Welcome to the AIoT Playbook! Let`s start by looking at the following questions: How did we get here? What do we mean by smart, connected products (and solutions)? What are the key plays of the AIoT Playbook? And how to best read the book?
About the AIoT Playbook and the AIoT User Group
In January 2020, a hand full of senior IT experts and enthusiasts from different companies met at the Bosch Connectory in Stuttgart to exchange their experiences and views on AI and IoT. AI was at the peak of a new hype, fueled by Alpha Go, advancements in autonomous driving, and not to forget about Cambridge Analytica. The general feeling was that - with the exception of autonomous driving - AI had not really arrived in the world of IoT. Sure, every IoT article or presentation in the last 10 years had mentioned predictive maintenance, but in reality many IoT applications were still much more basic. So how could one better utilize AI in the world of physical products, manufacturing, and equipment operations? The workshop was organized as an open exchange, with a mixture of presentations and group discussions. After three days, there was so much excitement about the topic and the way collaboration in the group worked, that it was decided to make this a regular thing. The result was the formation of the AIoT User Group, a loosely coupled, non-profit network of AI and IoT practitioners, who work together to exchange experiences and best practices on the application of AI in the IoT. Throughout 2021, local chapters were set up in Singapore (special thanks to CK and Thomas!), Shanghai (Nǐ hǎo, Gene and Cherry!) and Chicago (hi Fermin and Hans!).
Picture: First AIoT User Group meeting, with practitioners from Accenture, Bosch, Clariba, Deutsche Post, Evaco, mm1, Opitz, Recogizer, TH Köln and Tomorrow Labs, at the Bosch Connectory in Stuttgart.
Over time, it became clear that it would make sense to document the collected wisdom in a good practice framework - this is how the AIoT Playbook started. Content creation is driven by experts in different domains (see the AIoT Expert Network). The AIoT Editorial Board provides strategic guidance and management support. The basic working mode are so-called Unplugged-Sessions, where the real work on the Playbook is happening. All the material is developed as open source content (using CC BY 4.0), and is also used as foundation for different AIoT-related training courses.
If you are interested in joining the AIoT User Group, good starting points are the website aiot.rocks, as well as the AIoT User Group on LinkedIn. The main site for the AIoT Playbook is simply aiotplaybook.org.
If you are reading the AIoT Playbook as a PDF or on an eBook reader, you might sometimes find that not everything is perfect, like in a fully edited book. The reason is that the Playbook is constantly evolving, and so the decision was made to use a Wiki as the foundation for the playbook. The other book formats are derived as snapshots from the wiki, and the conversion is sometimes not perfect. Also, some content might sometimes still not be perfectly ready. This was a tradeoff between having the perfect book on the one hand, and having an open, digital platform that can evolve over time, on the other. Since the book formats are published in the open access format, this should hopefully also be acceptable to all readers of the offline versions.
Smart, connected products (and solutions)
So what is AIoT all about? AIoT combines AI and connectivity for physical products. These can be new product categories, or retrofit solutions for existing assets and equipment in the field. The general idea is summarized in the figure below: physical producs (e.g. a forklift) are end points with physical components and on-board computing (combining hardware, software and AI). The product is uniquely identified, captures relevant status information, and perceives its environment. The product is integrated with business processes (e.g. warehousing), and solves specific customer problems (e.g. optimizing warehousing tasks). The product is exchanging data in a closed loop with a backend (e.g. cloud or on-premise systems). In the cloud, the product has a digital twin, an anlytics component, and is controlled by AI or ML (Machine Learning). Via this setup, additional information can be exchanged with other customer systems (e.g. an ERP system).
In all of this, AI can enable new functionality either on-board the product, or in the backend. IoT provides the required connectivity between the product and the backend. Digital Twins are a digital presentation of the real, physical product - providing abstraction, standardization and a rich, semantic view on the AIoT data. AI can be used to help create Digital Twins, or to build applications which utilize them.
Of course this is a big vision, which will not become a reality for each product category over night. But it shows the potential of AIoT. Also, not all projects might look at such a high level of productization and deep integration - AIoT can also support more basic retrofit approaches.