Digital Transformation

Thomas Siebel, 2019

Transform or die
Waves of change
- Digitalization (using computers at all, up to the 80s)
- The Internet (90s – today)
Today’s digital transformation is the next wave
- Cloud Computing
- Big Data
- AI
- IoT
Change needs to be in a unified, holistic manner, recommends Centers of Excellence
It’s a big deal, don’t get fucked

This chapter is an overview of the four main topics: the cloud, data, AI, and IoT
Ten core requirements for success:
- Data aggregation
- Multi-Cloud Computing
- Edge Computing
- Platform Services
- Enterprise Semantic Model
- Enterprise Microservices
- Enterprise Data Security
- System Simulation Using AI and Dynamic Optimization Algorithms
- Open Platform
- Common Platform for Collaborative Development

Chapters 4 – 7 go into details for each of the main topics
Cloud Features:
- Infinite capacity
- On-demand self-service
- Broad network access
- Resource pooling
- Rapid elasticity
Different cloud platforms have different benefits, so be able to interface with all of them (Google, AWS, Azure)

History, how cloud computing helps
Size, Speed, and Shape:
- Size – amount of data is growing
- Speed – data being created more frequently
- Shape – many more forms (video, telemetry, voice, etc.)
Challenges
- Multiple systems
- Incorporating and contextualizing high-frequency data
- Working with data lakes
- Ensuring consistency, referential integrity, and continuous downstream use
- Enabling new tools and skills for new needs
  - “Enterprise IT and analytics teams need to provide tools that enable employees with different levels of data science proficiency to work with large data sets and perform predictive analytics using a unified data image.”

Why now:
- Moore’s Law (raw computing power and price)
- Internet and the cloud, more data being shared and saved
- Statistics and understanding of data science
- Open source movement, use of Python language

AI Today
- Machine Learning
- Optimization
- Logic
Machine Learning Workflow
- Data Assembly and Preparation
- Feature Engineering
- Labeling the Outcomes
- Setting Up the Training Data
- Choosing and Training and Algorithm
- Deploying Algorithm to Production [this is where we’ve struggled]
- Closed-Loop Continuous Improvement
AI translators are needed, not just data science teams, to help the business understand and use AI

Sensors or computing devices everywhere
Think of it doing for physical objects what the internet has done for information: searchable, sortable, remotely controllable, 24/7, generating data and metadata, everything becomes a software problem/feature

Use cases
- Smart electric grid
- Predictive Maintenance
- Inventory optimization
- Patient care
Impact
- Transforming the value chain
- Redefining industry boundaries
- Changing business models

ENGIE case study on transformation
- CEO vision
- established centers of excellence to collaborate with business unit leaders, define requirements, create roadmaps, develop and deploy
Enel case study: smart grid
Catepillar case study: enterprise data hub
John Deere case study: supply chain
3M case study: operational efficiency
US Airforce case study: predictive maintenance

How to aggregate and use all this data?
Don’t do it all yourself, the real game to play is building the connections between all the tools that exist

It’s bad because:
- It’s too complex to use well
- It’s brittle
- It’s not future proof
- Data integration is super hard
The cloud is problematic too, or rather using a single cloud

His recommendation: Model-Driven Architecture
- Start at the model you want to develop and work backwards:
  - Model
  - Tools needed to build the model
  - Data needed for those tools