Today, in order to have two machines talking to each other, one need to use a lot of human effort. Engineers are required to discover services and APIs, read their docs, write the integration code and maintain the integration as software evolves. These are the most boring engineering tasks that are delegated to cohorts of support engineers.
And that’s just about the engineering, whereas people also required to sign contracts and terms of service and pay bills. Common situation is that the upfront labour costs to integrate surpass the running costs of using the API for a couple of years.
Moreover, for many cases, such as AI services like machine translation, you need frequently swap service providers to have a cutting edge of technology. With the manual peer-to-peer API integration that’s just economically infeasible.
2. It’s not about SkyNet
It’s about business
and should be solved
Data Science Weekend, Moscow, 2017
3. New Digital Economy
MAIN FACTORS
cheaper information search
cheaper communication
MAIN OUTCOMES
more scale
digital transformation
more room in value chains
1995-…
Data Science Weekend, Moscow, 2017
4. New AI Economy
2016-…
MAIN FACTOR
cheaper decision making
MEANING
data collection, storage and processing
risk management and prediction
decision delegation,
implementation and monitoringData Science Weekend, Moscow, 2017
5. New AI Economy
2016-…
EFFICIENCY
for business built around decisions:
mortgage, insurance etc
AI TRANSFORMATION
business processes rebuilt around
decision-making automation
NEW MARKETS
more room in value chains
Data Science Weekend, Moscow, 2017
6. Decision-making in
XX
few trusted information sources
in-house analysts and experts
few trusted partners on pre-paid contracts
planning everything ahead
Data Science Weekend, Moscow, 2017
7. Decision-making in
XXIreal-time information discovery
outsourced analysts and experts
on-demand services with instant gratification
in situ decision making
COOPERATION
more
better
faster
cheaper
Data Science Weekend, Moscow, 2017
10. How many humans does it take
to make two programs talk?
Client App Service
Provider
Discover
API
Read
API docs
Write
integration
code
Maintain
integration
$394Bn
System Integration
Market*
$35Bn
Custom Software
Integration Market*
*2017 projection by Gartner Data Science Weekend, Moscow, 2017
11. How many humans does it take
to make two programs talk?
Client App Service
Provider
Discover
API
Read
API docs
Write
integration
code
Maintain
integration
+ legal + deals
Data Science Weekend, Moscow, 2017
12. Typical solutions
Custom integrations in private cloud
IPaaS (cloud integration for public services)
Crowdsourcing (service platforms)
are based on
manual directories
Unlikely to work for “50Bn connected devices by 2020”
Data Science Weekend, Moscow, 2017
(to avoid custom p2p integrations)
13. Like it was in early web
Personal
bookmarks
Curated
directories
Automated
search
but…
Data Science Weekend, Moscow, 2017
14. APIs are not documents,
delivery means not
reading,
but interacting
HERE COMES AI!
Data Science Weekend, Moscow, 2017
15. Two interesting projects
Viv Labs, acquired by Samsung in 2016: synthesises
middleware code for third-party services integrated to
the platform
Microsoft/Cambridge DeepCoder: combines code
samples from other programs to reach its goal
Data Science Weekend, Moscow, 2017
16. Two interesting projects
Viv Labs, acquired by Samsung in 2016: synthesises
middleware code for third-party services integrated to
the platform
Microsoft/Cambridge DeepCoder: combines code
samples from other programs to reach its goal
Intento: API Integration Platform, powered by
“artificial engineers”
THREE!
Data Science Weekend, Moscow, 2017
17. INTENT
Client
software
translate
CONTEXT
text
to
Intento Service Platform accepts requests with an intent and its context, routes
them to the appropriate API, receives the answer and translates it back to the
intent domain.
Service
providers
Microsoft Cognitive
Services
IBM Watson
Google Cloud Services
…
Indie providers
System
of Intents
Routing &
Integration
Billing
Data Science Weekend, Moscow, 2017
18. Focused on AI Services
machine translation
image recognition
voice synthesis
OCR
et ceteraWHY?
multiple providers per intent
arms race between service providers
context-based relevance
(currenly)
20. What it has to do with
Machine to Machine
Communication?
powered by “artificial engineers”, not manual
ideally, any new public service is integrated to the
platform in a matter of seconds, much like Google
does to websites
“standard library” of intents may be used as a lingua
franca for “I need to to do…” between machines
21. Please, Oh Please!
If you are interested in either
using the Intento Platform or
submitting your APIs, fill out
the form at
https://inten.to/api-platform