The challenges in data management are diverse. But so are the opportunities – Verena Fink and Dirk Hauke are certain of that. With their expertise, they want to help managers and executives lose their fear of innovative technologies. In the joint conversation it quickly becomes clear why both have a faible for AI solutions. And why this brief exchange is not enough to cover the entire topic.
Verena Fink has been a member of parsionate’s advisory board since 2019. Dirk Hauke has been a shareholder and senior partner of the company since February. We have summarised some parts of their conversation for you.
VERENA: Artificial intelligence or AI is an inflationary term right now. It feels like everything is being sold as AI. In the AI books I write for managers, I advocate that we demystify AI and actively look at the opportunities that real use cases offer.
Interested in Artificial Intelligence? My new book is here: “KI-Projekte einfach machen“
You know real-world use cases well enough, what is your AI mission that led you to parsionate?
DIRK: Mission? I would say courage for medium-sized companies, where we urgently need more innovative power to bring AI potential to the streets. With my work, I want to contribute to managers and executives recognising that digital products, data, and AI key skills must become part of the company.
VERENA: And how do you think you can realise your AI mission for medium-sized companies at and with parsionate?
DIRK: For AI to work, a correspondingly comprehensive data management is needed. parsionate is one of the European market leaders for data management. With the AI activities that we now want to expand, we are therefore focusing on the core competence of the company. There are two topics that are in the front of our minds here.
The first big point is: we know data and we know how to collect, structure and store data. We are already doing this today and we know the corresponding systems. We have the necessary resources and competences. We are now expanding this with innovative data science and AI topics.
The second area is in-depth knowledge of the customer’s business situation – with a view to corporate strategy and business processes. In addition to data management, parsionate relies on strong strategy consulting right from the start. Sharpening the corporate strategy with a view to the topic of AI and data is a necessary component of a successful AI-driven data strategy. From my own experience, I know how important it is to understand the company-wide business processes to start with AI where the result leverage is highest.
How do we proceed in projects? First, data structures and data must be collected and prepared. This is the only way data scientists can work with them efficiently. In my opinion, this is still a problem for most companies today: they are not able to prepare and structure their data in such a way that they can even fuel data science projects. That is already the strength of the parsionate.
The team that we are now additionally building at parsionate is made of employees who have at least five to ten years of professional experience as data scientists and have the required know-how. Therefore, I am deeply convinced that we can create great added value for our customers with our approach.
VERENA: I recently spoke with a consultant colleague who says: “What we do is not AI consulting, but mainly data sorting.” Do you have any provocative statements like that? Do you share the view that 90% of AI pilots are doomed to fail because companies are sitting on a mountain of data rubbish?
DIRK: Of course, the preparation of data is initially a necessary step to establish artificial intelligence as an agile component in the company. Nevertheless, I know from my professional experience that successful AI projects can also be implemented quite early in a company. It is possible that data has to be collected manually at first. But if you select the use cases well, don’t expect miracles from AI and are also prepared to move again at the process level if necessary, and provide resources especially for integration into the productive systems, you can implement successful AI projects.
VERENA: Can you say that 80% of the effort of an AI project is data work for you?
DIRK: In effort yes, in result no.
VERENA: If I go further with this and say: in most companies, it is underestimated how much project time is required to prepare the data. What is the benefit then of using a special team like parsionate? Can you massively speed up the process because you handle this 80% data work in the fast lane?
DIRK: Yes, exactly. If we are honest, AI is far less “miraculous” than one or the other might think at decision-making level. Quite often, you find open-source solutions that only need to be trained in a sensible way. A lot is done via statistical algorithms, which come from the 1950s and 1960s. In terms of content, this is not a high bar. But what is a high bar is indeed building up the data quickly and in a structured way so that a data science team can work well with it. That’s No. 1. and No. 2 is, you have to know which box to reach into, in other words: which tools work for my use case.
VERENA: You say it’s less “miraculous” than you think, but you have to know which box to reach into so that it all fits together in the end. And above all, you must see how to make this elephant, which is 80% of the time, gallop.
DIRK: That is one element. A second element is hidden on a completely different level. Data scientists are often colleagues who are very analytical. But basically, they have little idea of how business processes are structured and where the big handles are. They don’t know which approach can achieve good savings effects or automation effects or sales increase effects. Where one could effectively create benefits in the company. And that’s where parsionate’s consulting unit can help, of course.
We want to look at the entire value chain at every stage. parsionate knows the use cases in the customers’ value chains and where the greatest effects can be achieved. We recognise the cost drivers as well as the manual and redundant work processes. We see where the data structures have not been mapped correctly. And we know where additional data could increase customer satisfaction.
I am convinced that with parsionate we have an excellent basis for jointly designing a data-driven consulting unit. There are many consultants who are very good at process consulting but reach their limits when it comes to data-driven implementation. We, on the other hand, have an excellent consulting approach and we also understand data.
VERENA: Consultancies in the AI environment often want to co-sell their own solution and don’t deal with the business processes at all. Can you tell me where I (as a company) have a sufficiently large volume in the process to make such solutions profitable? Where is my leverage the greatest? Where can I get away with data protection requirements? Where is employee and customer acceptance the highest?
DIRK: We experience over and over again that we can generate quick customer benefits by combining organisational tasks with artificial intelligence, for example, to minimise the susceptibility to errors in the work process. Technically AI-oriented consultancies find this difficult. They sometimes try to jump too far and overwhelm the possibilities of the existing AI systems.
One example is quality control in the call centre. We know that it will take some time before chatbots can really understand and speak human language. But we can already use AI today to check the quality of a customer conversation. Have cross-selling products been offered? Was the customer satisfied? Could his problem be solved with the call? Even with such small steps, a company can significantly improve its processes with the use of AI.
VERENA: The application areas of AI in industry and production are also diverse and promising, from product development and quality management to process optimisation and logistics. Especially if a sufficient database has been created in advance, text, speech, image, and sound recognition can play just as important a role in manufacturing and production as, for example, action planning or multi-dimensional pattern recognition. In the area of “Business AI”, European companies have significant potential to harness the AI power due to our strong industrial and IoT background.
Do you already have a feeling for the use cases that will be easiest to implement for parsionate? Are they use cases in sales, in e-commerce, at the customer interface or rather in the backoffice? Are there already tendencies?
DIRK: There are already many standard products for everything that has to do with recomendations or uplifts. Here, however, the question arises for every company as to how the topic should be strategically positioned. If personalisation is only intended to enable the company to address the customer in a meaningful way, these standard products can be used very well. However, if the customer experience is to be an essential part of the company’s USP, it is more likely to pursue its own solutions with which it can better stand out in the market.
In addition, the derivation of order quantities, the reduction of handling costs in order processing and the optimisation of advertising print are not yet well solved in retail. I am convinced that we will develop even more exciting topics for our customers.
VERENA: Companies can start by asking themselves where each data point comes from and then evaluate in which use case you could use it. Every piece of data has a story. I always find it advisable to create a detailed description of the data with the help of experts. Based on this, the missing values can then be determined.
The idea is to get a full insight into data by focusing on mining, segmenting, and finding patterns before thinking about the use case.
In my opinion, whoever starts such projects should make sure that the external consultants have enough freedom to experiment with the treasure of data in the data warehouse or data lake.
What are the first places you would check when you look into a company?
DIRK: That depends on the industry. Basically, all stages of value creation and all core processes have to be analysed. In the beginning, you can start with individual use cases to learn. But you should quickly get to a point where the company processes are analysed in a structured way regarding the use of AI. This is not so easy, because on the one hand you need to know the processes and on the other hand you need sufficient knowledge about the possibilities that AI already offers today.
I think that all topics that fall into the omnichannel area are important for retailers. Everything that creates frequency, links the channels, and enhances the buying experience will be the top topics of the next 2-3 years.
For machinery and equipment manufacturers, it will be about sales activation and the development of CRM measures. In addition, areas such as service and aftersales must be made more efficient.
VERENA: Let’s say I’m a middle-class person with little money. What do I have to bring with me that I don’t feel ashamed because you are there with me and say, “You don’t have the maturity yet that we can build on”?
DIRK: In this case, we proceed on a use-case basis. The key question is always: what data is available? We would cut the big white elephant you mentioned earlier into slices. Then we would also slice the corresponding data. It will always be possible to identify two or three use cases in customer workshops that have potential. If the potential is great enough and you have the understanding that innovative projects can unfortunately also fail, you should address the issue.
VERENA: If you do it in sales, it’s easy to outline an ROI.
DIRK: Exactly, that’s where the uplift is always at its highest. But we are also looking for projects in other areas and we will find them. The goal is to achieve an increase in results of approximately 2-5 million euros per use case. If you have three of these projects and address them, you should be able to bring one of them to the finish line in such a way that it covers the total investment. We have already spoken with the first parsionate customers about AI topics and shown them good scenarios.
But we don’t have to necessarily sell AI solutions in customer workshops either. Through our strategic consulting approach, our core competence lies in the holistic view of data management. For a medium-sized company, it may make more sense to use a standard product. It is important to us that we can credibly establish the entire program in the company.
On the other hand, there is no industry today that does not deal with AI. Banks have already started to use AI for their customer data – for credit insurance or credit rating checks. In mechanical and systems engineering, the buzzword of the day is “predictive maintenance”. Many have already begun to address this, have built up data warehouses and installed the corresponding sensors. But there is still a lot of catching up to do. Work is being done on AI topics across all industries. But the potential remains a huge open field.
VERENA: Yes, of course. AI in production is gaining importance through predictive maintenance when algorithms for action planning monitor defined parameters and properties of the processes. Who wouldn’t like to determine the optimal maintenance time based on the wear level of operating resources and optimise maintenance activities?
AI can help manufacturing companies at various stages of product development. Processes can be simplified, scaled, and accelerated. Pattern recognition can help to evaluate test and simulation data more efficiently. At the same time, planning and optimisation algorithms can speed up development processes. How do you see companies in this field of possibilities?
DIRK: My impression is that it is not strategically planned or structured in companies. Such projects often arise where an idea and an internal resource were available. What is needed, however, is someone in charge who says: “I have a strategic vision. Our levers are here, there, and there. And that’s why we start with sub-project 1 at this point.” And it would be even better if there was a person sitting there who said, “I’m now building a data management structure that makes Data Scientists fast.”
Now, there are still too many MBAs in the decision-making boards who focus strongly on costs and too little on innovation. The good news is that we now feel that something is changing and that is great, because it also needs people who know the engine room and know which levers, they have to turn.
Hier geht es zum Interview: https://parsionate.com/en/magazine/artificial-intelligence-in-data-management-verena-fink-meets-dirk-hauke/