Motivations, Classification and Model Trial of Conversational Agents for
Insurance Companies
Falko Koetter
1
, Matthias Blohm
1
, Monika Kochanowski
1
, Joscha Goetzer
2
, Daniel Graziotin
2
and Stefan Wagner
2
1
Fraunhofer Institute for Industrial Engineering, Nobelstr. 12, 70569 Stuttgart, Germany
2
University of Stuttgart, Universit
¨
atsstr. 38, 70569 Stuttgart, Germany
Keywords:
Conversational Agents, Intelligent User Interfaces, Machine Learning, Nlp, Chatbots, Insurance.
Abstract:
Advances in artificial intelligence have renewed interest in conversational agents. So-called chatbots have
reached maturity for industrial applications. German insurance companies are interested in improving their
customer service and digitizing their business processes. In this work we investigate the potential use of con-
versational agents in insurance companies by determining which classes of agents are of interest to insurance
companies, finding relevant use cases and requirements, and developing a prototype for an exemplary insur-
ance scenario. Based on this approach, we derive key findings for conversational agent implementation in
insurance companies.
1 INTRODUCTION
With the digital transformation changing usage pat-
terns and consumer expectations, many industries
need to adapt to new realities. The insurance sector
is next in line to grapple with the risks and opportuni-
ties of emerging technologies, in particular Artificial
Intelligence (Nordman et al., 2017).
Fraunhofer IAO as an applied research institution
supports digital transformation processes in an ongo-
ing project with multiple insurance companies (Ren-
ner and Kochanowski, 2018). The goal of this project
is to scout new technologies, investigate them, rate
their relevance and evaluate them (e.g. in a model trial
or by implementing a prototype). While insurance
has traditionally been an industry with very low cus-
tomer engagement, insurers now face a young gener-
ation of consumers with changing attitudes regarding
insurance products and services (Pohl et al., 2017).
Traditionally, customer engagement uses channels
like mail, telephone and local agents. In 2016, chat-
bots emerged as a new trend (Guzmn and Patha-
nia, 2016), making it a topic of interest for Fraun-
hofer IAO and insurance companies.
With the rise of the smartphone, many insurers
started offering apps, but success was limited (Power,
J. D., 2017), which may stem from app fatigue (Schip-
pers, 2016). App use has plateaued, as users have too
many apps and are reluctant to add more (Gartner,
2015). In contrast, conversational agents require no
separate installation, as they are accessible via mes-
saging apps, which are likely to be already installed
on a user’s smartphone. Conversational agents are an
alternative to improve customer support and digitize
processes like claim handling.
The objective of this work is to facilitate the cre-
ation of conversational agents by defining the traits
of an agent more clearly using a (1) classification
framework, which is based on current literature and
research topics, and systematically analyzing (2) use
cases and requirements in an industry, shown in the
example insurance scenario. The applicability of this
approach is shown by implementing a prototype (3)
including an evaluation. Furthermore, we derive key
findings for conversational agent implementation in
insurance companies and open points for research.
2 RELATED WORK
In this section we investigate work in the area of con-
versational agents, dialog management, and research
applications in insurance.
McTear et al. (2016) offer detailed explanations
about background and history of conversational in-
Koetter, F., Blohm, M., Kochanowski, M., Goetzer, J., Graziotin, D. and Wagner, S.
Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies.
DOI: 10.5220/0007252100190030
In Proceedings of the 11th International Conference on Agents and Artificial Intelligence (ICAART 2019), pages 19-30
ISBN: 978-989-758-350-6
Copyright
c
 2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
19

terfaces as well as techniques to build and evaluate
own agent applications. Another literature review
about chatbots was provided by Cahn (2017), where
common approaches and design choices are summa-
rized followed by a case study about the functioning
of IBM’s chatbot Watson, which became famous for
winning the popular quiz game Jeopardy! against hu-
mans.
Many chatbot applications have already been built
nowadays with the goal to solve actual problems. One
example is PriBot, a conversational agent, which can
be asked questions about an application’s privacy pol-
icy, because users tended to skip reading the often
long and difficult to understand privacy notices. Also,
the chatbot accepts queries of the user which aim to
change his privacy settings or app permissions (Hark-
ous et al., 2016).
In the past there have already been several studies
with the goal to evaluate how a conversational agent
should behave for being considered as human-like as
possible. In one of them, conducted by Kirakowski
et al. (2009), fourteen participants were asked to talk
to an existing chatbot and to collect key points of con-
vincing and unconvincing characteristics. It turned
out that the bot’s ability to hold a theme over a longer
dialog made it more realistic. On the other hand, not
being able to answer to a user’s questions was re-
garded as an unsatisfying characteristic of the artifi-
cial conversational partner (Kirakowski et al., 2009).
In another experiment, which was done by
S
¨
orensen (2017), eight users had to talk to two differ-
ent kinds of chatbots, one behaving more human-like
and one behaving more robotic. In this context, they
had to fulfill certain tasks like ordering an insurance
policy or demanding an insurance certification. All
of the participants instinctively started to chat by us-
ing natural human language. In cases in which the
bot did not respond to their queries in a satisfying
way, the users’ sentences continuously got shorter un-
til they ended up with writing key words only. Thus,
according to the results of this survey, conversational
agents preferably should be created human-like, be-
cause users seem to be more comfortable when feel-
ing like talking to another human being, especially
in cases in which the concerns are crucial topics like
their insurance policies (S
¨
orensen, 2017).
Dialog management strategies (DM) define the
conversational behaviors of a system in response to
user message and system state McTear et al. (2016).
In industry applications, DM often consists of a
handcrafted set of rules and heuristics, which are
tightly coupled to the application domain (McTear
et al., 2016) and improved iteratively. One problem
with handcrafted approaches to DM is that it is chal-
lenging to anticipate every possible user input and
react appropriately, making development resource-
intensive and error-prone. But if few or no recordings
of conversations are available, these rule-oriented
strategies may be the only option.
As opposed to the rule-oriented strategies, data-
oriented architectures work by using machine learn-
ing algorithms that are trained with samples of di-
alogs in order to reproduce the interactions that are
observed in the training data. These statistical or
heuristical approaches to DM can be classified into
three main categories: Dialog modeling based on re-
inforcement learning, corpus-based statistical dialog
management, and example-based dialog management
(simply extracting rules from data instead of manually
coding them) (McTear et al., 2016; Spierling, 2005).
Spierling (2005) highlights neural networks, Hidden-
Markov Models, and Partially Observable Markov
Decision Processes as possible implementation tech-
nologies.
The following are common strategies for rule-
based dialog management:
• Finite-state-based DM uses a finite state machine
with handcrafted rules, and performs well for
highly structured, system-directed tasks (McTear
et al., 2016).
• Frame-based DM follows no predefined dialog
path, but instead allows to gather pieces of in-
formation in a frame structure and no specific or-
der. This is done by adding an additional entity-
value slot for every piece of information to be col-
lected and by annotating the intents in which they
might occur. Using frames, a less restricted, user-
directed conversation flow is possible, as data is
captured as it comes to the mind of the user (Rud-
nicky and Xu, 1999).
• Information State Update represents the informa-
tion known at a given state in a dialog and up-
dates the internal model each time a participant
performs a dialog move, (e.g. asking or answer-
ing). The state includes information about the
mental states of the participants (beliefs, desires,
intentions, etc.) and about the dialog (utterances,
shared information, etc.) in abstract representa-
tions. Using so-called update moves, applicable
moves are chosen based on the state (Traum and
Larsson, 2003).
• Agent-based DM uses an agent that fulfills con-
versation goals by dynamically using plans for
tasks like intent detection and answer genera-
tion. The agent has a set of beliefs and goals
as well as an information base which is updated
throughout the conversation. Within this informa-
ICAART 2019 - 11th International Conference on Agents and Artificial Intelligence
20

tion framework the agent continuously prioritizes
goals and autonomously selects plans that max-
imize the likelihood of goal fulfillment (Nguyen
and Wobcke, 2005).
Chu et al. (2005) describes how multiple DM ap-
proaches can be combined to use the best strategy for
specific circumstances.
A virtual insurance conversational agent is de-
scribed by Yacoubi and Sabouret (2018), utiliz-
ing TEATIME, an architecture for agent-based DM.
TEATIME uses emotional state as a driver for ac-
tions, e.g. when the bot is perceived unhelpful, that
emotion leads the bot to apologize. The shown exam-
ple bot is a proof of concept for TEATIME capable
of answering questions regarding insurance and react
to customer emotions, but does not implement a full
business process.
Kowatsch et al. (2017) describe a text-based
healthcare chatbot that acts as a companion for
weightloss but also connects a patient with healthcare
professionals. The chat interface supports non-textual
inputs like scales and pictorials to gather patient feed-
back. Study results showed a high engagement with
the chatbot as a peer and a higher percentage of auto-
mated conversation the longer the chatbot is used.
Overall, these examples show potential for con-
versational agents in the insurance area, but lack sup-
port for complete business processes.
3 CLASSIFICATION OF
CONVERSATIONAL AGENTS
The idea of conversational agents that are able to
communicate with human beings is not new: In
1966, Joseph Weizenbaum introduced Eliza, a virtual
psychotherapist, which was able to respond to user
queries using natural language and which could be
considered as the first chatbot (Weizenbaum, 1966).
Nowadays, the idea of speaking machines has experi-
enced a revival with the emergence of new technolo-
gies, especially in the area of artificial intelligence.
Novel machine learning algorithms allow developers
to create software agents in a much more sophisti-
cated way and in many cases they already outper-
form previous statistical NLP methods (McTear et al.,
2016). Additionally, the importance of messaging
apps such as WhatsApp or Telegram has increased
over the last years. In 2015, the total number of peo-
ple using these messaging services outran the total
number of active users in social networks for the first
time. Today, each of these app has about between 200
million and 1.5 billion users (Inc, 2018).
As a highly popular topic in 2016 (Guzmn and
Pathania, 2016), a great variety of different chatbots
evolved together with an equally wide range of ter-
minologies. For being able to draw a big picture of
the current trends in the area of conversational agents,
we divide them into the following four common cate-
gories:
• Chatterbots: Bots with focus on small talk
and realistic conversations, not task-oriented, e.g.
Cleverbot (Carpenter, 2018).
• (Virtual, Intelligent, Cognitive, Digital, Per-
sonal) assistants (VPAs): Agents fulfilling tasks
intelligently based on spoken or written user input
and with the help of data bases and personalized
user preferences (Cooper et al., 2008) (e.g. Ap-
ple’s Siri or Amazon’s Alexa (Dale, 2016)).
• Specialized digital assistants (SDAs): Focused
on a specific domain of expertise, goal-oriented
behavior (Dale, 2016).
• Embodied conversational agents (ECAs): Vi-
sually animated agents, e.g. in form of avatars
or robots (Radziwill and Benton, 2017), where
speech is combined with gestures and facial ex-
pressions.
Figure 1 shows the results of evaluating these four
classes in terms of different characteristics such as re-
alism or task orientationbased on own literature re-
search. Chatterbots provide a high degree of enter-
tainment since they try to imitate the behavior of hu-
man beings while chatting, but there is no specific
goal to be reached within the scope of these conversa-
tions. In contrast, general assistants like Siri or Alexa
are usually called by voice in order to fulfill a specific
task. Specialized assistants concentrate even more on
achieving a specific goal, which often comes at the
expense of realism and user amusement because their
ability to respond to not goal-oriented conversational
inputs like small talk is mostly limited. The best feel-
ing of companionship can be experienced by talking
to an embodied agent, since the reactions of these bots
are closest to human-like behavior.
When looking at these classification results, a
broad spectrum of various possible agents is offered.
Therefore, a restriction depending on the specific use
case has to be made first, before the realization of
a prototypical chatbot can be tackled. Since Fraun-
hofer IAO aims to investigate solutions supporting
processes in the insurance domain, creating a proto-
type with the properties of a SDA is necessary, be-
cause the main purpose in this scenario is to perform
and successfully complete a certain task (e.g. report-
ing a claim). Furthermore, adding additional chatter-
bot features such as the ability to do small talk in a
Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies
21

limited goal-oriented scope could lead to a more re-
alistic and human-like user experience. However, be-
fore considering detailed design and implementation
choices, it is helpful to take a general look at the role
of chatbots within the special environment of insur-
ance companies for identifying essential issues and
needs.
Companionship
Realism
Entertainment
Textual
Task Or ientation
Spoken
Chatterbots
General
Digital Assistants
Embodied
Conversational Agents
Specialized
Digital Assistants
0
1
2
3
4
5
6
7
Figure 1: Classification of conversational agents with their
characteristics (own presentation). Values between 0 and
7 indicate how strong a characteristic applies for the given
type of agent.
4 CHATBOTS IN INSURANCE
Insurance is an important industry sector in Germany,
with 560 companies that manage about 460 million
policies (Schwark and Theis, 2014). However, the in-
surance sector is under a high cost pressure, which
shows in a declining employee count and low mar-
gins (Stange and Reich, 2015). The insurance market
is saturated and has transitioned from a growth mar-
ket to a displacement market (Aschenbrenner et al.,
2010). For the greater part, German insurance com-
panies have used conservative strategies, caused by
risk aversion, long-lived products, hierarchical struc-
tures, and profitable capital markets (Zimmermann
and Richter, 2015). As these conditions change, so
must insurance companies.
Insurance is an industry with low customer en-
gagement, as an insurer traditionally has basically
two touch points to interact with customers: selling
a product and the claims process. A study found that
consumers interact less with insurers than with any
other industry, so the consumer experience with insur-
ers tends to lag behind others (Niddam et al., 2014).
Many insurance companies have heterogeneous
IT infrastructures incorporating legacy systems
(sometimes from two or more companies as the result
of a merger) (Weindelt, 2016). These grown architec-
tures pose challenges when implementing new data-
driven or AI solutions, due to issues like data qual-
ity, availability and privacy. Nonetheless, the high
amount of available data and complex processes make
insurance a prime candidate for machine learning and
data mining. The adoption of AI in the insurance sec-
tor is in early stages, but accelerating, as insurance
companies strive to improve service and remain com-
petitive (Nordman et al., 2017).
Conversational agents are one AI technology at
the verge of adoption. In 2017, ARAG launched
a travel insurance chatbot, quickly followed by bots
from other insurance companies (Gorr, 2018). While
these chatbots are still experimental and implement
narrow use cases, these first implementations prove
public interest and feasibility.
To identify areas of possible chatbot support, we
surveyed the core business processes of insurance
companies as described in Aschenbrenner et al.
(2010) and Horch et al. (2012). Three core areas
of insurance companies are customer-facing: market-
ing/sales, contract management and claim manage-
ment. Figure 2 shows the main identified processes
related to this area.
marketing/
sales
sales talk
underwriting
sell policy
contract
management
contract
change
change of
personal data
customer-
facing
processes
claim
management
damage claim
reporting
claim
assessment
cancellation
claim
settlement
claim
adjustment
billing
Figure 2: Customer-facing insurance processes (based
on Aschenbrenner et al. (2010) and Horch et al. (2012)).
We identified all these processes as possible use
cases for conversational agent support, in particular
support by SDAs.
Furthermore, we investigated general require-
ments for conversational agents in these processes:
Availability and Ease-of-use Conversational agents
are an alternative to both conventional customer sup-
ICAART 2019 - 11th International Conference on Agents and Artificial Intelligence
22

port (e.g. phone, mail) as well as conventional
applications (e.g. apps and websites). Compared
to these conventional solutions, chatbots offer more
availability than human agents and have less barri-
ers of use than conventional applications, requiring
neither an installation nor the ability to learn a new
user interface, as conventional messaging services are
used (Derler, 2017).
Guided Information Flow. Compared to websites,
which offer users a large amount of information they
must filter and prioritize themselves, conversational
agents offer information gradually and only after the
intent of the user is known. Thus, the search space is
narrowed at the beginning of the conversation without
the user needing to be aware of all existing options.
Smartphone Integration. Using messaging services,
conversational agents can integrate with other smart-
phone capabilities, e.g. making a picture, sending a
calendar event, setting a reminder or calling a phone
number.
Customer Call Reduction. Customer service func-
tions can be measured by reduction of customer calls
and average handling time (Guzmn and Pathania,
2016). SDAs can help here by automating conver-
sations, handling standard customer requests and per-
forming parts of conversations (e.g. authentication).
Human Handover. Customers often use social me-
dia channels to escalate an issue in the expectation of
a human response, instead of an automated one. A
conversational agent thus must be able to differenti-
ate between standard use cases it can handle and more
complicated issues, which need to be handed over to
human agents (Newlands, 2017). One possible ap-
proach is to use sentiment detection, so customer who
are already stressed are not further aggravated by a
bot (Guzmn and Pathania, 2016).
Digitize Claim Handling. Damage claim handling
in insurance companies is a complex process in-
volving multiple departments and stakeholders (Koet-
ter et al., 2012). Claim handling processes are
more and more digitized within the insurance com-
panies (Horch et al., 2012), but paper still dominates
communication with claimants, workshops and ex-
perts. (Fannin and Brower, 2017) defines maturity
levels of insurance processes, defining virtual han-
dling as a process where claims are assessed fully dig-
itally based on digital data from the claimant (e.g. a
video, a filled digital form), and touchless handling as
a fully digital process with no human intervention on
the insurance side. SDAs help moving towards these
maturity levels by providing a guided way to make
a claim digitally and communicate with the claimant
(e.g. in case additional data is needed).
Conversational Commerce is the use of Conversa-
tional Agents for marketing and sales related pur-
poses (Eeuwen, 2017). Conversational Agents can
perform multiple tasks using a single interface. Ex-
amples are using opportunities to sell additional prod-
ucts (cross-sell) or better versions of the product the
customer already has (up-sell) by chiming in with per-
sonalized product recommendations in the most ap-
propriate situations. One example would be to note
that a person’s last name has changed during an ad-
dress update customer service case and offer appro-
priate products if the customer has just married.
Internationalization is an important topic for large
international insurance companies. However, most
frameworks for implementing conversational agents
are available in more than one language. To the best
of our knowledge today the applied conversational
agents in German insurance are optimized only for
one language. So this topic is future work in respect
to the prototype.
Compliance to privacy (GDPR) is usually guaranteed
by the login mechanisms on the insurance sites, there-
fore the topic is out of scope for our research proto-
type. For broader scenarios not requiring identifica-
tion on the insurance site and the usage of the data for
non-costumers, this is an open area of research.
5 PROTOTYPE
Based on the work presented in the last sections and
our talks with insurance companies, we arrived at the
following non-functional requirements that the chat-
bot prototype ideally should fulfill:
• Interoperability: The agent should be able to
keep track of the conversational context over sev-
eral message steps and messengers.
• Portability: The agent can be run on different
devices and platforms (e.g. Facebook Messen-
ger, Telegram). Therefore it should use a unified,
platform-independent messaging format.
• Extensibility: The agent should provide a high
level of abstraction that allows designers to add
new conversational content without having to deal
with complicated data structures or code.
Additionally, the following functional require-
ments should be regarded in the implementation:
• Report a Claim: The system must provide the
possibility for a user to report a damage claim us-
ing the conversational agent (prototype scenario).
• Human Language Understanding: The system
should be able to understand and process the
Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies
23

user’s inputs and intents given in form of written
natural language (German).
• Response Generation: The system should be
able to generate an answer sentence in written
human language (German) according to the user
queries.
For dialog design within the prototype, experi-
menting with machine learning algorithms was the
preferred implementation strategy. For this purpose,
discussions with insurance companies were held to
assess the feasibility of receiving existing dialogs
with customers, for example for online chats, phone
logs or similar. However, such logs generally seem to
be not available at German insurers, as the industry
has self-regulated to only store data needed for claim
processing (GDV, 2012). As a research institute rep-
resents a third party not directly involved in claims
processing, data protection laws forbid sharing of data
this way without steps to secure personal data. During
our talks we have identified a need for automated or
assisted anonymization of written texts as a precon-
dition for most customer-facing machine learning use
cases, at least when operating in Europe (Kamarinou
et al., 2016). However, these issues go beyond the
scope of our current project, but provide many oppor-
tunities for future research.
To still build a demonstrator in face of these chal-
lenges, dialogs for the prototype were manually de-
signed without using real-life customer conversations
and fine-tuned by user testing with fictional damage
claims. As this approach entails higher manual effort
for dialog design, a narrower scenario was chosen to
still allow for the full realization of a customer-facing
process. The chosen scenario was a special case of
the damage claim process: The user has a damaged
smartphone or tablet and wants to make an insurance
claim.
Figure 5 shows the main components of the pro-
totype and their operating sequence when processing
a user message. To provide extensibility prototype
architecture strictly separates service integration, in-
ternal logic and domain logic.
The user can interact with the bot over different
communication channels which are integrated with
different bot API clients. To integrate a different mes-
saging service, a new bot API client needs to be writ-
ten. The remainder of the prototype can be reused.
Once a user has written a message, a lookup of
user context is performed to determine if a conversa-
tion with that user is already in progress. User con-
text is stored in a database so no state is kept within
external messaging services. Afterwards, a typing no-
tification is given to the user, indicating the bot has
received the message and is working on it. This pre-
vents multiple messages by a user who thinks the bot
is not responsive.
In the next step, the message has to be understood
by the bot. In case of a voice message, it is transcribed
to text using a Google speech recognition web ser-
vice.
For natural language understanding, we com-
pared four possible frameworks (Microsoft’s LUIS,
Google’s Dialogflow, Facebook’s wit.ai and IBM’s
Watson) regarding important criteria for prototype
implementation. A comparison table for these criteria
is shown in Table 1. As a result of the comparison,
Dialogflow was chosen as a basic framework.
Table 1: Comparison of Microsoft’s LUIS, Google’s Di-
alogflow, Facebook’s wit.ai, and IBM’s Watson (based
on Davydova (2017)).
LUIS Dialogflow Wit.ai Watson
Python bindings no yes yes yes
German language yes yes in Beta yes
Free service no yes yes no
Remember state yes yes yes yes
Service bound yes yes yes yes
Simple training with effort yes yes yes
Dialogflow is used for intent identification, which
determines the function of a message and based on
that a set of possible parameters (McTear et al., 2016).
For example, the intent of the message “the dis-
play of my smartphone broke” may have the intent
phone broken with the parameter damage type as
display damage, while the parameter phone type
is not given. Together, this information given by Di-
alogflow is a MessageUnderstanding
As soon as the message is understood, the user
context is updated. Afterwards, a response needs to
be generated. This process, which was labeled with
Plan and Realize Response in Figure 5, is shown in
detail in Figure 6.
In the prototype, an agent-based strategy was cho-
sen in order to combine the capabilities of the frame-
based entities and parameters in Dialogflow with a
custom dialog controller based on predefined rules in
a finite state machine. This machine allows to de-
fine rules that trigger handlers and state transitions
when a specific intent or entity-parameter combina-
tion is encountered. That way, both intent and frame
processing happen in the same logically encapsulated
unit, enabling better maintainability and extensibil-
ity. The rules are instances of a set of *Handler
classes such as an IntentHandler for the aforemen-
tioned intent and parameter matching, supplemented
by other handlers, e.g. an AffirmationHandler,
ICAART 2019 - 11th International Conference on Agents and Artificial Intelligence
24

Figure 3: Dialog excerpt of the prototype, showing the pos-
sibility to clarify the phone model via multiple-choice in-
put).
which consolidates different intents that all express
a confirmation along the lines of “yes”, “okay”,
“good” and “correct”, as well as a NegationHandler,
a MediaHandler and an EmojiSentimentHandler
(to analyze positive, neutral, or negative sentiment
of a message with emojis). Each implements their
own matches(MessageUnderstanding) method.
Rules (handlers) are used within the dialog state
machine:
1. Stateless handlers are checked independently of
the current state. For example, a RegexHandler
rule determines whether the formality of the ad-
dress towards the user should be changed (Ger-
man differentiates the informal “du” and the for-
mal “Sie”)
2. Dialog States map each possible state to a list
of handlers that are applicable in that state.
For instance, when the user has given an an-
swer and the system asks for explicit confirma-
tion in a state USER CONFIRMING ANSWER, then an
AffirmationHandler and a NegationHandler
capture “yes” and “no” answers.
3. Fallback handlers are checked if none of the ap-
plicable state handlers have yielded a match for
an incoming MessageUnderstanding. These
fallbacks include static, predefined responses
with lowest priority (e.g. small talk), as well as
handlers to repair the conversation by bringing the
user back on track or changing the topic.
At first, the system had only allowed a single state
to be declared at the same time in the router. How-
ever, this had quickly proven to be insufficient as users
are likely to want to respond or refer not only to the
most recent message, but also to previous ones in the
chat. With only a single contemporaneous state, the
user’s next utterance is always interpreted only in that
state. In order to make this model resilient, every
state would need to incorporate every utterance that
the user is likely to say in that context. As this is
not feasible, the prototype has state handlers that al-
low layering transitions on top of each other, allowing
multiple simultaneous states which may advance indi-
vidually.
To avoid an explosion of active states, the system
has state lifetimes: new states returned by callbacks
may have a lifetime that determines the number of di-
alog moves this state is valid for. On receiving a new
message, the planning agent decreases the lifetimes of
all current dialog states by one, except for the case of
utter non-understanding (“fallback” intent). If a state
has exceeded its lifetime, it is removed from the pri-
ority queue of current dialog states.
Figure 6 contains details about how the system
creates responses to user queries. Based on the ap-
plicable rule, the conversational agent performs chat
actions (e.g. sending a message), which are generated
from response templates, taking into account dialog
state, intent parameters, and information like a user’s
name, mood and preferred level of formality.
RuleHandlers, states and other dialog specific im-
plementations are encapsulated, so a new type of di-
alog can be implemented without needing to change
the other parts of the system.
Generated chat actions are stored in the user con-
text and performed for the user’s specific messenger
using the bot API. As the user context has been up-
dated, the next message by the user continues the con-
versation.
Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies
25

Before the user utters an intent to make a damage
claim, the prototype explains its functionality and of-
fers limited small talk. As soon as the user wants to
make a damage claim, the conversational agent gath-
ers the required information using a predetermined
questionnaire. Questions concern type of damage,
damaged phone, phone number, IMEI, damage time,
damage event details, etc. Answers are interpreted us-
ing dialog flow (e.g. determining a point in time). In-
terpretation results have to be confirmed by the user.
In case the answer is not understood, not correct or
not confirmed, the question is repeated. Alternatively,
for specific questions domain specific actions for clar-
ification are implemented. For example, a choice for
specific phone model is shown in Figure 3. For each
question, users may ask for details or for an example
answer. A skip intent is recognized and causes the
dialog to advance to the next question if the current
question is optional.
After the questionnaire is concluded, the bot
thanks the user and stores the data. In a real-life ap-
plication, claim management systems would be inte-
grated to automatically trigger subsequent processes.
6 EVALUATION
To evaluate the produced prototype’s quality and per-
formance, we conducted a model trial with the goal
to report a claim by using the chatbot without having
any further instructions available.
Of the 14 participants (who all had some techni-
cal background), 35.7% claimed to regularly use chat-
bots, 57.1% to use them occasionally, and only 7.1%
stated that they had never talked to a chatbot before.
However, all participants were able to report a claim
within a range of about four minutes, resulting in an
overall task completion rate of 100%.
Additionally, the users had to rate the quality of
their experiences with the conversational agent by fill-
ing out a questionnaire. For each question they could
assign points between 0 (did not apply at all) and
10 (did apply to the full extent). The most impor-
tant quality criteria, whose choice was oriented on the
work of Radziwill and Benton (2017), are listed with
their average ratings in Figure 4 and are discussed in
detail.
With an average of 8 points for Ease of Use, the
users had no problems with using the bot to solve the
task. In the same way, 8.3 points for Appropriate For-
mality indicate that the participants were comfortable
with the formal and informal language the bot talked
to them. Only one user stated that he felt worried
about permanently being called by his first name af-
ter he told it. Fewer points were given for the bot’s
degree of human-like behavior: The rating for con-
vincing Natural Interaction with 7.9 points may be
due to the fact that the conversation was designed in
a strongly questionnaire-oriented way, which might
have restricted the feeling of having a free user con-
versation. Also, the satisfaction with given answers to
users’ domain specific questions was considered quite
(but not totally) convincing with 7.6 points. The least
convincing experience was that chatbot’s Personality,
which was rated with only 5.2 points on average. This
is not surprising, since during this work we put com-
paratively less efforts in strengthening the agent’s per-
sonal skills as it does not even introduce itself with a
name, but instead mainly acts on a professional level,
always concentrating on the fulfillment of its task.
With 7.2 points, talking to the chatbot was experi-
enced as quite Funny & Interesting, but still with a
lot of room for further improvement. Similarly, the
agent’s Entertainment capabilities, which are at 7.7
points on average at the moment, could be upgraded
by extending the conversational contents with addi-
tional enjoyable features not related to the question-
naire. For the future we plan to do another larger eval-
uation on a bigger and more heterogeneous group of
participants.
Ease of Use
Appropriate Formality
Natural Interaction
Response Quality
Personality
Funny & Interesting
Entertainment
0
2
4
6
8
10
8
8.3
7.9
7.6
5.2
7.2
7.7
Average rating points
Figure 4: Survey results: average user experience ratings
(fourteen participants, 0..10 points).
7 CONCLUSIONS AND
OUTLOOK
In this work we have investigated the potential of us-
ing conversational agents in insurance companies by
ICAART 2019 - 11th International Conference on Agents and Artificial Intelligence
26

Service Wrappers Internal System
User text
User
User text
Databases
Databases
Message
User Lookup
User
Unify Update
Add Update
<< get or create >>
Us t
Add Update
Store Update
Context
Build next Actions
dd ChatActions
Store ChatActions
Perform Actions
Message
g
g
Figure 5: Sequence diagram of the conversational agent prototype.
general research and implementing a prototype. We
determined which classes of conversational agents are
useful for insurance companies, which insurance pro-
cesses can be supported, and what the requirements
and motivations for using conversational agents in in-
surance are. These findings can be used to facilitate
the development of conversational agents. We found
a need for Specialized Digital Assistants in customer
facing processes.
Based on these findings we formulated require-
ments for conversational agents in insurance and se-
lected the smartphone damage claim as an example
scenario. We implemented this scenario in a proto-
type, using machine learning for intent recognition
but relying on manual dialog design. Instead of a sin-
gle dialog state, we implemented a system of multiple
conversational states enabling more flexible conversa-
tions. We evaluated our prototype with real users and
gathered their reactions with a questionnaire. Overall,
we found that the prototype is able to handle the ex-
ample scenario to the user’s satisfaction. Possible im-
provements in the prototype scenario are a better de-
termination of the desired degree of formality as well
as defining a consistent persona for the agent (a first
Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies
27

ResponseComposer
Response Templates
Figure 6: Detailed sequence diagram of the response generation in the conversational agent prototype.
ICAART 2019 - 11th International Conference on Agents and Artificial Intelligence
28

step would be to provide a name).
The findings indicate technology is ready to im-
plement conversational agents for insurance customer
service scenarios. However, as real-life scenarios are
broader than the example scenario, considerably more
effort is necessary to design the dialogs. One area not
covered by the prototype is human handover in case
the conversational agent cannot complete and interac-
tion to the user’s satisfaction.
Data protection and privacy remain open areas for
research. Legal and practical questions regarding data
collection, storage and processing must be worked
on alongside technical requirements, as they tend to
be complex and have limited precedent (Smart-Data-
Begleitforschung, 2018).
In future research, we would like to extend the
prototype to different scenarios as well as perform a
real-life evaluation with an insurance partner to quan-
tify the benefits of agent use, e.g. call reduction, suc-
cess rate, and customer satisfaction.
REFERENCES
Aschenbrenner, M., Dicke, R., Karnarski, B., and Schweig-
gert, F. (2010). Informationsverarbeitung in Ver-
sicherungsunternehmen. Springer.
Cahn, J. (2017). Chatbot: Architecture, design, & develop-
ment.
Carpenter, R. (2018). Cleverbot.
https://www.cleverbot.com/.
Chu, S.-W., O’Neill, I., Hanna, P., and McTear, M. (2005).
An approach to multi-strategy dialogue management.
In Ninth European Conference on Speech Communi-
cation and Technology, pages 865–868.
Cooper, R. S., McElroy, J. F., Rolandi, W., Sanders, D.,
Ulmer, R. M., and Peebles, E. (2008). Personal virtual
assistant. US Patent 7,415,100.
Dale, R. (2016). Industry watch: The return of the chatbots.
Natural Language Engineering, 22(5):811–817.
Davydova, O. (2017). 25 chatbot platforms: A compar-
ative table. https://chatbotsjournal.com/25-chatbot-
platforms-a-comparative-table-aeefc932eaff.
Derler, R. (2017). Chatbot vs. app vs. website chatbots
magazine. https://chatbotsmagazine.com/chatbot-vs-
app-vs-website-en-e0027e46c983.
Eeuwen, M. (2017). Mobile conversational commerce:
messenger chatbots as the next interface between busi-
nesses and consumers. Master’s thesis, University of
Twente.
Fannin, T. and Brower, B. (2017). 2017 future of claims
study. Technical report, LexisNexis.
Gartner (2015). Market trends: Mobile
app adoption matures as usage mellows.
https://www.gartner.com/newsroom/id/3018618.
GDV (2012). Verhaltensregeln f
¨
ur den Umgang mit
personenbezogenen Daten durch die deutsche
Versicherungswirtschaft. http://www.gdv.de/wp-
content/uploads/2013/03/GDV Code-of-
Conduct Datenschutz 2012.pdf. Datum des Aufrufes
des Dokumentes: 11.02.2015.
Gorr, D. (2018). Ein Versicherungsroboter f
¨
ur
gewisse Stunden. Versicherungswirtschaft
Heute. http://versicherungswirtschaft-
heute.de/schlaglicht/ein-versicherungsroboter-fur-
gewisse-stunden/.
Guzmn, I. and Pathania, A. (2016). Chatbots in customer
service. Technical report, Accenture.
Harkous, H., Fawaz, K., Shin, K. G., and Aberer, K. (2016).
Pribots: Conversational privacy with chatbots. In
Twelfth Symposium on Usable Privacy and Security
(SOUPS 2016), Denver, CO. USENIX Association.
Horch, A., Kintz, M., Koetter, F., Renner, T., Weidmann,
M., and Ziegler, C. (2012). Projekt openXchange:
Servicenetzwerk zur effizienten Abwicklung und Opti-
mierung von Regulierungsprozessen bei Sachsch
¨
aden.
Fraunhofer Verlag, Stuttgart.
Inc, S. (2018). Most popular messaging apps 2018.
https://www.statista.com/statistics/258749/most-
popular-global-mobile-messenger-apps/.
Kamarinou, D., Millard, C., and Singh, J. (2016). Machine
learning with personal data. Queen Mary School of
Law Legal Studies Research Paper, (247).
Kirakowski, J., O’Donnell, P., and Yiu, A. (2009). Estab-
lishing the hallmarks of a convincing chatbot-human
dialogue. In Maurtua, I., editor, Human-Computer In-
teraction, chapter 09. InTech, Rijeka.
Koetter, F., Weisbecker, A., and Renner, T. (2012). Busi-
ness process optimization in cross-company service
networks: architecture and maturity model. In SRII
Global Conference (SRII), 2012 Annual, pages 715–
724. IEEE.
Kowatsch, T., Nißen, M., Shih, C.-H. I., R
¨
uegger, D., Vol-
land, D., Filler, A., K
¨
unzler, F., Barata, F., Hung, S.,
B
¨
uchter, D., et al. (2017). Text-based healthcare chat-
bots supporting patient and health professional teams:
Preliminary results of a randomized controlled trial on
childhood obesity. In Persuasive Embodied Agents for
Behavior Change (PEACH2017). ETH Zurich.
McTear, M., Callejas, Z., and Griol, D. (2016). The Con-
versational Interface, volume 6. Springer.
Newlands, M. (2017). 10 ways ai
and chatbots reduce business risks.
https://www.entrepreneur.com/article/305073.
Nguyen, A. and Wobcke, W. (2005). An agent-based ap-
proach to dialogue management in personal assistants.
In Proceedings of the 10th international conference on
Intelligent user interfaces, pages 137–144. ACM.
Niddam, M., Barsley, N., Gard, J.-C., and Cotro-
neo, U. (2014). Evolution and revolution:
How insurers stay relevant in a digital future.
https://www.bcg.com/publications/2014/insurance-
technology-strategy-evolution-revolution-how-
insurers-stay-relevant-digital-world.aspx.
Nordman, E., DeFrain, K., Hall, S. N., Karapiperis, D., and
Obersteadt, A. (2017). How artificial intelligence is
changing the insurance industry.
Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies
29

Pohl, V., Kasper, H., Kochanowski, M., and Renner, T.
(2017). Zukunftsstudie 2027 #ichinzehnjahren (in
german). http://s.fhg.de/zukunft2027.
Power, J. D. (2017). 2017 u.s. auto claims satisfaction
study. http://www.jdpower.com/resource/jd-power-
us-auto-claims-satisfaction-study.
Radziwill, N. and Benton, M. (2017). Evaluating quality of
chatbots and intelligent conversational agents. arXiv
preprint arXiv:1704.04579.
Renner, T. and Kochanowski, M. (2018). Innovationsnet-
zwerk digitalisierung f
¨
ur versicherungen (in german).
http://s.fhg.de/innonetz.
Rudnicky, A. and Xu, W. (1999). An agenda-based dia-
log management architecture for spoken language sys-
tems. In IEEE Automatic Speech Recognition and Un-
derstanding Workshop, volume 13.
Schippers, B. (2016). App fatigue.
https://techcrunch.com/2016/02/03/app-fatigue/.
Schwark, P. and Theis, A. (2014). Statistisches Taschen-
buch der Versicherungswirtschaft. Gesamtverband der
Deutschen Versicherungswirtschaft (GDV), Berlin.
Smart-Data-Begleitforschung (2018). Smart Data
- Smart Solutions. https://www.digitale-
technologien.de/DT/Redaktion/DE/Downloads/-
Publikation/2018 06 smartdata smart solutions.pdf.
S
¨
orensen, I. (2017). Expectations on chatbots among
novice users during the onboarding process.
http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-
202710.
Spierling, U. (2005). Interactive digital storytelling: to-
wards a hybrid conceptual approach. Worlds in
Play: International Perspectives on Digital Games
Research.
Stange, A. and Reich, N. (2015). Die zukunft der deutschen
assekuranz: chancenreich und doch ungewiss. In
Change Management in Versicherungsunternehmen,
pages 3–9. Springer Gabler, Wiesbaden.
Traum, D. R. and Larsson, S. (2003). The information state
approach to dialogue management. In Current and
new directions in discourse and dialogue, pages 325–
353. Springer.
Weindelt, B. (2016). Digital transformation of industries.
Technical report, World Economic Forum and Accen-
ture.
Weizenbaum, J. (1966). Eliza - a computer program for
the study of natural language communication between
man and machine. Commun. ACM, 9(1):36–45.
Yacoubi, A. and Sabouret, N. (2018). Teatime: A formal
model of action tendencies in conversational agents.
In ICAART (2), pages 143–153.
Zimmermann, G. and Richter, S.-L. (2015). Gr
¨
unde f
¨
ur
die ver
¨
anderungsaversion deutscher versicherungsun-
ternehmen. In Change Management in Versicherung-
sunternehmen, pages 11–35. Springer.
ICAART 2019 - 11th International Conference on Agents and Artificial Intelligence
30