15
Employing Chatbots 30/08/2016

Employing chatbots

Embed Size (px)

Citation preview

Page 1: Employing chatbots

Employing Chatbots30/08/2016

Page 2: Employing chatbots

Employing Chatbots Why the Chatbot explosion?

• Messaging is THE most popular communication channel

• Simultaneous conversations in multiple channels

• Facebook Messenger has over 900 million users!

• EaaS/ IoT

• Advancement of NL and AI technologies

Gihan Munasinghe
I want to talk here what chatbots have become popular nowi.e Messaging has become popular Simultaneous conversations Cost Saving Ways of linking with automation (Everything is as a Service these days, open API’s) NLP and AI have come a long way
Page 3: Employing chatbots

Employing Chatbots What is the role of NLP?

Gihan Munasinghe
Role of NLP.. NLP needs to be flexible, adaptation wise less users need to train better then in-house adaptation will be Supervised Learning is good, You do not need to have all powerful NLP system from day oneNLP is a just the enabler, there bit more than NLP when building a Useful botWe need to add the picture with the bot on Hart
Page 4: Employing chatbots

Automation

User ExperienceTrust

Employee to Bot

Training

Employing Chatbots

Page 5: Employing chatbots

Employing Chatbots

Taming the Bot • NLP has come a long way but its not 100% perfect

• Always have back up, start with NLP will fail

• GIGO (Garbage in Garbage out)

Gihan Munasinghe
NLP is good, its not perfectNLP need lot of training. Don’t be afraid to use the humans .. interactions should be naturalShould put the picture of bot and human touching hands one in the blog
Page 6: Employing chatbots

Employing Chatbots Making The Bot Productive

• User Experience - should the bot behave separately to a human handling the same issue or query

• Linking with other services

• Context Awareness

• Multi channel, same bot should be able to handle different messaging platforms

Page 7: Employing chatbots

Employing Chatbots

A bot to try & do everything or A team of specialised bots

Employing a Bot?

Page 8: Employing chatbots

Customer Service User Case

Bob (accounting)Steve(support)

Brenda (sales)

Customer

Employing Chatbots

Page 9: Employing chatbots

Know it all TREVOR

Employing Chatbots

Page 10: Employing chatbots

Easy to understand

Easy to train

You can model this on a single business unit

Smoother Bot - Human Hybrid model

Easily adoptable

Trial and error without affecting the rest of the processes

Defined clear boundaries

Single Purpose BotsEmploying Chatbots

Page 11: Employing chatbots

You have only one big bot to manage

One place of interaction

User experience will not vary

Same with single purpose model you start with one use case

Gradually increase bots responsibility

Multi Purpose BotEmploying Chatbots

Page 12: Employing chatbots

How to get started:

Start simple - try Hello World! Choose the problem you want to solve Find a good NLP base Monitor usage between customers and employees Hook in to current workflows

Employing Chatbots

Page 13: Employing chatbots

What is the Future ?

Employing Chatbots

More emphasis on sentiment

Semi supervised learning

Bot to Bot communicate

Bot discovery

Ontology specifications for bots

Gihan Munasinghe
Sentiment Analysis Faster learning cycles Bot taking bots over NLP see there is no difference
Page 14: Employing chatbots

Employing Chatbots

Page 15: Employing chatbots

Converse.ai, Inc.548 Market Street

San Francisco, CA 94104+1 (415) 919-7891 | [email protected]

Thank You