Interview – There is no stand-alone strategy for AI, it must be part of the company-wide strategy

Ronny FehlingRonny Fehling is Partner and Associate Director for Artificial Intelligence as the Boston Consulting Group GAMMA. With more than 20 years of continually progressive experience in leading business and technology innovation, spearheading digital transformation, and aligning the corporate strategy with Artificial Intelligence he industry-leading organizations to grow their top-line and kick-start their digital transformation.

Ronny Fehling is furthermore speaker of the Predictive Analytics World for Industry 4.0 in May 2020.

Data Science Blog: Mr. Fehling, you are consulting companies and business leaders about AI and how to get started with it. AI as a definition is often misleading. How do you define AI?

This is a good question. I think there are two ways to answer this:

From a technical definition, I often see expressions about “simulation of human intelligence” and “acting like a human”. I find using these terms more often misleading rather than helpful. I studied AI back when it wasn’t yet “cool” and still middle of the AI winter. And yes, we have much more compute power and access to data, but we also think about data in a very different way. For me, I typically distinguish between machine learning, which uses algorithms and statistical methods to identify patterns in data, and AI, which for me attempts to interpret the data in a given context. So machine learning can help me identify and analyze frequency patterns in text and even predict the next word I will type based on my history. AI will help me identify ‘what’ I’m writing about – even if I don’t explicitly name it. It can tell me that when I’m asking “I’m looking for a place to stay” that I might want to see a list of hotels around me. In other words: machine learning can detect correlations and similar patterns, AI uses machine learning to generate insights.

I always wondered why top executives are so frequently asking about the definition of AI because at first it seemed to me not as relevant to the discussion on how to align AI with their corporate strategy. However, I started to realize that their question is ultimately about “What is AI and what can it do for me?”.

For me, AI can do three things really good, which humans cannot really do and previous approaches couldn’t cope with:

  1. Finding similar patterns in historical data. Imagine 20 years of data like maintenance or repair documents of a manufacturing plant. Although they describe work done on a multitude of products due to a multitude of possible problems, AI can use this to look for a very similar situation based on a current problem description. This can be used to identify a common root cause as well as a common solution approach, saving valuable time for the operation.
  2. Finding correlations across time or processes. This is often used in predictive maintenance use cases. Here, the AI tries to see what similar events happen typically at some time before a failure happen. This way, it can alert the operator much earlier about an impending failure, say due to a change in the vibration pattern of the machine.
  3. Finding an optimal solution path based on many constraints. There are many problems in the business world, where choosing the optimal path based on complex situations is critical. Let’s say that suddenly a severe weather warning at an airport forces an airline to have to change their scheduling because of a reduced airport capacity. Delays for some aircraft can cause disruptions because passengers or personnel not being able to connect anymore. Knowing which aircraft to delay, which to cancel, which to switch while causing the minimal amount of disruption to passengers, crew, maintenance and ground-crew is something AI can help with.

The key now is to link these fundamental capabilities with the business context of the company and how it can ultimately help transform.

Data Science Blog: Companies are still starting with their own company-wide data strategy. And now they are talking about AI strategies. Is that something which should be handled separately?

In my experience – both based on having seen the implementations of several corporate data strategies as well as my upbringing at Oracle – the data strategy and AI strategy are co-dependent and cannot be separated. Very often I hear from clients that they think they first need to bring their data in order before doing AI project. And yes, without good data access, AI cannot really work. In fact, most of the time spent on AI is spent on processing, cleansing, understanding and contextualizing the data. However, you cannot really know what data will be needed in which form without knowing what you want to use it for. This is why strategies that handle data and AI separately mostly fail and generate huge costs.

Data Science Blog: What are the important steps for developing a good data strategy? Is there something like a general approach?

In my eyes, the AI strategy defines the data strategy step by step as more use cases are implemented. Rather than focusing too quickly at how to get all corporate data into a data lake, it will be much more important to start creating a use-case, technology and data governance. This governance has to be established once the AI strategy is starting to mature to enable the scale up and productization. At the beginning is to find the (very few) use-cases that can serve as light house projects to demonstrate (1) value impact, (2) a way to go from MVP to Pilot, and (3) how to address the data challenge. This will then more naturally identify the elements of governance, data access and technology that are required.

Data Science Blog: What are the most common questions from business leaders to you regarding AI? Why do they hesitate to get started?

By far it the most common question I get is: how do I get started? The hesitations often come from multiple sources like: “We don’t have the talent in house to do AI”, “Our data is not good enough”, “We don’t know which use-case to start with”, “It’s not easy for us to embrace agile and failure culture because our products are mission critical”, “We don’t know how much value this can bring us”.

Data Science Blog: Most managers prefer to start small and with lower risk. They seem to postpone bigger ideas to a later stage, at least some milestones should be reached. Is that a good idea or should they think bigger?

AI is often associated (rightfully so) with a new way of working – agile and embracing failures. Similarly, there is also the perception of significant cost to starting with AI (talent, technology, data). These perceptions often lead managers wanting to start with several smaller ambition use-cases where failure isn’t that grave. Once they have proven itself somehow, they would then move on to bigger projects. The problem with this strategy is on the one side that you fragment your few precious AI resources on too many projects and at the same time you cannot really demonstrate an impact since the projects weren’t chosen based on their impact potential.

The AI pioneers typically were successful by “thinking big, starting small and scaling fast”. You start by assessing the value potential of a use-case, for example: my current OEE (Overall Equipment Efficiency) is at 65%. There is an addressable loss of 25% which would grow my top line by $X. With the help of AI experts, you then create a hypothesis of how you think you can reduce that loss. This might be by choosing one specific equipment and 50% of the addressable loss. This is now the measure against which you define your failure or non-failure criteria. Once you have proven an MVP that can solve this loss, you scale up by piloting it in real-life setting and then scaling it to all the equipment. At every step of this process, you have a failure criterion that is measured by the impact value.

Virtual Edition, 11-12 MAY, 2020

The premier machine learning
conference for industry 4.0

This year Predictive Analytics World for Industry 4.0 runs alongside Deep Learning World and Predictive Analytics World for Healthcare.

Stop processing the same mistakes! Four steps to business & IT alignment

Digitization. Agility. Tech-driven. Just three strategy buzzwords that promise IT transformation and business alignment, but often fade out into merely superficial change. In fact, aligning business and IT still vexes many organizations because company leaders often forget that transformation is not a move from A to B, or even from A to Z––it’s a move from a fixed starting point, to a state of continual change.

Read this article in German:

Mit den richtigen Prozessen zum Erfolg: vier Schritte zum Business-IT Alignment


Within this state of perpetual flux, adaptive technology is necessary, not only to keep up with industry developments but also with the expansion of technology-enabled customer experiences. After all, alignment assumes that business and technology are separate entities, when in fact they are inextricably linked!

Metrics that matter: From information technology to business technology

Information technology is continuing to challenge the way companies organize their business processes, communicate with customers and potential customers, and deliver services. Although there is no single dominant reorganization strategy, common company structures lean towards decentralizing IT, shifting it closer to end-users and melding the knowledge-base with business strategy. Business-IT alignment is more than ever vital for market impact and growth.

This tactic means as business goals pivot, IT can more readily respond with permanent solutions to support and maintain enterprise momentum. In turn, technological advances and improvements are hardwired into current and future strategies and initiatives. As working ecosystems replace strict organizational structures, the traditional question “Which department do you work in?” has been replaced by, “How do you work?”

But how does IT prove its value and win the trust of the C-suite? Well, according to Gartner, almost 20% of companies have already invested in tools capable of monitoring business-relevant metrics, with this number predicted to reach 60% by 2021. The problem is many infrastructure and operations (I&O) leaders don’t know where to begin when initiating an IT monitoring strategy.

Reach beyond the everyday: Four challenges to alignment

With this, CIOs are under mounting pressure to address digital needs that grow and transform, as well as to renovate the operational environment with new functions. They also must still demonstrate how IT is meeting a given business strategy. So looking forward, no matter how big or small your business is, technology can deliver tangible and intangible benefits (like speed and performance) to hit revenue and operational targets efficiently, and meet your customers’ expectations of innovation.

Put simply, having a good technological infrastructure enriches the culture, efficiency, and relationships of your business.

Business and IT alignment: The rate of change

This continuous strategic loop means enterprises function better, make more profit, and see better ROI because they achieve their goals with less effort. And while there may be no standard way to align successfully, an organization where IT and business strategy are in lock-step can further improve agility and operational efficiencies. This battle of the ‘effs’, efficiency vs. effectiveness, has never been so critical to business survival.

In fact, successful companies are those that dive deeper; such is the importance of this synergy. Amazon and Apple are prime examples—technology and technological innovation is embedded and aligned within their operational structure. In several cases, they created the integral technology and business strategies themselves!

Convergence and Integration

These types of aligned companies have also increased the efficiency of technology investments and significantly reduced the financial and operational risks associated with business and technical change.

However, if this rate of change and business agility is as fast as we continually say, we need to be talking about convergence and integration, not just alignment. In other words, let’s do the research and learn, but empower next-level thinking so we can focus on the co-creation of “true value” and respond quickly to customers and users.

Granular strategies

Without this granular strategy, companies may spend too much on technology without ever solving the business challenges they face, simply due to differing departmental objectives, cultures, and incentives. Simply put, business-IT alignment integrates technology with the strategy, mission, and goals of an organization. For example:

  • Faster time-to-market
  • Increased profitability
  • Better customer experience
  • Improved collaboration
  • Greater industry and IT agility
  • Strategic technological transformation

Hot topic

View webinar recording Empowering Collaboration Between Business and IT, with Fabio Gammerino, Signavio Pre-Sales Consultant.

The power of process: Four steps to better business-IT alignment

While it may seem intuitive, many organizations struggle to achieve the elusive goal of business-IT alignment. This is not only because alignment is a cumbersome and lengthy process, but because the overall process is made up of many smaller sub-processes. Each of these sub-processes lacks a definitive start and endpoint. Instead, each one comprises some “learn and do” cycles that incrementally advance the overall goal.

These cycles aren’t simple fixes, and this explains why issues still exist in the modern digital world. But by establishing a common language, building internal business relationships, ensuring transparency, and developing precise corporate plans of action, the bridge between the two stabilizes.

Four steps to best position your business-IT alignment strategy:

  1. Plan: Translate business objectives into measurable IT services, so resources are effectively allocated to maximize turnover and ROI – This step requires ongoing communication between business and IT leaders.
  2. Model: IT designs infrastructure to increase business value and optimize operations – IT must understand business needs and ensure that they are implementing systems critical to business services.
  3. Manage: Service is delivered based on company objectives and expectations – IT must act as a single point-of-service request, and prioritize those requests based on pre-defined priorities.
  4. Measure: Improvement of cross-organization visibility and service level commitments – While metrics are essential, it is crucial that IT ensures a business context to what they are measuring, and keeps a clear relationship between the measured parameter and business goals.

Signavio Says

Temporarily rotating IT employees within business operations is a top strategy in reaching business-IT alignment because it circulates company knowledge. This cross-pollination encourages better relationships between the IT department and other silos and broadens skill-sets, especially for entry-level employees. Better knowledge depth gives the organization more flexibility with well-rounded employees who can fill various roles as demand arises.

Get in touch

Discover how Signavio can lead your business to IT transformation and operational excellence with the  Signavio Business Transformation Suite. Try it for yourself by registering now for a free 30-day trial.

Data Leader Days 2018

Daten bilden das Fundament der digitalen Transformation. Die richtige Nutzung von Daten entwickelt sich daher zu einer Kernkompetenz und macht im Wettbewerb den Unterschied. Dies gilt sowohl für ganz Unternehmen als auch für einzelne Mitarbeiter, die mit Datennutzung ihre Karriere vorantreiben können.

Erfahrungen von Pionieren und führenden Anwenderunternehmen sind dafür unverzichtbar. Mit den Data Leader Days am 14. und 15. November 2018 in der Digital-Hauptstadt Berlin haben Sie die Chance, direkt von Spitzenkräften aus der Wirtschaft zu lernen und wichtige Impulse für Ihre digitale Weiterentwicklung zu erhalten.

Die Data Leader Days sind das Entscheider-Event für die Datenwirtschaft, das den Schwerpunkt auf die tatsächlichen Nutzer und Anwender-Unternehmen legt. Die Fachkonferenz hat sich seit Gründung im Jahr 2016 als eines der exklusivsten Events rund um die Themen Big Data und künstliche Intelligenz etabliert. In diesem Jahr werden die Data Leader Days erstmalig auf zwei Tage mit unterschiedlichen Schwerpunkten erweitert:

14. November 2018: Commercial & Finance Data

15. November 2018: Industrial & Automotive Data


Die Agenda ist stets aktuell direkt auf zu finden.


Speaker der Data Leader Days 2018




Die Data Leader Days finden dieses Jahr zum dritten Mal statt und haben sich zur Pflichtveranstaltung für Geschäftsführer, Führungskräfte und Professionals aus den Bereichen IT, Business Intelligence und Data Analytics etabliert und empfehlen sich ebenfalls für Leiter der Funktionsbereiche Einkauf, Produktion, Marketing und Finance, die das hier brachliegende Potenzial ausschöpfen wollen.

Zum Event anmelden können sich Teilnehmer direkt auf oder via (Klick).

Ständig wachsende Datenflut – Muss nun jeder zum Data Scientist werden?

Weltweit rund 163 Zettabyte – so lautet die Schätzung von IDC für die Datenmenge weltweit im Jahr 2025. Angesichts dieser kaum noch vorstellbaren Zahl ist es kein Wunder, wenn Anwender in Unternehmen sich überfordert fühlen. Denn auch hier muss vieles analysiert werden – eigene Daten aus vielen Bereichen laufen zusammen mit Daten Dritter, seien es Dienstleister, Partner oder gekaufter Content. Und all das wird noch ergänzt um Social Content – und soll dann zu sinnvollen Auswertungen zusammengeführt werden. Das ist schon für ausgesprochene Data Scientists keine leichte Aufgabe, von normalen Usern ganz zu schweigen. Doch es gibt eine gute Nachricht dabei: den Umgang mit Daten kann man lernen.

Echtes Datenverständnis – Was ist das?

Unternehmen versuchen heute, möglichst viel Kapital aus den vorhandenen Daten zu ziehen und erlauben ihren Mitarbeitern kontrollierten, aber recht weit gehenden Zugriff. Das hat denn auch etliche Vorteile, denn nur wer Zugang zu Daten hat, kann Prozesse beurteilen und effizienter gestalten. Er kann mehr Informationen zu Einsichten verwandeln, Entwicklungen an den realen Bedarf anpassen und sogar auf neue Ideen kommen. Natürlich muss der Zugriff auf Informationen gesteuert und kontrolliert sein, denn schließlich muss man nicht nur Regelwerken wie Datenschutzgrundverordnung gehorchen, man will auch nicht mit den eigenen Daten dem Wettbewerb weiterhelfen.

Aber davon abgesehen, liegt in der umfassenden Auswertung auch die Gefahr, von scheinbaren Erkenntnissen aufs Glatteis geführt zu werden. Was ist wahr, was ist Fake, was ein Trugschluss? Es braucht einige Routine um den Unsinn in den Daten erkennen zu können – und es braucht zuverlässige Datenquellen. Überlässt man dies den wenigen Spezialisten im Haus, so steigt das Risiko, dass nicht alles geprüft wird oder auf der anderen Seite Wichtiges in der Datenflut untergeht. Also brauchen auch solche Anwender ein gewisses Maß an Datenkompetenz, die nicht unbedingt Power User oder professionelle Analytiker sind. Aber in welchem Umfang? So weit, dass sie fähig sind, Nützliches von Falschem zu unterscheiden und eine zielführende Systematik auf Datenanalyse anzuwenden.

Leider aber weiß das noch nicht jeder, der mit Daten umgeht: Nur 17 Prozent von über 5.000 Berufstätigen in Europa fühlen sich der Aufgabe gewachsen – das sagt die Data-Equality-Studie von Qlik. Und für Deutschland sieht es sogar noch schlechter aus, hier sind es nur 14 Prozent, die glauben, souverän mit Daten umgehen zu können. Das ist auch nicht wirklich ein Wunder, denn gerade einmal 49 Prozent sind (in Europa) der Ansicht, ausreichenden Zugriff auf Daten zu haben – und das, obwohl 85 Prozent glauben, mit höherem Datenzugriff auch einen besseren Job machen zu können.

Mit Wissens-Hubs die ersten Schritte begleiten

Aber wie lernt man denn nun, mit Daten richtig oder wenigstens besser umzugehen? Den Datenwust mit allen Devices zu beherrschen? An der Uni offensichtlich nicht, denn in der Data-Equality-Studie sehen sich nur 10 Prozent der Absolventen kompetent im Umgang mit Daten. Bis der Gedanke der Datenkompetenz Eingang in die Lehrpläne gefunden hat, bleibt Unternehmen nur die Eigenregie  – ein „Learning by Doing“ mit Unterstützung. Wie viel dabei Eigeninitiative ist oder anders herum, wieviel Weiterbildung notwendig ist, scheint von Unternehmen zu Unternehmen unterschiedlich zu sein. Einige Ansätze haben sich jedoch schon bewährt:

  • Informationsveranstaltungen mit darauf aufbauenden internen und externen Schulungen
  • Die Etablierung von internen Wissens-Hubs: Data Scientists und Power-User, die ihr Know-how gezielt weitergeben: ein einzelne Ansprechpartner in Abteilungen, die wiederum ihren Kollegen helfen können. Dieses Schneeball-Prinzip spart viel Zeit.
  • Eine Dokumentation, die gerne auch informell wie ein Wiki oder ein Tutorial aufgebaut sein darf – mit der Möglichkeit zu kommentieren und zu verlinken. Nützlich ist auch ein Ratgeber, wie man Daten hinterfragt oder wie man Datenquellen hinter einer Grafik bewertet.
  • Management-Support und Daten-Incentives, die eine zusätzliche Motivation schaffen können. Dazu gehört auch, Freiräume zu schaffen, in denen sich Mitarbeiter mit Daten befassen können – Zeit, aber auch die Möglichkeit, mit (Test-)Daten zu spielen.

Darüber hinaus aber braucht es eine Grundhaltung, die sich im Unternehmen etablieren muss: Datenkompetenz muss zur Selbstverständlichkeit werden. Wird sie zudem noch spannend gemacht, so werden sich viele Mitarbeiter auch privat mit der Bewertung und Auswertung von Daten beschäftigen. Denn nützliches Know-how hat keine Nutzungsgrenzen – und Begeisterung steckt an.


Nothing Found

Sorry, no posts matched your criteria