Chatbots Market Study - Daybreak Insights

Chatbots are widely adopted across different industries and for various use cases due to their ability to engage leads and serve customers 24/7. With the advent of generative AI, the chatbot development process is undergoing a massive transformation.

We have leveraged our years of research and hands-on product trials of over 400 chatbot companies to distill the implications of generative AI to the industry in this market study. This report presents the full set of generative AI features seen across enterprise chatbots, as well as their benefits to different chatbot user groups. We have also selected some notable chatbot companies and scored their generative AI capabilities.

1. Market Study
Implications of Generative AI
in Enterprise Chatbots
June2024

2. TABLE OF CONTENTS
2
Limitations………………………...…..………….………..….………
Key Takeaways………………………...…..….…..…..….………
Glossary of Terms…………………………...………...….………
About Daybreak Insights……………….……...….………
Industry Background & Introduction……………….…….…..
Chatbot Building Process………………………………………………
Gen AI Benefits……………………..….…………………………………….
Gen AI Feature Descriptions………..……….…………...….………
I.Chatbot Builders
II.Chatbot Testers
III.End Customers
IV.Live Agents
Notable Companies………………………...…..……………...….………
I.Cognigy
II.Kore.ai
III.Voiceflow
3
4
5
6
12
18
19
20
21

3. INDUSTRY BACKGROUND & INTRODUCTION
3
Chatbots are widely
adopted for various use
cases such as customer
service, HR, sales, and
marketing–and across
different industries
including banking, retail,
healthcare, and
manufacturing.
They greatly assist the
enterprise in providing
autonomous agents that
can engage leads and
resolve customer issues
24/7.
Daybreak Insights has
conducted years of
research and hands-on
product trials of the
chatbot landscape of 400+
companies, gaining in-
depth industry
knowledge.
This market study examines the Generative AI* features of chatbot companies
and highlights the implications to the industry.

4. 4
CHATBOT BUILDING PROCESS
Before the advent of generative AI, chatbot builders
created chatbots by establishing conversation flows on a
canvas that included nodes that were linked using logic
connections.
For each node, chatbot builders would then create intents*
and utterances*; intents being actions that the chatbot
would perform to fulfill a customer’s request and
utterances being example customer sentences that would
trigger the intent if matched with a customer input.
After building the chatbot flow, chatbot testers generally
ran a wide range of tests on potential customer inputs
against the chatbot to identify and fix potential issues.
With generative AI, all these processes are
undergoing transformations.
2
3
1

5. 5
GEN AI BENEFITS
Generative AI offers benefits across four chatbot user groups.
User Group Benefits Generative AI Features
Chatbot
Builders
Simplified and accelerated chatbot development
process
Utterance Generation
Entity Generation
Flow Generation
Knowledge Extraction
Zero-Shot/Few-Shot Modeling
Chatbot TestersSimplified and accelerated chatbot testing
experience
Testing Data Generation
End CustomersMore natural and personalized conversations with
the chatbot
Conversation Generation
Output Rephrasing
Live Agents Streamlined live agent experience Conversation Summary
Generation

6. GEN AI FEATURE DESCRIPTIONS -CHATBOT BUILDERS
6
Chatbot companies integrate with generative AI to offer the following features that reduce
the time and effort needed in the development process.
Utterance Generation
The training utterances for the various intents can be
automatically generated. The chatbot builder just needs to
provide the intent name and description for generative AI to
suggest relevant utterances.
Entity* Generation
Generative AI can automatically generate entities; the
chatbot builder just needs to provide the entity name and
description for the relevant entities to be suggested.
•For example, the entity name “World airport
codes”would generate suggestions such as
“SFO,” “LAX,” “LHR,” “SYD,”etc.

7. GEN AI FEATURE DESCRIPTIONS -CHATBOT BUILDERS
7
Flow Generation
Generative AI can generate entire chatbot flows
complete with all the necessary nodes and logic. The
chatbot builder only needs to input a name for the
chatbot flow, and a description or a sample transcript.
Knowledge Extraction
For informational chatbots, generative AI can
automatically parse documents such as CSVs, PDFs,
or webpages and use their parsed contents to answer
relevant customer questions.
Zero-Shot/Few-Shot Modeling
Generative AI enables customer inputs to be matched
to the appropriate intents by their semantic similarity
to the intent names alone, and requires the chatbot
builder to create few or no utterances at all.

8. GEN AI FEATURE DESCRIPTIONS -CHATBOT TESTERS
8
After having built the chatbot, the chatbot testers would need to test the chatbot and
generative AI can also help facilitate this process.
Testing Data Generation
Generative AI can automatically generate potential
customer inputs and simulated conversations to test if the
inputs match with the expected intents and if all the flows
execute correctly.
•For example, for the intent “Book flight,”
potential customer inputs that could be
generated include “Can you get me a flight from
Miami to New York?” “I’d like 2 economy seats
from Lisbon to Tokyo tomorrow at noon,” and
“Can I book United Premium Economy from LAX
to JFK please?”

9. GEN AI FEATURE DESCRIPTIONS -END CUSTOMERS
9
Generative AI can independently answer customer questions or enhance the chatbot’s
conversations by making them more natural and humanlike.
Conversation Generation
The chatbot builder can insert a conversation generation
node in the chatbot’s conversational flow, define the
chatbot’s task, customize its personality, and then let the
chatbot freely and automatically generate dialogues with
the customer.
•For example, the chatbot task can be
“Welcome the customer to the bank’s
website and also give them a piece of
financial fraud prevention advice.” and its
personality can be set as “Be a cheerful and
professional banking assistant.”

10. GEN AI FEATURE DESCRIPTIONS -END CUSTOMERS
10
Generative AI can independently answer customer questions or enhance the chatbot’s
conversations by making them more natural and humanlike.
Output Rephrasing
For the responses that the chatbot delivers, generative
AI can make them more empathetic and contextual by
rephrasing and responding with the right emotions to
the customer’s inputs.
•For example, the customer says “I just got into
a car accident and need to file an insurance
claim.” Instead of responding robotically to the
input with “Please provide me with insurance
number,” the chatbot can now deliver a more
empathetic response such as “I’m so sorry to
hear about the accident! Can you please
provide me with your insurance number so
that I can help you to file an insurance claim?”

11. 11
When the customer is transferred to a live agent, the agent will need to read the customer’s interaction
history with the chatbot to better serve them. Generative AI can expedite this handoff process.
Conversation Summary
This feature can generate a concise and insightful
summary of the conversation between the chatbot and the
customer.
•An example conversation summary can be
“Joseph wants to change the date of his flight to
November 15, 2024. When searched through
the flight reservation database, no economy
seats were found on Joseph’s desired date.
Joseph asked to be transferred to a live agent.”
GEN AI FEATURE DESCRIPTIONS -LIVE AGENTS

12. Generative AI Feature
Vendor
Score
Rationale
Utterance Generation Generates utterances from the intent name and description.
Entity Generation Generates entities from the entity name and description.
Flow Generation Generates chatbot flows from name, description, or transcript.
Knowledge Extraction Parses documents and webpages to answer customer questions.
Zero-Shot/Few-Shot Modeling Enables its NLU to be trained using fewer utterances.
Testing Data Generation Does not generate testing data.
Conversation Generation Automatically conducts full conversations with the customer.
Output Rephrasing Rephrases the outputs based on previous customer inputs.
Conversation Summary Generation Generates conversation summaries for live agent handoff.
Advanced Effective Limited Unavailable
12
NOTABLE COMPANIES -COGNIGY

13. NOTABLE COMPANIES -COGNIGY
13
•Cognigy enables the integration of their chatbots
with generative AI to benefit the chatbot builders,
end customers, and live agents.
•The chatbot builder can integrate with a variety of
generative AI providers, including OpenAI, Azure
OpenAI, Anthropic Claude, Google Vertex AI, and
Aleph Alpha.
•Cognigy’s knowledge extraction feature parses
the chatbot builder’s uploaded content into
“chunks” of information, which can range from a
sentence to a paragraph. The chatbot builder can
view and modify these chunks in a Chunk Editor in
order to adjust the chatbot’s responses to
potential customer questions.

14. Generative AI Feature
Vendor
Score
Rationale
Utterance Generation Generates utterances from the intent name and description.
Entity Generation Can sometimes generate entities when generating utterances.
Flow Generation Generates chatbot flows from the flow name and description.
Knowledge Extraction Parses PDF documents to answer customer questions.
Zero-Shot/Few-Shot Modeling Detects intents using similarity between input and intent name.
Testing Data Generation Generates potential customer inputs to test intent detection.
Conversation Generation Creates limited conversations to collect entities from customer.
Output Rephrasing Rephrases the outputs based on previous customer inputs.
Conversation Summary Generation Generates conversation summaries for live agent handoff.
Advanced Effective Limited Unavailable
14
NOTABLE COMPANIES -KORE.AI

15. NOTABLE COMPANIES -KORE.AI
15
•Kore.ai enables the integration of their chatbots
with generative AI to benefit the chatbot builders,
chatbot testers, end customers, and live agents.
•The chatbot builder can integrate with OpenAI,
Azure OpenAI, Anthropic, Kore.ai’s proprietary
GPT, and any other custom LLM that they choose.
•For the Testing Data Generation feature, Kore.ai
can automatically generate potential customer
inputs to test if they will trigger the relevant
intents. It can also provide customer input
suggestions when creating simulated conversations
to test the chatbot’s flow executions.

16. Generative AI Feature
Vendor
Score
Rationale
Utterance Generation Generates utterances from the intent name and description.
Entity Generation Generates entities from the entity name and description.
Flow Generation Does not generate chatbot flows.
Knowledge Extraction Parses documents and webpages to answer customer questions.
Zero-Shot/Few-Shot Modeling Does not offer zero-shot/few-shot modeling capabilities.
Testing Data Generation Does not generate testing data.
Conversation Generation Automatically conducts full conversations with the customer.
Output Rephrasing Rephrases the outputs based on previous customer inputs.
Conversation Summary Generation Does not generate conversation summaries.
Advanced Effective Limited Unavailable
16
NOTABLE COMPANIES -VOICEFLOW

17. NOTABLE COMPANIES -VOICEFLOW
17
•Voiceflow enables the integration of their chatbots
with generative AI providers OpenAI and Anthropic to
benefit the chatbot builders and the end customers.
•For Voiceflow’s Knowledge Extraction and
Conversation Generation features, it provides a
Preview function which allows the chatbot builder to
test the feature with questions and view the chatbot’s
responses from different generative AI providers.
Based on the quality of the chatbot’s response, they
can select the optimal model and adjust the feature
settings accordingly.
•Voiceflow’s Conversation Memory feature enables the
chatbot to rephrase its responses using the context of
the previous 10 exchanges in the conversation.

18. LIMITATIONS
Despite the benefits of generative AI, there are some limitations to its capabilities.
Zero-shot modeling does not allow the chatbot builder to tweak the model if the chatbot’s responses are not what
they’re looking for. Therefore this feature should be used when the chatbot only has a limited number of intents
with descriptive intent names.
Generative AI also has the tendency to hallucinate and invent answers if it cannot find the requested information
in its data sources. Hence the chatbot builder must take care to set appropriate AI guardrails.
When integrating with a third-party generative AI provider, there might be data privacy concerns. The customer
should be notified that their data will be shared with these providers before using the chatbot, and be given the
option to delete their data from these providers after using the chatbot.
18

19. KEY TAKEAWAYS
19
✔Generative AI is rapidly transforming the
enterprise chatbot
•Simplifying and accelerating the chatbot building
and testing process.
•Making the chatbot’s conversations more natural
and personalized.
•Streamlining the live agent experience by providing
conversation summaries.
✔We expect more progress and innovation from
generative AI in the chatbot industry, especially with
regards to the ability of customizing the chatbot for
specific industry use cases and making their
conversations indistinguishable from real human agents.
✔While generative AI has made some headway
in the chatbot industry, it has not fully
automated the process and also comes with
some limitations.

20. 20
GLOSSARY OF TERMS
•Entity
Keywords that the chatbot builder has defined for the chatbot to extract from customer inputs in order to help fulfill their requests.
Examples of entities include airport codes, cities, and product names.
•Generative AI
A specific type of artificial intelligence that is capable of generating new and original content, such as texts, images, videos, and
computer code.
•Intent
The goal or objective that the customer is communicating through their input. The intent is recognized by the chatbot in order to help
the customer fulfill their requests in the conversation flow.
•Large Language Model (LLM)
A type of artificial intelligence model that is trained on vast amounts of data and thus can generate new content by replicatingthe
patterns it learned from its trained data.
•Natural Language Understanding (NLU)
A type of artificial intelligence that is capable of understanding and processing human language. Chatbots can utilize natural language
understanding to recognize the intents, entities, and sentiments in customer inputs.
•Token
A single unit of text that the chatbot segments the customer input or its own output into in order to process it. It has a variable length
and can range from one character to an entire word.
•Utterance
A sentence that the chatbot builder has predefined as an example customer input that expresses their intent to complete a
specific objective. For example, for the intent of booking a table at a restaurant, utterances can be “I’d like to book a table for
2 at 9.” or “Do you have a table for 2 people at 9 o’clock?”

21. ABOUT DAYBREAK INSIGHTS
21
Our custom research empowers you to understand the latest technology trends, identify
possible vendors, or scout for innovative startups.
Know the companies that comprise
any technology category and stay
up-to-date on the market trends
Identify Companies and Trends
Contact [email protected] get the insights and the data you need.
Explore the product features and
technologies for a category and see
how companies score and rank
Analyze Key Product Features
Understand the differentiators of
each company and how they grade
across each capability
Examine Company Profiles

22. INDUSTRIES COVERED
22
We have professional expertise in the following emerging technology industries.
3D
Printing
Health
Technology
Artificial
Intelligence
HR
Technology
Blockchain
Technology
Insurance
Technology
Energy
Technology
Internet of
Things
Financial
Technology
Marketing
Technology
Real Estate
Technology
Video
Technology
Regulatory
Technology
Virtual
Reality
Retail
Technology
Your Custom
Industry
Security
Technology
Transportation
Technology