27. The Importance Early Innovation
Pharmacies /
Drugstores
Music Industry Publishing Industry
Lodging / Travel Mobility / Public
Transport
Marketplace
28. Amazon – Innovation Role Model
Online Bookstore
eCommerce
Platform Web Services eBook Reader
1 2 3 4
29. Agenda
1 Importance of early innovation
2 Chances and pitfalls: 3 company types
3 Key success factors
4 Innovation @ Gini
5 Q&A
30. Agenda
1 Importance of early innovation
2 Chances and pitfalls: 3 company types
3 Key success factors
4 Innovation @ Gini
5 Q&A
31. 3 Company types
Large multinational
enterprise
(≈150k employees)
Big international corporation
+ innovation lab
(<30k employees)
Startup
(<100 employees)
41. Agenda
1 Importance of early innovation
2 Chances and pitfalls: 3 company types
3 Key success factors
4 Innovation @ Gini
5 Q&A
42. Agenda
1 Importance of early innovation
2 Chances and pitfalls: 3 company types
3 Key success factors
4 Innovation @ Gini
5 Q&A
43. Key Success Factors
Customer focus – think outside the box!
Innovate when ahead of the game
Dedication – time, resources, commitment
Collaboration & culture
Disrupt yourself
As creative as possible & as efficient as needed
Awareness & communication
KPI-driven
Strategy alignment
44. Agenda
1 Importance of early innovation
2 Chances and pitfalls: 3 company types
3 Key success factors
4 Innovation @ Gini
5 Q&A
45. Agenda
1 Importance of early innovation
2 Chances and pitfalls: 3 company types
3 Key success factors
4 Innovation @ Gini
5 Q&A
47. Experiments
Customer focus – think outside the box!
There are a lot of methods
that help you to break out of
the box:
- Systematic Inventive Thinking (SIT)
- 6 hats
- Newspaper method
- Design Thinking
- …
49. Zoom in: Idea Generation
Idea
- Features
- Product
enhancements
- Small effort
(≤ 2 Sprints)
- No-brainer
- Must haves
Decision:
Product faculty
- Bigger effort /
high invest
- Unclear
business case
- New markets
- New solutions
- New products
- Etc…
Inno Process
by Inno Lab
Consulting
?
Inno LabProduct faculty
Sales Faculty
50. Zoom in: Idea Generation
https://ginis.ideas.aha.io/ideas/new
🡨 Internal ideas form
External ideas form
🡨
(Product Roadmap SW)
Idea Contributor kicks off
Inno Process by filling out this
form or getting in touch with
Tobias to do this together
51. Map Your Business Model Network
Customer A
Customer‘s
Customer
Channel Partner
Competitor A
Competitor B
Competitor C
Customer B
Strategic Partner
& Service
Providers
Customer C
54. Andrey Bachvarov
T O N I G H T ’ S S P E A K E R
Serial Entrepreneur with
passion for innovation
55. Agenda
▪ Why innovate the GRID?
▪ Redesigned Value Chain and Ecosystem
▪ The Risks of Digital Complexity
▪ Grid Use Case example
▪ Role of AI
56. Value chain redesigned
Information
from Smart
Devices
and
Appliances
Power
Generation
Power
Transmission
Power
Distribution
Energy
Retail
Electric
Devices and
Appliances
End-use
Customers
Traditional Electricity Value Chain
Emerging Electricity Value Chain
Power
Generation
Power
Transmission
Power
Distribution
Energy
Retail
Electric
Devices and
Appliances
End-use
Customers
Information Flow
• Two-way power flow
• Overflow of information from smart devices
57. Digital Consumer of the Future
Utility of the future looks more like an internet service provider or a mobile carrier
and includes cloud, social media, mobile apps and data analytics
58. Rate of Digitalization
Sources: Navigant Research, government data
Europe
$11.6 bln in smart grid technology
revenue by 2023
80% smart meter penetration by
2023
4,138 electricity suppliers and
distribution system operators
216 mln households
USA & Canada
$17.5 bln in smart grid technology
revenue by 2023
60% smart meter penetration by
2023
1000+ electricity suppliers and
distribution system operators
138 mln households
Asia & Australia
$30.3 bln in smart grid technology
revenue by 2023
45% smart meter penetration by
2023
514 mln households in China,
Japan, Australia and New Zealand
Utilities are expected to spend $37 bln on customer information systems and analytics between 2014 and 2023
65. 69
Electric Distribution
Companies (EDCs)
have to inspect
millions of
kilometers of power
lines every year and
they lose money
doing it ineffectively
und inefficiently.
1 The current inspection process is increasingly slow and
costly
2 Areas are hard to access and grid elements are not
properly inspected
3 The data is very difficult to analyze because it is on
paper
Problem
66. 1 2 3
We use drones
for technical
inspection
Our drones fly to places
where humans or
vehicles cannot go
We automatically
annotate problems
for preventive
maintenance
Solution
67. Company Purpose
We are data analytics company using
drones and high fidelity, visual, laser
and thermal monitoring to help
utilities, renewables and any other
asset that requires any kind of visual
inspection for its operation and
maintenance.
What do we do?
71. PowerDrone Digital Twin
Image recognition
current accuracy rates
100% for transformer stations
for poles and insulators
for commutation equipment
for different sub-types of insulators (ceramic, glass, silicon)
for different sub-types of metal poles (stretcher, holder, angle)
99%
97,8%
96,6%
90,55%
72. Success stories
- Inspection of two MV power lines nearby
Sofia, Bulgaria
- 35 unplanned shut-downs for 2,5 months
before drone inspection & analytics
- After detecting the problematic spots and
executed preventive repairs – only 5 shut-
downs on both power lines for a period of 2.5
months
- Result: 700% improvement
Achievements
73. Choosing the right Neural Network is
not obvious
• Feedforward NN
• Convolutional NN
• Fully Convolutional NN
• Deconvolutional NN
• GANs
• Deep Learning
• Intuitive
• Binary > Instances
Neural networks for image processing
74. Image Classification and Recognition Processing via
CNNs
• Similar to the human brain
• Specially designed for
Image Processing
• Requires less pre-
processing
• No hand filters
• Learns during training
• Convolution layers –filtering the input image and extracting
specific features such as edges, curves, and colors
• Pooling layers – Improve the detection of unusually placed
objects.
• Normalization layers – boost performance by normalizing the
inputs of previous layer.
• Fully connected layers – neurons have full connections to all
activations in the previous layer
• Feature Extraction
• Classification
Convolutional Neural Network
75. • Adding visual noise kills
most NNs
• Perturbation
• Need for deep learning
Example of adversarial image misclassification. Image credit: OpenAI
Adversarial Images
76. • Helps with Adversarial
Images
• Huge reduction of
human inputs for fake vs
true discrimination
• Ability to train on
scenarios not existing in
the training set
Generative Adversarial Networks