creates content, which she says brings it structurally closer to human intelligence and makes adoption easier because companies no longer need large teams of data engineers to label data. She notes that at least seven of the world's ten largest public companies by market cap are AI-first, and situates the moment against the 2023 OpenAI leadership crisis, the Hollywood writers' protests against AI, and the EU AI Act. She lays out a framework for positioning a company in the generative AI ecosystem across infrastructure, models and applications, data ownership and access, and industry vertical, and recommends companies start with business-process automation before moving to more intelligent products. She then details persistent gaps: large language models lack deep engineering, chemical and geometric domain knowledge, struggle with real-time short-term forecasting, and cannot reliably handle escalated customer complaints or troubleshooting, so she recommends a holistic data governance strategy and treating GenAI as one part of a larger system rather than a complete solution.

Key takeaways

  • Generative AI adds content creation to prior AI's classification and recognition strengths, which Feiyu Xu says brings it structurally closer to human intelligence.
  • GenAI lowers the adoption barrier because companies no longer need large teams of data engineers to label training data.
  • Most European companies sit only at the application layer of the GenAI value chain, which creates dependency on foreign infrastructure and model providers.
  • Feiyu Xu recommends companies start GenAI adoption with business-process automation before moving to smarter products and hardware.
  • Current large language models lack strong domain data in engineering, chemistry and geometry, and are not suited to short-term, real-time forecasting.

Detailed notes

The context: hype, power struggles and regulation

Xu opens by noting the surge of investment and capability in generative AI and foundation models since late 2022, helped by the open-source movement, which she says lets both large companies bypass slow procurement processes and startups run early experiments cheaply. She points to the prior year's leadership crisis at OpenAI as a conflict between researchers who wanted to slow commercialization and better understand the technology's "black box" and executives pursuing large-language-model revenue, alongside Hollywood writers protesting AI over job displacement and the European Union's newly released, risk-based AI Act.

Why GenAI is a bigger shift than earlier AI

Xu argues generative AI differs from prior AI, which was mainly good at classification and recognition, by adding the ability to create content such as images, text, signals and answers, which she says brings machines structurally closer to how humans recognize, learn and create. She gives three reasons industries see particular potential in GenAI: first, this classification-plus-creation combination mirrors human capability; second, adoption is easier because companies no longer need large teams of data engineers to label training data, since models can learn more directly from raw data; and third, foundation models offer one broad model serving many capabilities, compared with the narrow, single-purpose AI interfaces still typical even at large companies such as Google's separate search, image search and translation products. She also argues regulation is necessary because even model developers cannot fully predict the emerging capabilities of large models, so laws alone are not enough and processes and technical monitoring are also needed, including across national borders.

Positioning a company in the GenAI ecosystem

Xu offers a framework for companies and investors to position themselves: first, the value chain layer, from computing infrastructure and hardware, to large-language-model providers, to applications built on top, noting that most European companies operate only at the application layer, which she flags as a dependency risk; second, the data dimension, covering data ownership versus data access and the legal conditions that govern whether a company may use text, image, video, audio or code data it holds; and third, the industry vertical, such as finance, automotive or aerospace. She recommends companies use these dimensions to define a clear position, for example building multimodal foundation models specifically for the automotive industry.

Where to apply GenAI: processes, products and promotion

Xu recommends companies start applying AI to business processes such as customer relationship management with chatbots, supply chain and logistics network optimization, and human resources, before moving on to making products themselves more intelligent, for example autonomous driving or in-car digital assistants, or infusing intelligence into hardware so that device manufacturers become combined hardware-software companies, or building entirely new AI-dependent products such as drones. She urges companies to actively promote this progress internally and externally rather than treating it modestly, noting that marketing AI-driven improvements is a continuous effort many European companies underinvest in.

The gap between demos and scaled production

Xu warns that running convincing generative AI proofs of concept is very different from delivering scaled, high-quality processes and products, and that hiring AI experts before establishing data access and governance risks losing those hires quickly. She stresses working with legal, IT and business units to determine which data can be used, under which conditions, and for which purposes, since obligations differ for continuous improvement of existing products versus creating new ones, and for structured versus unstructured data. She identifies specific technical gaps: current large language models have strong data coverage in software and medicine but little in engineering, chemistry or geometry, which limits applications such as automotive simulation; trustworthiness and hallucination depend heavily on how complete a model's domain data is, not only on training and instructions; and models are not close enough to real time to support short-term forecasting such as supply chain demand prediction. She also cautions that generative AI should not be used to handle escalated, angry customers or full troubleshooting, saying it can competently cover roughly 80 percent of fact-driven customer requests but the remainder still needs human experts.

Recommendations for scaling GenAI

Xu argues an unscaled proof of concept counts as a failure, not a success, and that scaling requires a holistic corporate strategy: clear data strategy and governance, defined ownership across ideation, deployment and continuous improvement, and a pricing and marketing approach that starts by confirming user satisfaction before monetizing further. She recommends leveraging the broader AI ecosystem rather than building every technology in-house, and advises companies choosing a foundation-model provider to check what data it was trained on, its governance and testing practices, and its energy use and sustainability. She closes by saying vertical, industry-specific AI and AI embedded into hardware are strongly needed, alongside continued vigilance about deepfakes and misinformation.

Questions answered in this talk

Why is generative AI a bigger shift for industry than earlier AI?

According to Feiyu Xu at Rise of AI 2024, generative AI adds the ability to create content on top of the classification and recognition that earlier AI was good at, which she argues brings machines structurally closer to human intelligence and makes adoption easier since companies no longer need large data-engineering teams to label training data.

Where should a company start when adopting generative AI?

According to Feiyu Xu at Rise of AI 2024, companies should start by applying GenAI to business processes such as customer relationship management, supply chain optimization and HR, then move on to making their products and hardware more intelligent, and should scale from proofs of concept only after confirming users are satisfied.

Can generative AI handle customer service escalations and troubleshooting?

According to Feiyu Xu at Rise of AI 2024, generative AI can competently cover an estimated 80 percent of fact-driven customer requests, but escalated, angry customers and full troubleshooting should still be handled by human experts.

Mentioned

Generative AI, Foundation models, EU AI Act, Retrieval-augmented generation, Data governance, Vertical AI, Deepfakes

Keep exploring

More from Feiyu Xu

More from Rise of AI Conference 2024

More on Enterprise AI & Adoption

Transcript

Auto-generated transcript, not edited. · 3,542 words

Show the full transcript

good afternoon I hope you have collect your energy for my talk I want to share with you my experience and also Vision how we can leverage artificial intelligence and g&i for the intelligent transformation of your company and the industry perhaps you as startups contribute to the intelligent transformation of Industry since November 22 a lot of things happened with AI with our world with our society a lot of investment went into G Ai Ai and also the recent large language models Foundation model they are becoming more and more intelligent more and more impressive I think most of you have tested GPD 4 already right perhaps you use slides use GPD 4 for your slides already and uh in particular I'm so happy to observe the open source movement with large language model this really help large companies to access the most recent Innovative large language models without the long procurement approval process and and also for startups companies right you can use open source large langage model do your first Innovative experiments however we know artificial

intelligence is also a big change has a big impact of our society last year we observed the struggle of open AI this is not only the power struggle this is a struggle of people who think scientist scientists they think we have to slow down the commercialization of open AI we have to understand better the black box and uh Sam otman and many of us many industry they have they want to make big money with large langage model foundational models and also we know the writers in Hollywood they protected against large language model they are scared of AI where replace their jobs and we know this year European Union released the Air Act the air regulation which is risk driven you see it's so complex around Ai and uh I think the most impressive part is in addition to previous AI for automation classification now ai go to the knowledge cre cre ation part right create pictures images signals texts answers and the instructions this really make a big difference if we want to persuade Industries

to be part of the game of AI you just have to look at the 10 largest public treaded companies by Mark cap in the world at least seven of them are AI first companies including hyperscalers platform companies such as meta Microsoft and also infrastructure companies like NVIDIA right and also car companies but Tesla now is facing challenges from China however if you want to be successful and skill I think without Ai No company can really be successful when I talk with Industries in particular board members and the senior Executives they don't know what is AI they think perhaps even calculator is AI right and uh but AI actually uh serve very special capabilities of human societies for example AI enable machines to behave like humans like like robots and uh in many science fiction film we fall in love in robots robots fall in love in us I think uh two years ago most people were not fall in love in robots in machine right in AI machines but today after the large Range model

after the foundation model I think the challenge is there perhaps some of us were falling off in AI enabled assistant because you can really talk and chat with AI for a long time right and in most use cases most applications the industries and the way leverage AI Technologies be not because they simulate our capabilities they are much powerful than us right for example AI can recognize hundred thousands hundred thousands of faces for me it's always so much such a challenge I feel often very bad I saw a face but I cannot associate his name or her names uh in particular at conference I think the AI assistant will do much better job even help us and also now ai Empower individuals all of us we are bet we become poet we can write poem now and uh now I talk also with the and entertainment industry now ai also enable us to write songs to create films and uh create pictures AI Empower individuals and our employees to have such much more capabilities beyond our

imagination right we can become really poet if we have good ideas if we create good prompts and write songs and compose some musician pieces right which we can never could not did uh do before for many companies I think AI really can help the human machine interactions both for their devices and also for their communication between the customers and their products and services why industry now sees such a big potential in Genai there are technology reasons first the trade additional machine learning Technologies they are very good in classification recognition the new technology they are also very good in creation right the classification recognition plus creation this is what we humans do right we recognize things learn things and we can create content this means AI is much more close to human intelligence than before the second very big uh reason is this makes make also the adoption of AI in the industry much easier why because most companies and most smes they don't have data Engineers before we need high quality labored data we need

data Engineers we need data data Engineers work with the domain experts like very easy right for the Fe recognition you have a picture you have to say this is the eye this is the nose this is the month this is year today if we have high quality raw data the machine can recognize itself this is a really lower the barrier for the adoption for industry this is actually one of the major reasons why commercialization and the industrialization adoption get easier you don't need a big team of data engineer to label the data annotate data machine can then themselves and the third one what I what we think now ai is much more close to human intelligence we humans we can talk we can recognize faces we can give answers if we speak several language we can do translation before if you look at Google right some uh two months ago I complained the product owner of bath and see your user interface your still narrow AI if we go to Google we have a search

we have image search we have machine translation actually the large langry model the foundational approach means you have one foundational model one lar model you can serve different capabilities now we have broad AI before narrow AI but this is all still a challenge for many companies even AI first companies what the user interface for a broad AI approach still we have very narrow AI interface the last one I think why we need AI regulation why we do have to monitor what happened with AI because even the owners the developers of large language model of foundation model they cannot really predict the capabilities which the large language model and the foundation model really can bring us the emerging capabilities right the emerging capabilities is on the one hand so exciting because they can really contribute to the Innovation and also to contribute to the science on the other hand we have to monitor it carefully but only with regulation with laws are not enough we need processes we need even technology at The Meta level in

order to monitor this emerging capabilities and even also cross the boundary of the nation Nations right this is very very important and now I want to share with you I think I think many of you know them perhaps are boring for you already and uh for Industries the applications of large engine model if we uh classify them based on the data structure from text point of view right many of you work with lar model already writing emails Nob brainer formulating perhaps answers of interviews and also translation I guess most of you use it right right who do not use the text functions of large langage model most of you and also I think for many companies very important they leverage codex in order to help their software development but do you know what is the most useful uh use case the most useful use case is most companies if they want to deliver softwares they want and they need also to understand the demand of the users right the understanding of the customers and the

find the existing software adapt to the new demand this is in the in the future very important not only creating the software codes and uh also for videos and films I talk to many investors and on many Industries many indust in particular big Industries they want to Define their air strategy but since they see hey it's so complicated there's so many Air Technologies and uh how can we Define our air strategy and the investor say oh we have money but we don't understand the ecosystem where should I put our money into it it's very simple I want to share with you the classification of the Gen ecosystem first from the value chain point of view actually we have the on the bottom infrastructure for computing and cloud and then even Hardware companies right and then we have a companies who provide large langage models the AI Technologies most companies in the Europe even most of us right we build applications on top of infrastructure and uh Air Technologies and The Leverage applications however the challenge

for Europe all of you know it if we are only stay with the applications we are too dependent on the infrastructure and the core Technologies of our other regions and countries and then after you know your position in the value chain you have to see which kind of data the we as companies own or we have the access to perhaps owning the data is even less important than the data access most companies have to spend a lot of energy invest of the data access the data access to your customers the data access to your own data right even the data is in your own company many data you are not allow to use without the legal conditions very very important and uh texture data image data video data audio data uh code data right and uh the third dimension is Industries which industry do you belong down to game industry car industry Aerospace given this three dimensions actually every company can position them the notar what is the notar for my company is my company

I want to be the number one in finance building the automation optimization intelligent Finance industry applications all I belong to the company I want to build multimodal foundational models for car industry and also for many companies from they don't know how to what AI really contribute to most companies right very simple Mo AI contribute to most company most companies start with the business processes they want to make the customer relationship management more easy with chatbot sted right supply chain optimization logistic Network optimization and the human resource processes I also strongly recommend most companies who want to start with AI should start with the processes and then it will be great if companies can leverage AI to make their product more intelligent even car industry right you have autonomous driving you have digital assistant in the car and for many device company Manufacturing Company AI can really Infuse make your Hardware more intelligence right many company become a hardware software uh combination company and the many companies create even new products based off the AI right

drone without AI computer vision you cannot devel any drones right and and I recommend every company don't be too modest if you start with your process optimization with AI promote it make visible make your company visible any time see my company is a intelligent company I make the process more efficient and my product more intelligent and the promotion and the marketing with AI for every company is a continuous Improvement this is also many European company they are missing this effort also this is not part of the European company culture now since um because of uh GNA larg Lang model this make the access to AI Technologies much easier it's m low touch no touch right however be able to use large langage model and uh do some prompts the does not mean you can really improve your process in a high quality way efficient way and does not mean you can deliver high quality processes and high quality services and products the gap between leveraging G AI have some poc's and make the executive board

happy for one moment is completely different from skill your processes and skill your product because if you want to really Ru out AI enabled products you need to also come up with many other things right the quality control perhaps in the beginning the human machine collaboration is very important you have to set the threshold of the quality and the perhaps in the beginning you have only 20% automation later 90% and uh every company PS should be happy with prompt Engineering in the beginning and and this can be a very good promotion for GNA in the company however even you invested money into a a lab hide the people these people will leave you very fast if you don't have the data therefore before you spend money hide the uh the uh the a experts it's the most comfortable part even it's hard to find the right experts however you have to work with the legal experts and also it and the business unit identify where are the data in your company and uh which data

are needed and uh under which condition you are allow to use the data there are very very complex legal conditions but you can work out for example which data are you are allow to use continuously Improvement of your products which data you are allowed to create new products they are completely different different legal framework and how can you deal with structured data and under which condition you can deal with unstructured data for unstructured data in which process you are not allow to look at the data this kind of a competence you have to build up because every company we have the responsibility to minimize the risk minimize the risk for the company to developing use cases without uh um losing the control of the data governance on on the other hand also minimize risk for your employees for your people because this kind of risk must be taken over by the uh company not the employees not the um developers and the researchers I know recck is very popular now right uh you com combine

with embedding search and the prompt engineering however the biggest challenge for most Industries in order to really develop uh high value products and services first if we look at the large langage model the current the common CR right it have a lot of uh uh data about software great therefore can help us and a lot of data of in medical area therefore even I ask for Chinese medicine uh gbd4 can give me some advice for winter in B however there's no know about engineering and chemical industry and also if we for example want to build applications for car industry right simulation you need also knowledge about the geometric there's no knowledge behind it how can we Infuse this knowledge into it we have to find out lot of research is ahead furthermore trustworthiness right TR trust wor is not not only only dependent on instruction behavior training also the completeness of the domain the Hallucination is not only the hallucination because the Jai really want to please you give you just an answer because the

missing of the data therefore high quality data as complete as possible about one area is so important I think one company who really can be the leader in particular industry is really depend whether they can have the right data quality data and the the last one is also right if you want to do supply chain demand forecast short time um short-term prediction realtime data really important and uh we know today we are happy with Spring train uh um data the data of course we know that in larg langage model the data is more closer to today still there's Gap right if you want to do short-term prediction in the last two days the current large Lang model will not help you it's very important to work out how can we Infuse real time data in order to make the predic successful I want to share I want don't want to scare with you right because we are not yet so far with the industry foundational model if you look at the data in a company

in a car company we have a business process data they are like trees or graphs they are not pure natural language and and if we most the data are transaction data structure data in tables right and then we have also data in graph if we give report right and also chemical data geometric data all these data are needed in order to make industry application successful even I'm fan of large L model generative AI Foundation model we still have to be very very clear this is only Technologies and in order to make our company successful our application successful we have to treat large en model as only part of the solutions the overall solution is complexer therefore uh knowledge cph and also planning collaboration between machine and the human are very important there we we have don't have to treat large energ model as a magic to can solve all the problems and uh many applications for business you have million with right um like Rec for Knowledge Management many companies have started and I think

uh gen AI for simulation of product design is still not yet so far because we uh the larg the MTI model larg model for industry are still under development for check part also for uh digital system for customer service uh for this part I also really strongly recommend you be very careful don't consider J AI can really solve all the problem of customer support don't let AI to solve the escalation if your customer is angry with your product your service let the human treat it otherwise you lose it right and also I don't think U the current as uh generative A system can really handle trouble shooting because Tribble shooting means also the as system should at first understand what the human has have done and also do the inferencing but for the fact driven question and also instruction how to question AI can really deal with they cover perhaps 80% of the customer request already you save a lot of money already but another 10% please leverage humans human experts I don't have so

much time but I do very quickly uh I think for many companies uh uh they start with a application use cases and uh I always say a POC which we not skill is a failure it's not a success but in order to make AI application successful and skill you need a holistic a strategy for Corporation data strategy data governance very very important a clear process and governance who owns what who is who owns uh the ideation uh for deployment and the continuous Improvement and the pricing also marketing right many companies are very eager say oh I have ai I want to uh get more money be cool at first get the feedback of your users right and whether the users are happy with the intelligent solution and the and then if the customers are happy with the result perhaps you can have this log in and ask for more money therefore I think it's and also the ecosystem not every company can develop every Air Technologies right leverage as much as possible the air uh

ecosystem I will skip the a regulation I guess in this conference many people uh have repeatedly talked about it and if you choose a large langage model foundational model company please be careful which data they use which kind of air governance which kind of testing uh evaluation method we have how they leverage image uh energy whether they are sustainable very important check these companies before you use it blandly and I think we have a bright future in front us still we have to be careful vertical AI I think is heavily needed in the industry and AI into the hardware and also this information de fake everybody of us we have to be careful thank you very much

Rise of AI 2027 · Berlin · May 2027

Be in the room for the next one.

Ten years of talks started as conversations in one curated room in Berlin. The next edition brings 300 selected leaders together in May 2027.