The document discusses intelligent CPaaS (communications platform as a service) and the role of AI/ML capabilities. It notes key factors for CPaaS success include rich capabilities, fast development time, developer friendliness, cloud infrastructure, and pay as you go pricing. It describes new developments like serverless functions and visual builders. It outlines many potential uses of AI/ML in CPaaS for capabilities like speech recognition, call automation, agent assistance, and call analytics. Challenges in implementing AI/ML include resource needs and data sharing concerns. Emerging technologies like 5G and AV1 may further enable new CPaaS uses.
2. Voximplant 2019
CPaaS success factors
Rich set of capabilities (for voice, sms, video, etc)
Fast time to market (days-weeks-months)
Developer friendly (programmable)
Cloud infrastructure (global availability)
Pay as you go pricing
IDC: “CPaaS will
continue to grow at
strong CAGR of 39.2%
from 2017 to 2022”
3. Voximplant 2019
What’s new - Serverless
Functions
Call control is still done with some
ML (markup language), but it’s
hosted somewhere on the platform
Cloud App Engine
Call control is done with some app
written using some programming
language, the app is running inside an
cloud app engine
4. Voximplant 2019
What’s new - Serverless
Functions
Benefits:
● less or no infrastructure is required
● faster development
● graceful failover
Example: Twilio
Cloud App Engine
Benefits:
● same as with Functions
● real-time call control and debugging
● increased flexibility
Example: Voximplant
6. Voximplant 2019
What’s new - Сloud IDE
All platforms offer web interface that is used by
developers to control the platform, but there is
plenty of space for improvement
Building and debugging apps using Cloud IDE
speed up and simplify the development
7. Voximplant 2019
AI/ML in CPaaS
Speech Recognition/Synthesis - modern TTS/STT solutions (e.g. Google Speech API) are
powered by AI/ML
Automation (NLU/NLP) - connectors to Google Dialogflow, IBM Watson, Amazon Lex,
Microsoft LUIS, etc.
Agent Augmentation (CCaaS) - helping agents in real-time with hints and tips
Emotion/Sentiment Analysis - understanding the quality of customer-agent interaction
8. Voximplant 2019
AI/ML in CPaaS
Audio/Video Enhancement - removing background noise, blurring background in video
conferencing, etc.
Video Analysis - what’s happening in the video
Call Analytics - after the call transcription has been created ML can be used to figure out
what happened during the conversation (classification, segmentation, etc.)
Call Quality Analysis - big data related to call quality (packet loss, rtt, etc.) and routing
(carriers, networks, etc.)
9. Voximplant 2019
AI/ML in CPaaS
Other Specific Capabilities
Example: Voicemail Detection
Voximplant implemented internally by utilizing machine learning
Less than 2 seconds is required to detect if it’s a voicemail prompt or a live person
99%+ accuracy
Requires additional training for specific cases (music instead of ring, etc.)
10. Voximplant 2019
AI/ML in CPaaS - Challenges
AI/ML requires significant resources including talented developers and data scientists, it’s
not easy to compete with major cloud players
Workaround: partnership instead of the competition
11. Voximplant 2019
AI/ML in CPaaS - Challenges
Concerns: partner’s control over your costs and need to share user data
Possible workaround: work with more partners offering similar functionality
It’s worth it: gross margin for this services can be much higher compared to telco
12. Voximplant 2019
Other important things for CPaaS
5G - everything eventually will go to data channel, where more CPaaS
capabilities will be accessible
AV1 - royalty-free video codec that can be an enabler for additional use cases,
like streaming and video conferencing