Tech Strategist Meetup Sydney November 2019: https://www.meetup.com/en-AU/tech-strategists-syd/events/266150820/
AI is transforming every aspect of our daily lives and the data landscape is becoming increasing open and transparent. Between the high level academia and low level algorithms, where should the modern business leader start on their AI journey and harness true value from their data?
Contino has formulated a simple 6 step approach to Enterprise AI Strategy that is currently ranked as Featured Snippet in Google search. On the back of our recent successful data and AI transformation case study at Transport NSW, let us share with you our step by step, proven methdology towards enterprise-wide AI adoption
4. AI in Year 2019
General AI? Not quite there yet...
4
Source: still from Blade Runner 1982
5. 5
Demystified Definition of AI
Hot Dog!
COGNITION
DATA
What we do have is Narrow AI which tackles specific problems by inputting data and applying
learnings to produce knowledge and cognition
6. 6
Rapid AI Evolution in the Age of Implementation
2010
2012ish - frameworks
Nowish - autoML
7. Commodised AI and Open Data enables all
industries to create unique Virtuous Cycles and
Defensive Data Moats
7
Better
Product
More Quality
Data
More Happy
Users
V
AWS
13. 13
Enterprise AI challenges are not really about
algorithms or technology.
Coherent Strategy Design &
Operating Model
Execution
The key ingredients of AI adoption are simple:
Customers IT & Business
1 2 3
16. Working backwards from Strategy
Capabilities
Data Engineering,
Analytics and ML
Capabilities on
Cloud
Cloud Native
Development Skills
Agile Product Mode
of Delivery
API and Full Stack
Engineering
SME and Partner
Collaboration
Intelligent Transport Networks,
managed with Data
1. Data Science Incubator and
open data expansion
2. Trial AI to improve network
management and customer
service
3. Integrate predictive
maintenance
Outcome
1. 5 minutes near
real time
2. 80% accuracy
3. Exec + Operational Personas
4. Ambitious time to value
5. Productivity and cost
optimisation = “serverless-first”
Objectives & Goals
Crowdsourced Vision
17. DATA ASSETS
TACTICSOBJECTIVES
Predict Patronage in advance using Weather
and Network Usage data
GOALS
85% accuracy for predictions
over a 24 hours time frame
PEOPLE
Exec Sponsor: ED of
Innovation
ETHICS & LEGAL
1
2
3
4 5
PROCESSES TECH ASSETS7 86 9
MEASURES
AI PRODUCT CANVAS 2019Date:Network Patronage PredictionTopic:
21. Picking the Right Lighthouse MVP
21
LargeLow
Low
Small
Lon
g
Hig
h
Short
Business
Sponsorsh
ip
Duratio
n
Importanc
e
Project
Size
Pick this
project
Data
Qualit
y
22. ETHICS & LEGAL
MEASURES
TACTICS
DATA ASSETS
OBJECTIVES
Predict Patronage in advance using Weather
and Network Usage data
GOALS
85% accuracy for predictions
over a 24 hours time frame
PEOPLE
Exec Sponsor: ED of
Innovation
1
2
3
4 5
PROCESSES TECH ASSETS7 86 9
AI PRODUCT CANVAS 2019Date:Network Patronage PredictionTopic:
Opal
Data
Ingestion
Weather
Data
Ingestion
Ingest
Weather
data per
hour
Opal
ingestion
per 5min
85%
Accuracy
for 48
hours
24. 24
Working back from the Customer
1.Data
Platform
2.Analytics and
visualisation
3.AI/ML
AND implement this way
3 Months not 3
Years
START from here
Business Use
Case
27. MLOps / DataOps
So work backwards towards best practice AI
Architecture
ML Engineers / Data Scientists
Application Engineers Model
Notebook
Training
Job
AppSync Inference
Lambda
De-Identified
Data Lake
Inference
Auth
27
Weather
Events
Time Tables
Static
Web
Warehouse
Opal
Exec
Operations
29. 29
An Operating Model for Enterprise-Wide AI
Innovation, Collaboration and Consumption
Business IT
AI Product 3AI Product 2
Process
Ethics &
Legal
People
3Cs
Customers
Culture
Comms
AI Product 1
Data Assets
Technology Assets
AI Product 4
31. 31
Machine Learning Life Cycle
Product
OGTM
Split Data
Train ModelTest ModelDeploy Model
Monitor and
Validate
AI Problem
Definition
Validation Data Training DataTest DataLive Data
Collect and
Prepare Data
Pretrained
Model / Service
32. DATA ASSETS
ETHICS & LEGAL
• Privacy and PII
• Opt In Opt out
• Diversity & Bias
MEASURES
TACTICSOBJECTIVES
Predict Patronage in advance using Weather
and Network Usage data
GOALS
85% accuracy for predictions
over a 24 hours time frame
PEOPLE
• Exec: ED of Innovation
• Owner: Head of Operations
• Team: Lucius
• Customers: Exec and
Operators
1
2
3
4 5
PROCESSES TECH ASSETS7 86
9
Opal
Data
Ingestion
Weather
Data
Ingestion
Patronage
(DeepAR)
BOM
Data
(HTTP)
Legal & Compliance Review
Training
Client Engagement
Communications Strategy
Sales & Marketing Team
CHANGE MANAGEMENT
Ingest
Weather
data per
hour
Opal
ingestion
per 5min
85%
Accuracy
for 48
hours
AI PRODUCT CANVAS 2019Date:Network Patronage PredictionTopic:
34. Iteratively build up your capabilities
34
$$
$
Value
Time
$$
$$
Phase 1 (3 weeks)
Service Design and UX
Data Platform Foundations
Phase 2 (6-9 weeks)
Minimal Viable Product
ML algorithm tuning and training
User Testing and Metrics Validation
Phase 3
Continuous Innovation
Defensible AI Asset
Continuous Journey