AI is transforming seemingly all kinds of human activities, and software development is the next frontier. OutSystems is leading this transformation, focused on augmenting developers with built-in, intelligent assistance that makes you more productive and doing less grunt work. Learn how in this session!
A Comprehensive Look at Generative AI in Retail App Testing.pdf
How to Be More Productive With AI Development Assistance
1. | Change session name in master slide
How to be more productive with
AI-Assisted Development
NextStep 2019
2. | Change session name in master slide| Change session name in master slide
@ bhavin.patel@outsystems.com
Bhavin Patel
Sr. Solution Architect | OutSystems
@ antonio.alegria@outsystems.co
m
Antonio Alegria
Head of AI | Company
@antonioalegria
3. | Change session name in master slide| Change session name in master slide
OutSystems
One Core Goal
4. | Change session name in master slide| Change session name in master slide
Extreme Agility
without Limits
5. | Change session name in master slide| Change session name in master slide
Explorer
Novice
Professional
Master
6. | Change session name in master slide| Change session name in master slide
What if we could accelerate Pros and
help beginners learn while doing?
7. | Change session name in master slide| Change session name in master slide
Explorer
Novice
Professional
Master
Guide Accelerate
8. | Change session name in master slide| Change session name in master slide
40%
of new projects will have AI Co-Developers in the Team, by 2022
Gartner
9. | Change session name in master slide| Change session name in master slide
Workflow Builder
Architecture
Dashboard
AI Infusion across the
Software Development Lifecycle
Plan
Design
Prototype
Develop
Assure
Quality
Deploy
Release
Monitor Operate
Service Studio
10. | Change session name in master slide| Change session name in master slide
AI-Assisted Development
presenting
11. | Change session name in master slide| Change session name in master slide
● Inline next step suggestions in
any flow, anywhere in the flow
● Between 1 and 6 suggestions
● Suggestions specific to your
business logic
● Attempts to auto-fill node
properties
● Models are evolving every
week
12. | Change session name in master slide| Change session name in master slide
Lets see it in action!
Demo
13. | Change session name in master slide| Change session name in master slide
How do we do it?
Anonymized
Flow Data
1
X-Ray
Deep Code Analysis &
Normalization
2
14. | Change session name in master slide| Change session name in master slide
GetProducts
Check If GetProducts
is Empty
ExcelToRecordList
For Each
ExcelToRecordList
CreateProduct
ProductRecord
Set ProductRecord
X-Ray
Deep Code Analysis
& Normalization
15. | Change session name in master slide| Change session name in master slide
Get List of Entity X
Check if A is Empty
ExcelToRecordList
of X
For Each X in B
Create X with D
Set D
D
READ
READ
READ
WRITE
WRITE
WRITE
WRITE
X-Ray
Deep Code Analysis
& Normalization
Aggregate
Output:
RecordList of
X
RecordList of
X
Local Var
Record of X
Record of X
A
B
C
D
16. | Change session name in master slide| Change session name in master slide
How do we do it?
Anonymized
Flow Data
1
X-Ray
Deep Code Analysis &
Normalization
2
Code DNA
Graph Neural Code
Understanding
3
17. | Change session name in master slide| Change session name in master slide
● Learns patterns and micro-
patterns from code examples
● State-of-the-Art Graph Neural
Networks
● Model ensemble: multiple Deep
Learning models for more
accurate predictions
● Core technology underpinning
other AI infusion
Code DNA
Graph Neural Code Understanding
18. | Change session name in master slide| Change session name in master slide
...
71%
15%
3%
0%
...
Prediction of
Next Node
X-Ray
Deep Code Analysis &
Normalization
Anonymized
Flow Data
How do we do it?
Code DNA
Graph Neural Code
Understanding
19. | Change session name in master slide| Change session name in master slide
How do we do it?
Continuous improvements learned, every week
Anonymized
Flow Data
1
X-Ray
Deep Code Analysis &
Normalization
2
Code DNA
Graph Neural Code
Understanding
3
Assistant Usage Data &
Feedback Loop
4
20. | Change session name in master slide| Change session name in master slide
65% Accuracy in Top-1 Suggestion
When confident a single super
accurate suggestion
Suggests CRUD Wrappers
instead of default CRUDs
Keyboard support
Suggest main entity for
Aggregates
Auto-fill properties
Inline suggestions anywhere in the
flow
If & For-Each Suggestions
AI-Assisted
Development
Highlight high-confidence suggestions
1 to 6 suggestions
Suggest common Actions
Suggest the right Entity Actions
Drag connector for suggestions
Works in any flow kind
Automatic binding in
assigns
90% Accuracy in Top-3 suggestions
21. | Change session name in master slide| Change session name in master slide
Available today in Service Studio beta
GA in November 2019
22. | Change session name in master slide| Change session name in master slide
Data Mapping
What’s coming?
Pattern Suggestions
Infusion in other
OutSystems Tools
Beginner Guidance
We learn and improve by going through multiple projects and scenarios and integrating patterns on how to deal with different programming challenges.
Ever since we started programming, through different programming languages until we use OutSystems.
Every experience is reintegrated with more and more pattern matching.
What’s New?
Market-leading prediction accuracy
(First suggestion is right ⅔ of the time)
Specific suggestions for Aggregates, Actions & more!
(e.g. Get Customer, CreateAndUpdate Order, List Append, etc...)
Automatic binding of properties in some cases(e.g. Action params, Aggregate source and filters, Assign vars, etc...)
New sophisticated & streamlined UI
(more beautiful, easier to read, without affecting your regular flow)
Keyboard Enabled!
(more beautiful, easier to read, without affecting your regular flow)
In order to power our current AI Assistant we leverage anonymous Service Studio usage data on flows and development patterns. No personal data or application-level data is collected (e.g. data that is produced and consumed by customers applications).
Furthermore, the usage of the assistant will also provide us feedback data to continuously improve the Machine Learning algorithms and the resulting predictions.
What’s New?
Market-leading prediction accuracy
(First suggestion is right ⅔ of the time)
Specific suggestions for Aggregates, Actions & more!
(e.g. Get Customer, CreateAndUpdate Order, List Append, etc...)
Automatic binding of properties in some cases(e.g. Action params, Aggregate source and filters, Assign vars, etc...)
New sophisticated & streamlined UI
(more beautiful, easier to read, without affecting your regular flow)
Keyboard Enabled!
(more beautiful, easier to read, without affecting your regular flow)
What’s coming down the pipeline?
Further improved accuracy, specific suggestions & automatic binding for more cases!
Proactive Assistance(when it has high confidence, show user what she’s missing)
Novice Assistance
(explain suggestions & use Assistant to inject enablement for key friction points)
Pattern Suggestions
(go from next step to next pattern)
Also, these base models are powering some new capabilities used for architecture analysis, refactoring and pattern discovery.
Data Matching