No edit summary
 
(6 intermediate revisions by 2 users not shown)
Line 12: Line 12:
</imagemap>
</imagemap>


The Digital OEM combines physical product design, engineering and manufacturing with Software-as-a-Service in order to provide smart, connected products. Artificial Intelligence (AI) and Internet of Things (IoT) are the two key enablers. This section will introduce the concept of the Digital OEM in detail, following again the ''why'', ''what'', ''how'' structure from the introduction
The Digital OEM combines physical product design, engineering and manufacturing with Software-as-a-Service in order to provide smart, connected products. Artificial Intelligence (AI) and the Internet of Things (IoT) are the two key enablers. This section will introduce the concept of the Digital OEM in detail, following again the ''why'', ''what'', ''how'' structure from the introduction


__TOC__
__TOC__
Line 19: Line 19:
The motivation for adopting a digital OEM business model can vary widely. Many incumbent OEMs are seeking ways to build upon their existing business. New market entrants are looking at disruptive new business models enabled by the combination of physical products with AIoT.
The motivation for adopting a digital OEM business model can vary widely. Many incumbent OEMs are seeking ways to build upon their existing business. New market entrants are looking at disruptive new business models enabled by the combination of physical products with AIoT.


While AI and IoT are exciting technical enablers, anybody embarking on the AIoT journey should always start by looking at the "why": What is the purpose? And what are the expected business outcomes? From a strategic (and emotional) point of view, the purpose of the AIoT initiative should be clearly articulated: What is the belief? The mission? Why is this really done?
While AI and IoT are exciting technical enablers, anybody embarking on the AIoT journey should always start by looking at the "why": Why do this? What is the purpose? And what are the expected business outcomes? From a strategic (and emotional) point of view, the purpose of the AIoT initiative should be clearly articulated: What is the belief? The mission? Why is this truly done?


== Digital OEMs - business models ==
== Digital OEMs - business models ==
At the core of the digital OEM business model is the physical asset or product. The interesting question is which new opportunities are arising through the combination of physical products with digital solutions. Examples include:
At the core of the business model of the Digital OEM is the physical asset or product. An interesting question is which new opportunities arise through the combination of physical products with digital solutions. Examples include:
* Data-driven business, e.g. building on user-generated data or asset/product performance-related data. Examples include usage-based car insurance (UBI), data-driven aftermarket services, or [[Hybrid_Models#drone|drone-based building facade inspection]].
* Data-driven business, e.g., building on user-generated data or asset/product performance-related data. Examples include usage-based car insurance (UBI), data-driven aftermarket services, or [[Hybrid_Models#drone|drone-based building facade inspection]].
* Digital add-on services, e.g. an optional autopilot service for an electric vehicle, or cooking recipe add-ons for a smart kitchen appliance
* Digital add-on services, e.g., an optional autopilot service for an electric vehicle, or cooking recipe add-ons for a smart kitchen appliance
* Asset-as-a-Service, e.g. car-seat-heating-as-a-service, or the famous [https://www.rolls-royce.com/media/press-releases-archive/yr-2012/121030-the-hour.aspx "power-by-the-hour"] for Rolls-Royce aircraft engines
* Asset-as-a-Service, e.g., car-seat-heating-as-a-service, or the famous [https://www.rolls-royce.com/media/press-releases-archive/yr-2012/121030-the-hour.aspx "power-by-the-hour"] for Rolls-Royce aircraft engines
* Smart Maintenance, including predictive, preventive and prescriptive maintenance, enabled by deep analytics of asset/machine data via AIoT
* Smart Maintenance, including predictive, preventive and prescriptive maintenance, enabled by deep analytics of asset/machine data via AIoT


Line 31: Line 31:


The diagram shows key elements of two worlds:
The diagram shows key elements of two worlds:
* The OEM (Original Equipment Manufacturer) is an organization which makes devices from component parts bought from other organizations. This can be a car maker, a manufacturer of household appliances, or a manufacturer of manufacturing equipment, such as robots or laser cutting tools.
* The OEM (Original Equipment Manufacturer) is an organization that makes devices from component parts bought from other organizations. This can be a car maker, a manufacturer of household appliances, or a manufacturer of manufacturing equipment, such as robots or laser cutting tools.
* The suppliers of the OEM are usually referred to as "tier 1", "tier 2", etc., depending on their position in the value chain
* The suppliers of the OEM are usually referred to as "tier 1", "tier 2", etc., depending on their position in the supply chain
* On the other side, we have the digital ecosystems. Today, large hyperscalers are dominating cloud-based infrastructure (Infrastructure-as-a-Service, or IaaS) and platforms (Platform-as-a-Service). IaaS includes storage, networking, and virtual compute resources. PaaS includes Internet-based tools and middleware for building applications
* On the other side, we have the digital ecosystems. Today, large hyperscalers are dominating cloud-based infrastructure (Infrastructure-as-a-Service, or IaaS) and platforms (Platform-as-a-Service). IaaS includes storage, networking, and virtual compute resources. PaaS includes Internet-based tools and middleware for building applications
* Software-as-a-Service (or SaaS) are applications delivered over the internet.
* Software-as-a-Service (or SaaS) are applications delivered over the internet.
Line 38: Line 38:


== Incumbent OEMs - business improvements ==
== Incumbent OEMs - business improvements ==
Especially for incumbent OEMs, the idea of improving existing business by adding digitally-enabled solutions is attractive. Generating ARR (Annual Recurring Revenue) via digital services is very interesting, since ARR is seen as a more stable and predictable revenue stream. However, the opportunity to improve existing business -- and especially EBIT -- with digital solutions as a short-term measure should not be underestimated, since unproven, new business models can have inherent risks and realization of new, ARR-like revenues might take longer than hoped for.
Especially for incumbent OEMs, the idea of improving existing business by adding digitally enabled solutions is attractive. Generating ARR (Annual Recurring Revenue) via digital services is very interesting, since ARR is seen as a more stable and predictable revenue stream. However, the opportunity to improve existing business -- and especially EBIT -- with digital solutions as a short-term measure should not be underestimated, since unproven, new business models can have inherent risks and realization of new, ARR-like revenues might take longer than hoped for.
[[File:0.2.1 EBIT ARR.png|600px|frameless|center|link=|Business Outcomes]]
[[File:0.2.1 EBIT ARR.png|600px|frameless|center|link=|Business Outcomes]]


= WHAT =
= WHAT =
What can be done with AIoT from the perspective of the digital OEM? Usually, the answer is building smart, connected products. These combine physical products with smartness enabled by AI and connectivity enabled by IoT. To build smart, connected products, the digital OEM needs to combine product engineering and manufacturing capabilities with edge and cloud software development capabilities.
What can be done with AIoT from the perspective of the digital OEM? Usually, the answer is building smart, connected products. These combine physical products with smartness enabled by AI and connectivity enabled by the IoT. To build smart, connected products, the digital OEM needs to combine product engineering and manufacturing capabilities with edge and cloud software development capabilities.


[[File:0.2.1 Digital OEM and SCP.png|800px|frameless|center|link=|WHAT: Digital OEM and smart, connected products]]
[[File:0.2.1 Digital OEM and SCP.png|800px|frameless|center|link=|WHAT: Digital OEM and smart, connected products]]


== Smart, connected products - enabled by AIoT==
== Smart, Connected Products - Enabled by AIoT==
Smart, connected products usually combine edge and cloud computing capabilities: Edge computing is anything that happens on (or near) the asset/product in the field. Edge computing capabilities are usually dedicated to a single asset/product, or sometimes a specific cluster of assets/products operating in close proximity. Cloud computing in an AIoT scenario on the other hand can enable insights or functionality which relates to an entire fleet (or "swarm") of assets/products. Consequently, in AIoT we also differentiate between two types of intelligence: asset/product intelligence vs. swarm intelligence.
Smart, connected products usually combine edge and cloud computing capabilities: Edge computing is anything that happens on (or near) the asset/product in the field. Edge computing capabilities are usually dedicated to a single asset/product or sometimes a specific cluster of assets/products operating in close proximity. Cloud computing in an AIoT scenario, on the other hand, can enable insights or functionality that relates to an entire fleet (or "swarm") of assets/products. Consequently, in AIoT, we also differentiate between two types of intelligence: asset/product intelligence vs. swarm intelligence.


[[File:0.2.1 SCP.png|800px|frameless|center|link=|Smart, connected products]]
[[File:0.2.1 SCP.png|800px|frameless|center|link=|Smart, connected products]]


== Example: Robot vacuum cleaner ==
== Example: Robot Vacuum Cleaner ==
A good example for a smart, connected product is a robot vacuum cleaner. These products [https://www.theverge.com/2020/8/25/21377585/irobot-ai-software-update-home-intelligence-genius-app use AI] to identify room layouts and obstacles, and to compute efficient routes and methods. For example, the robot can decide to make a detour vs. switching into the build-in „climb over obstacle“-mode. Another example is the automatic activation of a „carpet boost“ mode. IoT-connectivity to the cloud enables integration with user interface technology such as smart mobile devices or smart home appliances for voice control („clean under the dining room table“).
A good example of a smart, connected product is a robot vacuum cleaner. These products [https://www.theverge.com/2020/8/25/21377585/irobot-ai-software-update-home-intelligence-genius-app use AI] to identify room layouts and obstacles and to compute efficient routes and methods. For example, the robot can decide to make a detour vs. switching into the built-in "climb over obstacle" mode. Another example is the automatic activation of a "carpet boost" mode. IoT connectivity to the cloud enables integration with user interface technology such as smart mobile devices or smart home appliances for voice control ("clean under the dining room table").


[[File:0.2.1 Robo Vacuum.png|800px|frameless|center|link=|Example: Robot Vacuum Cleaner]]
[[File:0.2.1 Robo Vacuum.png|800px|frameless|center|link=|Example: Robot Vacuum Cleaner]]


The robot vacuum example will be looked at in great detail in the [[AIoT_Business_Viewpoint|product design]] section.
The vacuum robot example will be examined in great detail in the [[AIoT_Business_Viewpoint|product design]] section.


== Example: Kitchen Appliance==
== Example: Kitchen Appliance==
Another good example for smart, connected products is a smart kitchen appliance. Here, the intelligence could start with data gathered from users of the kitchen appliances, in combination with user-generated ratings. This data could be combined to make targeted recommendations (created via AI), e.g. for cooking recipes. More advanced version of the smart kitchen appliance could also use AI on the product, e.g. for better device control and maintenance.
Another good example for smart, connected products is a smart kitchen appliance. Here, the intelligence could start with data gathered from users of kitchen appliances in combination with user-generated ratings. These data could be combined to make targeted recommendations (created via AI), e.g., for cooking recipes. A more advanced version of the smart kitchen appliance could also use AI on the product, e.g., for better device control and maintenance.
[[File:0.2.1 Kitchen Appliance.png|800px|frameless|center|link=|Example: Kitchen Appliance]]
[[File:0.2.1 Kitchen Appliance.png|800px|frameless|center|link=|Example: Kitchen Appliance]]


== Example: Automatic wiper control ==
== Example: Automatic Wiper Control ==
In this example, AI is utilizing images from the autopilot camera to determine the local weather situation. This is then used to automatically match wiper speed to the intensity of rain or snow. This is how [https://electrek.co/2019/10/14/tesla-deep-rain-neural-net-automatic-wipers/ Tesla] is doing it, and it's an area which is also starting to get the attention of the [https://www.researchgate.net/publication/335095551_Video-Based_Windshield_Rain_Detection_and_Wiper_Control_Using_Holistic-View_Deep_Learning research community].
In this example, AI utilizes images from the autopilot camera to determine the local weather situation. This is then used to automatically convert the wiper speed to the intensity of rain or snow. This is how [https://electrek.co/2019/10/14/tesla-deep-rain-neural-net-automatic-wipers/ Tesla] is doing it, and it is an area that is also starting to receive the attention of the [https://www.researchgate.net/publication/335095551_Video-Based_Windshield_Rain_Detection_and_Wiper_Control_Using_Holistic-View_Deep_Learning research community].


What is interesting about this example is that some Tesla customers have been initially complaining that this is not as accurate as other systems using rain sensors. So Tesla was using their [[OTA_Updates|Over-the-Air Update (OTA)]] capabilities to enhance this function.
What is interesting about this example is that some Tesla customers initially complained that this was not as accurate as other systems using rain sensors. Over time, Tesla was using their [[OTA_Updates|Over-the-Air Update (OTA)]] capabilities to enhance this function, using continuous model improvements and retraining.


[[File:0.2.1 Wiper control.png|800px|frameless|center|link=|Example: Windshield wiper control]]
[[File:0.2.1 Wiper control.png|800px|frameless|center|link=|Example: Windshield wiper control]]


== Example: Physical product design improvements ==
== Example: Physical Product Design Improvements ==
Another interesting use of AIoT is for the advanced analytics of product performance, based on data from assets in the field. For example, the team developing the electric motor for the wiper blades from the previous example could use this approach to better understand how their product performs in the field, e.g. at 200 kph on a highway at heavy rain. This information can then be used to improve the next generation of the motor. In this case, it might sometimes not be clear whether we are talking about advanced analytic or real AI (e.g. using ML), but it is still an important use case.
Another interesting use of AIoT is for the advanced analytics of product performance, based on data from assets in the field. For example, the team developing the electric motor for the wiper blades from the previous example could use this approach to better understand how their product performs in the field, e.g., at 150 kph on a highway under heavy rain. This information can then be used to improve the next generation of the motor. In this case, it might sometimes not be clear whether we are talking about advanced analytic or real AI (e.g., using ML), but it is still an important use case.


[[File:0.2.1 Product Design.png|800px|frameless|center|link=|Example: Physical product improvements]]
[[File:0.2.1 Product Design.png|800px|frameless|center|link=|Example: Physical product improvements]]


== Example: Smart tightening tool ==
== Example: Smart Tightening Tool ==
Another example is the smart tightening tool (e.g. the Bosch Rexroth [https://www.boschrexroth.com/en/us/products/product-groups/tightening-technology/topics/nexo-cordless-wi-fi-nutrunner/index Nexo cordless Wi-Fi nutrunner]). This is a type of tool used by industrial customers, e.g. for ensuring the quality of safety relevant joints.
Another example is the smart tightening tool (e.g., the Bosch Rexroth [https://www.boschrexroth.com/en/us/products/product-groups/tightening-technology/topics/nexo-cordless-wi-fi-nutrunner/index Nexo cordless Wi-Fi nutrunner]). This is a type of tool used by industrial customers, e.g., for ensuring the quality of safety relevant joints.


On the tightening tool, AI/ML can be used to control the proper execution of tightening programs (controlling torque and angle for specific combinations of materials). In the cloud, data from fleets of tightening tools can be analyzed to help automatically detect tightening anomalies, classify these anomalies, and make recommendations for handling of these anomalies.
On the tightening tool, AI/ML can be used to control the proper execution of tightening programs (controlling torque and angle for specific combinations of materials). In the cloud, data from fleets of tightening tools can be analyzed to help automatically detect tightening anomalies, classify these anomalies, and make recommendations for handling these anomalies.


[[File:0.2.1 Tightening.png|800px|frameless|center|link=|Example: Smart, connected tightening tool]]
[[File:0.2.1 Tightening.png|800px|frameless|center|link=|Example: Smart, connected tightening tool]]


= WHY revisited =
= WHY revisited =
Let`s revisit the "WHY" perspective with what we have learned so far about AIoT and the different use cases implemented by Digital OEMs.
Let us revisit the "WHY" perspective with what we have learned thus far about AIoT and the different use cases implemented by Digital OEMs.
=== Aligning the Product Lifecycle with the Customer Journey ===
== Aligning the Product Lifecycle with the Customer Journey ==
A key feature of AIoT is that it helps with aligning the product lifecycle and the customer journey. In the past, most OEMs lost contact with their products once they left the factory. Although many OEMs try to stay in touch with their customers and support them in the aftermarket, in most cases the customer relationship was based on service contracts, but not a digital relationship. This is of course changing with AIoT, which enables a much higher level of customer intimacy because OEMs cannot learn how their products are used in the field, and how they are performing. The data obtained and analyzed from the products in the field via AIoT can be augmented with other data, e.g. customer feedback from the Internet.
A key feature of AIoT is that it helps align the product lifecycle and the customer journey. In the past, most OEMs lost contact with their products once they left the factory. Although many OEMs try to stay in touch with their customers and support them in the aftermarket, in most cases, the customer relationship was based on service contracts but not a digital relationship. This is of course changing with AIoT, which enables a much higher level of customer intimacy because OEMs can now learn how their products are used in the field and how they are performing. The data obtained and analyzed from the products in the field via AIoT can be augmented with other data, e.g., customer feedback from the Internet.


AIoT also gives the OEM the opportunity to react to what he is learning about his products in the field, by constantly updating existing digital features or even creating new ones, deployed via [[OTA_Updates|Over-the-Air Updates (OTA)]]. Naturally, OTA in an AIoT setting will have to support updates of both software and AI models.
AIoT also gives the OEM the opportunity to react to what he is learning about his products in the field, by constantly updating existing digital features or even creating new ones, deployed via [[OTA_Updates|Over-the-Air Updates (OTA)]]. Naturally, OTA in an AIoT setting will have to support updates of both software and AI models.
Line 90: Line 90:
[[File:0.2.1 LCM.png|800px|frameless|center|link=|WHY revisited: Product LCM and Customer Journey]]
[[File:0.2.1 LCM.png|800px|frameless|center|link=|WHY revisited: Product LCM and Customer Journey]]


This topic was actually recently discussed by Uli Homann of Microsoft at the BCW.on session with Microsoft CEO Satya Nadella and Bosch CEO Volkmar Denner. Full video [https://bit.ly/3a9uUkt here].
This topic was recently discussed by Uli Homann of Microsoft at the BCW.on session with Microsoft CEO Satya Nadella and Bosch CEO Volkmar Denner. Full video [https://bit.ly/3a9uUkt here].


[[File:Homann AIoT Cycle.png|800px|frameless|center|Uli Homann discusses AIoT Cycle|link=https://www.youtube.com/watch?v=eE4NeZwmAeU]]
[[File:Homann AIoT Cycle.png|800px|frameless|center|Uli Homann discusses AIoT Cycle|link=https://www.youtube.com/watch?v=eE4NeZwmAeU]]


Uli Homann, Corporate Vice President, Microsoft: ''The digital feedback loop is essential for successful product development, and OEMs and manufacturers are now also starting to embrace it. For example, we are seeing more and more connected vehicles on the street, which are bringing data into a centralized cloud environment. And the cloud is then able to reason over that data and deduce information. Tesla is one of the very famous users of this digital feedback loop already, where they actually use two components. One, the car itself, where it brings information back based upon telemetry, instrumentation, and so forth. So how hard were the brakes being used, if the autopilot is going around the corner? How tight was the corner taken? And then human feedback. Elon is very, very active on Twitter, sometimes very positive, sometimes, to distraction. But he's very, very active for a very good reason, because he's looking for feedback. One very famous case was people complaining that the Model 3 was taking the corners too hard from their perspective. And so he took the feedback, they compared it with the feedback from the car, and then they made adjustments to the auto drive. And that is really what we call the digital feedback loop. Because on the one side, you have instrumentation from the car, but you have other channels as well that you bring together and that allow you to start to really think about the lifecycle of the customer journey, the customer buying the car, finding the right car, servicing the car and those kinds of things, and bringing all that data, all of this awareness back into the engineering cycle from design to manufacturing, to the sales and after sales, after market opportunities, etc. Bringing this together in an intelligent way based upon data, utilizing AIoT , is really the key piece here. And the last dimension of making this happen is are open platforms - both from an approach to software development as an open ecosystem, with open tools and a lot of open source in the cloud, and also open standards coming together not only in the cloud, but extending this reach into the car. The resulting programming model has platform capabilities underneath that are derived from the Cloud, optimized for the car. Making this consistently happen will not only allow us to enabled AIoT in the cloud, but also bringing cloud into the car or into the manufacturing capability. I think the digital feedback loop, the platform tooling and then bringing it into a consistent end-to-end perspective really will help ensure that we can get digital services at your fingertips. And again, Microsoft is part of an open ecosystem here. We are working together with Bosch and other players to actively bring this to bear, to real life so that we can really drive this vision forward.''
Uli Homann, Corporate Vice President, Microsoft: ''The digital feedback loop is essential for successful product development, and OEMs and manufacturers are now also starting to embrace it. For example, we are seeing an increasing number of connected vehicles on the street, which are bringing data into a centralized cloud environment. The cloud is then able to reason over that data and deduce information. Tesla is one of the very famous users of this digital feedback loop already, where they actually use two components. One is the car itself, where it brings information back based upon telemetry, instrumentation, and so forth. So how hard were the brakes being used, if the autopilot is going around the corner? How tight was the corner taken? And then human feedback. Elon is very, very active on Twitter, sometimes very positive, sometimes, to distraction. However, he's very, very active for a very good reason: because he's looking for feedback. One very famous case was people complaining that the Model 3 was taking corners too hard, from their perspective. And so he took the feedback, they compared it with the feedback from the car, and then they made adjustments to the auto drive. And that is truly what we call the digital feedback loop. Because on the one side, you have instrumentation from the car, but you have other channels as well that you bring together and that allow you to start to really think about the lifecycle of the customer journey, the customer buying the car, finding the right car, servicing the car and those kinds of things, and bringing all that data, all of this awareness back into the engineering cycle from design to manufacturing, to the sales and after sales, after market opportunities, etc. Bringing this together in an intelligent way based upon data, utilizing AIoT, is truly the key piece here. The last dimension of making this happen are open platforms, both from an approach to software development as an open ecosystem, with open tools and a lot of open source in the cloud, and also open standards coming together not only in the cloud but also extending this reach into the car. The resulting programming model has platform capabilities underneath that are derived from the Cloud, and optimized for the car. Making this happen consistently will not only allow us to enable AIoT in the cloud but also bringing cloud into the car or into the manufacturing capability. I think the digital feedback loop, the platform tooling and then bringing it into a consistent end-to-end perspective truly will help ensure that we can get digital services at your fingertips. Again, Microsoft is part of an open ecosystem here. We are working together with Bosch and other players to actively bring this to bear, to real life so that we can truly drive this vision forward.''


== Benefits ==
== Benefits ==
The benefits of this approach are manifold, including shorter time-to-market, improved differentiation, improved sales (including recurring revenues), improved customer experience, and consequently also improved customer loyalty.
The benefits of this approach are manifold, including shorter time-to-market, improved differentiation, improved sales (including recurring revenues), improved customer experience, and consequently improved customer loyalty.
[[File:0.2.1 Benefits.png|800px|frameless|center|link=|WHY: Benefits]]
[[File:0.2.1 Benefits.png|800px|frameless|center|link=|WHY: Benefits]]


= HOW =
= HOW =
Now let`s take a closer look at how the Digital OEM must go about implementing this with AIoT. This will include a discussion of key design decisions, technical constraints, and considerations for execution and delivery.
Now let us take a closer look at how the Digital OEM must go about implementing this with AIoT. This will include a discussion of key design decisions, technical constraints, and considerations for execution and delivery.


== Key design decisions <span id="KeyDecisions"></span>==
== Key Design Decisions <span id="KeyDecisions"></span>==
From the product manager`s perspective, a key question in the future will be - for each feature - whether this feature should be implemented in hardware, software, or AI - or combinations thereof. Implementing a feature in hardware (including HMI, processing, etc.) will have an impact on usability (for example, sometimes it will still be preferable to activate a feature via a physical control), but also on engineering and design complexity. Implementing the same feature completely in software (e.g. as a feature activated via a smart app) can often mean lower cost of delivery (no manufacturing / supply costs beyond the initial development), and also means that the feature can be updated via OTA in the future. Finally, if the feature can be implemented virtually, then the next big question is whether it should be implemented as a set of hard-coded rules (software development), or as a data-centric AI function which uses inference to make a decision based on its training.
From the product manager's perspective, a key question in the future will be -- for each feature -- whether this feature should be implemented in hardware, software, or AI/data, or combinations thereof. Implementing a feature in hardware (including HMI, processing, etc.) will have an impact on usability (for example, sometimes it will still be preferable to activate a feature via a physical control) but also on engineering and design complexity. Implementing the same feature completely in software (e.g., as a feature activated via a smart app) can often mean a lower cost of delivery (no manufacturing/supply costs beyond the initial development) and also means that the feature can be updated via OTA in the future. Finally, if the feature can be implemented virtually, then the next big question is whether it should be implemented as a set of hard-coded rules (software development) or as a data-centric AI function that uses inference to make a decision based on its training.


The decision for using AI, Software, or Hardware for a specific feature will have two main implications: first of all, the quality of the User Experience (UX), and secondly, the required technology pipeline to deliver the feature.
The decision to use AI, Software, or Hardware for a specific feature will have two main implications: first, the quality of the User Experience (UX), and second, the required technology pipeline to deliver the feature.


[[File:0.2.1 How Features.png|800px|frameless|center|link=|HOW: Key design decisions|class=Digital_OEM#Conclusion]]
[[File:0.2.1 How Features.png|800px|frameless|center|link=|HOW: Key design decisions|class=Digital_OEM#Conclusion]]


== Technical constraints ==
== Considerations for Execution and Delivery==
The decisions for the system design of each feature will also have to take technical constraints into consideration, as shown below. A more detailed discussion of technical constraints and how they must be addressed as part of the system design can be found in the section of product/solution design, specifically the [[AIoT_Data_and_Functional_Viewpoint#componentization|functional viewpoint]].
 
 
[[File:0.2.1 How Constraints.png|800px|frameless|center|link=|HOW: Technical Constraints]]
 
== Considerations for execution and delivery==
For the digital OEM, execution and delivery will require a holistic view, including business model, leadership & organization, sourcing and co-creation, User Experience (UX) and Human/Machine Interfaces (HMI), data strategy, AIoT architecture, DevOps, Digital Trust and Security, Quality Management, Compliance and Legal, Productization and Sales; and how AIoT will impact them.  
For the digital OEM, execution and delivery will require a holistic view, including business model, leadership & organization, sourcing and co-creation, User Experience (UX) and Human/Machine Interfaces (HMI), data strategy, AIoT architecture, DevOps, Digital Trust and Security, Quality Management, Compliance and Legal, Productization and Sales; and how AIoT will impact them.  


What is usually less relevant for the Digital OEM are aspects such as retrofit (assuming the approach here will be a line-fit one), site preparation and rollout. These are all important aspects for the Digital Equipment Operations, which will be discussed next.
What is usually less relevant for the Digital OEM are aspects such as retrofit (assuming the approach here will be a line-fit approach), site preparation and rollout. These are all important aspects for the Digital Equipment Operations, which will be discussed next.


[[File:0.2.1 How SCP.png|800px|frameless|center|link=|HOW: Execution and delivery|class=Digital_Equipment_Operator#]]
[[File:0.2.1 How SCP.png|800px|frameless|center|link=|HOW: Execution and delivery|class=Digital_Equipment_Operator#]]

Latest revision as of 05:15, 16 November 2021

More...More...More...More...More...More...Digital OEM

The Digital OEM combines physical product design, engineering and manufacturing with Software-as-a-Service in order to provide smart, connected products. Artificial Intelligence (AI) and the Internet of Things (IoT) are the two key enablers. This section will introduce the concept of the Digital OEM in detail, following again the why, what, how structure from the introduction

WHY

The motivation for adopting a digital OEM business model can vary widely. Many incumbent OEMs are seeking ways to build upon their existing business. New market entrants are looking at disruptive new business models enabled by the combination of physical products with AIoT.

While AI and IoT are exciting technical enablers, anybody embarking on the AIoT journey should always start by looking at the "why": Why do this? What is the purpose? And what are the expected business outcomes? From a strategic (and emotional) point of view, the purpose of the AIoT initiative should be clearly articulated: What is the belief? The mission? Why is this truly done?

Digital OEMs - business models

At the core of the business model of the Digital OEM is the physical asset or product. An interesting question is which new opportunities arise through the combination of physical products with digital solutions. Examples include:

  • Data-driven business, e.g., building on user-generated data or asset/product performance-related data. Examples include usage-based car insurance (UBI), data-driven aftermarket services, or drone-based building facade inspection.
  • Digital add-on services, e.g., an optional autopilot service for an electric vehicle, or cooking recipe add-ons for a smart kitchen appliance
  • Asset-as-a-Service, e.g., car-seat-heating-as-a-service, or the famous "power-by-the-hour" for Rolls-Royce aircraft engines
  • Smart Maintenance, including predictive, preventive and prescriptive maintenance, enabled by deep analytics of asset/machine data via AIoT
WHY: Digital OEM Business Models

The diagram shows key elements of two worlds:

  • The OEM (Original Equipment Manufacturer) is an organization that makes devices from component parts bought from other organizations. This can be a car maker, a manufacturer of household appliances, or a manufacturer of manufacturing equipment, such as robots or laser cutting tools.
  • The suppliers of the OEM are usually referred to as "tier 1", "tier 2", etc., depending on their position in the supply chain
  • On the other side, we have the digital ecosystems. Today, large hyperscalers are dominating cloud-based infrastructure (Infrastructure-as-a-Service, or IaaS) and platforms (Platform-as-a-Service). IaaS includes storage, networking, and virtual compute resources. PaaS includes Internet-based tools and middleware for building applications
  • Software-as-a-Service (or SaaS) are applications delivered over the internet.
  • The digital OEM will combine physical product development with Software-as-a-Service to deliver smart, connected products

Incumbent OEMs - business improvements

Especially for incumbent OEMs, the idea of improving existing business by adding digitally enabled solutions is attractive. Generating ARR (Annual Recurring Revenue) via digital services is very interesting, since ARR is seen as a more stable and predictable revenue stream. However, the opportunity to improve existing business -- and especially EBIT -- with digital solutions as a short-term measure should not be underestimated, since unproven, new business models can have inherent risks and realization of new, ARR-like revenues might take longer than hoped for.

Business Outcomes

WHAT

What can be done with AIoT from the perspective of the digital OEM? Usually, the answer is building smart, connected products. These combine physical products with smartness enabled by AI and connectivity enabled by the IoT. To build smart, connected products, the digital OEM needs to combine product engineering and manufacturing capabilities with edge and cloud software development capabilities.

WHAT: Digital OEM and smart, connected products

Smart, Connected Products - Enabled by AIoT

Smart, connected products usually combine edge and cloud computing capabilities: Edge computing is anything that happens on (or near) the asset/product in the field. Edge computing capabilities are usually dedicated to a single asset/product or sometimes a specific cluster of assets/products operating in close proximity. Cloud computing in an AIoT scenario, on the other hand, can enable insights or functionality that relates to an entire fleet (or "swarm") of assets/products. Consequently, in AIoT, we also differentiate between two types of intelligence: asset/product intelligence vs. swarm intelligence.

Smart, connected products

Example: Robot Vacuum Cleaner

A good example of a smart, connected product is a robot vacuum cleaner. These products use AI to identify room layouts and obstacles and to compute efficient routes and methods. For example, the robot can decide to make a detour vs. switching into the built-in "climb over obstacle" mode. Another example is the automatic activation of a "carpet boost" mode. IoT connectivity to the cloud enables integration with user interface technology such as smart mobile devices or smart home appliances for voice control ("clean under the dining room table").

Example: Robot Vacuum Cleaner

The vacuum robot example will be examined in great detail in the product design section.

Example: Kitchen Appliance

Another good example for smart, connected products is a smart kitchen appliance. Here, the intelligence could start with data gathered from users of kitchen appliances in combination with user-generated ratings. These data could be combined to make targeted recommendations (created via AI), e.g., for cooking recipes. A more advanced version of the smart kitchen appliance could also use AI on the product, e.g., for better device control and maintenance.

Example: Kitchen Appliance

Example: Automatic Wiper Control

In this example, AI utilizes images from the autopilot camera to determine the local weather situation. This is then used to automatically convert the wiper speed to the intensity of rain or snow. This is how Tesla is doing it, and it is an area that is also starting to receive the attention of the research community.

What is interesting about this example is that some Tesla customers initially complained that this was not as accurate as other systems using rain sensors. Over time, Tesla was using their Over-the-Air Update (OTA) capabilities to enhance this function, using continuous model improvements and retraining.

Example: Windshield wiper control

Example: Physical Product Design Improvements

Another interesting use of AIoT is for the advanced analytics of product performance, based on data from assets in the field. For example, the team developing the electric motor for the wiper blades from the previous example could use this approach to better understand how their product performs in the field, e.g., at 150 kph on a highway under heavy rain. This information can then be used to improve the next generation of the motor. In this case, it might sometimes not be clear whether we are talking about advanced analytic or real AI (e.g., using ML), but it is still an important use case.

Example: Physical product improvements

Example: Smart Tightening Tool

Another example is the smart tightening tool (e.g., the Bosch Rexroth Nexo cordless Wi-Fi nutrunner). This is a type of tool used by industrial customers, e.g., for ensuring the quality of safety relevant joints.

On the tightening tool, AI/ML can be used to control the proper execution of tightening programs (controlling torque and angle for specific combinations of materials). In the cloud, data from fleets of tightening tools can be analyzed to help automatically detect tightening anomalies, classify these anomalies, and make recommendations for handling these anomalies.

Example: Smart, connected tightening tool

WHY revisited

Let us revisit the "WHY" perspective with what we have learned thus far about AIoT and the different use cases implemented by Digital OEMs.

Aligning the Product Lifecycle with the Customer Journey

A key feature of AIoT is that it helps align the product lifecycle and the customer journey. In the past, most OEMs lost contact with their products once they left the factory. Although many OEMs try to stay in touch with their customers and support them in the aftermarket, in most cases, the customer relationship was based on service contracts but not a digital relationship. This is of course changing with AIoT, which enables a much higher level of customer intimacy because OEMs can now learn how their products are used in the field and how they are performing. The data obtained and analyzed from the products in the field via AIoT can be augmented with other data, e.g., customer feedback from the Internet.

AIoT also gives the OEM the opportunity to react to what he is learning about his products in the field, by constantly updating existing digital features or even creating new ones, deployed via Over-the-Air Updates (OTA). Naturally, OTA in an AIoT setting will have to support updates of both software and AI models.

WHY revisited: Product LCM and Customer Journey

This topic was recently discussed by Uli Homann of Microsoft at the BCW.on session with Microsoft CEO Satya Nadella and Bosch CEO Volkmar Denner. Full video here.

Uli Homann discusses AIoT Cycle

Uli Homann, Corporate Vice President, Microsoft: The digital feedback loop is essential for successful product development, and OEMs and manufacturers are now also starting to embrace it. For example, we are seeing an increasing number of connected vehicles on the street, which are bringing data into a centralized cloud environment. The cloud is then able to reason over that data and deduce information. Tesla is one of the very famous users of this digital feedback loop already, where they actually use two components. One is the car itself, where it brings information back based upon telemetry, instrumentation, and so forth. So how hard were the brakes being used, if the autopilot is going around the corner? How tight was the corner taken? And then human feedback. Elon is very, very active on Twitter, sometimes very positive, sometimes, to distraction. However, he's very, very active for a very good reason: because he's looking for feedback. One very famous case was people complaining that the Model 3 was taking corners too hard, from their perspective. And so he took the feedback, they compared it with the feedback from the car, and then they made adjustments to the auto drive. And that is truly what we call the digital feedback loop. Because on the one side, you have instrumentation from the car, but you have other channels as well that you bring together and that allow you to start to really think about the lifecycle of the customer journey, the customer buying the car, finding the right car, servicing the car and those kinds of things, and bringing all that data, all of this awareness back into the engineering cycle from design to manufacturing, to the sales and after sales, after market opportunities, etc. Bringing this together in an intelligent way based upon data, utilizing AIoT, is truly the key piece here. The last dimension of making this happen are open platforms, both from an approach to software development as an open ecosystem, with open tools and a lot of open source in the cloud, and also open standards coming together not only in the cloud but also extending this reach into the car. The resulting programming model has platform capabilities underneath that are derived from the Cloud, and optimized for the car. Making this happen consistently will not only allow us to enable AIoT in the cloud but also bringing cloud into the car or into the manufacturing capability. I think the digital feedback loop, the platform tooling and then bringing it into a consistent end-to-end perspective truly will help ensure that we can get digital services at your fingertips. Again, Microsoft is part of an open ecosystem here. We are working together with Bosch and other players to actively bring this to bear, to real life so that we can truly drive this vision forward.

Benefits

The benefits of this approach are manifold, including shorter time-to-market, improved differentiation, improved sales (including recurring revenues), improved customer experience, and consequently improved customer loyalty.

WHY: Benefits

HOW

Now let us take a closer look at how the Digital OEM must go about implementing this with AIoT. This will include a discussion of key design decisions, technical constraints, and considerations for execution and delivery.

Key Design Decisions

From the product manager's perspective, a key question in the future will be -- for each feature -- whether this feature should be implemented in hardware, software, or AI/data, or combinations thereof. Implementing a feature in hardware (including HMI, processing, etc.) will have an impact on usability (for example, sometimes it will still be preferable to activate a feature via a physical control) but also on engineering and design complexity. Implementing the same feature completely in software (e.g., as a feature activated via a smart app) can often mean a lower cost of delivery (no manufacturing/supply costs beyond the initial development) and also means that the feature can be updated via OTA in the future. Finally, if the feature can be implemented virtually, then the next big question is whether it should be implemented as a set of hard-coded rules (software development) or as a data-centric AI function that uses inference to make a decision based on its training.

The decision to use AI, Software, or Hardware for a specific feature will have two main implications: first, the quality of the User Experience (UX), and second, the required technology pipeline to deliver the feature.

HOW: Key design decisions

Considerations for Execution and Delivery

For the digital OEM, execution and delivery will require a holistic view, including business model, leadership & organization, sourcing and co-creation, User Experience (UX) and Human/Machine Interfaces (HMI), data strategy, AIoT architecture, DevOps, Digital Trust and Security, Quality Management, Compliance and Legal, Productization and Sales; and how AIoT will impact them.

What is usually less relevant for the Digital OEM are aspects such as retrofit (assuming the approach here will be a line-fit approach), site preparation and rollout. These are all important aspects for the Digital Equipment Operations, which will be discussed next.

HOW: Execution and delivery