Sebastian Petry, the Head of b.telligent’s “Data Science & AI” Competence Center talks about technology, confidence, and know-how on setting up AI in an enterprise.
Beside a good use case, what are the decisive factors that ensure the success of AI in an enterprise?
Of three key factors, the foremost is technology, or essentially cloud technologies, big data, bots, and AI frameworks, plus a host of diverse solutions and tools on the market. If you want to set up one or more of these in your enterprise, you must first define the scope of AI you wish to have. The question is, are you looking at a special product, or a particular part of your enterprise, or something you’d like to scale up and consider as your data or AI strategy too? How do you want to use, run, and advance your AI? You must address these issues first – or you’ll be disadvantaged right from the start, if you implement a solution that does not fit into your current infrastructure, data landscape, corporate philosophy, or workflows. The other way around, you’ll be a winner.
Control or confidence? Which strategy is better with AI in your enterprise?
This is the next key factor, although both play a decisive role, depending on which field we are talking about. If you integrate AI into your enterprise as a part of a campaign set, for instance, the goal is to let this technology handle tasks previously done by humans. For that, you must gain confidence in the respective field – will the machine be as good as you? Yes, but supposedly better. Then how do you control something that will be better than you could ever be? To get useful answers here, you need not just high-quality AI, but also reports and KPIs that the AI can assess and control. To judge the AI’s functioning in the overall context, you have to ask yourself how high a ROI you generate there. You can gain more confidence by ensuring that a data scientist or AI expert ensures proper internal communications in that department. In large enterprises, it will become increasingly important to work with other departments like the one for BI reports, by having the AI transmit performance data using self-service BI tools to the various departments.
Know-how in your enterprise is the third key factor. What should one heed?
If you choose to set up a complex thing like AI in your enterprise, you need to have a realistic view of how much you yourself can shape. Do you wish to develop AI with something like TensorFlow and embed deep learning in the depths of your enterprise? If so, you’ll need a highly qualified methodology expert or a data scientist. Or would it be okay to use tools that you could configure and optimize for yourself? In that case, you won’t require a particular deep learning expert, but rather a professional who knows how to handle the tools. Generally, this does not call for expertise in algorithms. Anyway, no matter which route you take, to succeed with AI in your enterprise you still need clarity on such issues.
What other aspects must you consider when going for AI at your enterprise?
The GDPR (Data Protection Regulation) just highlighted how poorly prepared most enterprises were for its implementation. AI is not a brand or marketing issue for tech companies. It will penetrate ever more into our daily lives at almost every enterprise. We recommend that you should set up your own data science department responsible for AI, built up on a very good BI and IT base. If you start from scratch, it may take up to three years to collect enough data, integrate the right technologies, build up internal know-how, and establish the right processes.
Small enterprises, in particular, can’t afford to avoid working with AI and its precursors. Any desire to use AI for your data means bearing a great responsibility. You just can’t assign this to a fresh recruit from a university. You need specialists with experience and knowledge of what can go wrong. Successful AI projects are often underpinned by the experiences of employees in this field, because in almost every AI project at the moment you will encounter virgin territory. Although AI is fascinating, especially in view of the algorithms, it is certainly something highly technical that must be embedded in an enterprise and resolved through big data technologies. Deploying AI and operating modern methods pose challenges for many enterprises.
b.telligent decided to rename its “Data Science” competence center to “Data Science & AI”. How come?
This was really a conscious decision. For a decade now, we have been automating processes and integrating intelligence into systems that we set up, manage, and continue to develop for our clients. We are good at this. Today, AI is defined as emulating intelligent actions with the aid of mathematical algorithms, under special consideration of how to deal with insecurity, and learning continually from new situations. That’s precisely where b.telligent’s history is quite mature. Our team of 15 data scientists and AI experts has grown along with the complexity of such processes. Many of us were statisticians and have a sound knowledge of the methods and algorithms of conventional machine learning processes. For us, deep learning is a natural extension of the data science toolbox – whose common designation we have adapted, along with the times, from business analytics, to data science, to artificial intelligence. Earlier, in our initial meetings, whenever our clients and job applicants asked us whether we applied artificial intelligence in our projects, we always answered affirmatively. Now we simply call it Data Science & AI.
How does b.telligent compare against other firms offering AI consulting services?
We have steadily enhanced our expertise with every project we conducted. The greatest benefit for you is that for AI issues we can pull in colleagues from our other competence centers, such as those for data integration, big data, and cloud services. This interdisciplinary exchange within our firm is the backbone of our strength. Our Data Science & AI competence center stands out from others, primarily in terms of application, development, and implementation of mathematical algorithms. For numerous enterprises, we have already served as the door opener into this novel world.