Chatbots for Insurance – Progessive, Allstate, GEICO, and More

Niccolo Mejia

Niccolo is a content writer and Junior Analyst at Emerj, developing both web content and helping with quantitative research. He holds a bachelor's degree in Writing, Literature, and Publishing from Emerson College.

Chatbots for Insurance - Progessive, Allstate, GEICO, and More

Although it’s clear that AI is only beginning to make its way into the insurance sector, Accenture estimates that big investments in AI could increase the annual profitability of insurers by 20%.

It follows that applications such as chatbots would find their way into the insurance world, especially because we are already seeing a rise in the use of AI for behavioral premium pricing, improving customer experience, and speeding up the claims process. Insurance giants like Progressive of Gieco have massive volumes of customer service requests from chat, phone, and email, and these reems of data are exactly what a conversational interface would be trained on.

We researched the space to better understand where chatbots come into play in the insurance industry and to answer the following questions:

  • What types of chatbot applications are currently in use in insurance?
  • What tangible results have chatbots driven in insurance?
  • Are there any common trends among these innovation efforts? How could these trends affect the future of InsurTech?

This report covers insurance enterprises like Progressive, Allstate, and Geico, as well Progress, an enterprise that claims to offer a chatbot software for insurers. We’ll begin by analyzing the chatbot initiatives at Progressive:


Progressive offers a chatbot called Flo, which the company claims can help customers file claims, move payment dates, and get auto insurance quotes. Progressive claims that the chatbot uses machine learning and a cloud-based API that pulls data from social media responses and training data.

We can infer the machine learning model behind the software was trained on thousands of insurance questions from potential and returning customers involving numerous topics such as filing claims, billing, and deductible rates. This text data would have been labeled as claims-related, billing-related, deductible-related, or related to other categories Progressive chose. The labeled text data would then be run through the software’s machine learning algorithm.

This would have trained the algorithm to discern the chains of text that, to the human brain, might be interpreted as a claims-, billing-, or deductible-related question as displayed in a question on facebook messenger. Progressive claims users can interact with the chatbot from its Facebook page using Facebook Messenger. Users can ask the chatbot insurance questions regarding numerous topics such as filing claims, billing, and premium rates. Then, the user’s question runs through the algorithm behind the Flo chatbot, and the algorithm coaxes out the data points in that question that correlate to a certain type of customer service question based on its training data.

For example, the algorithm may have been trained on thousands of customer support tickets with a “deductible-related” label. It would then have “learned” what makes a customer service ticket a deductible-related ticket, and it would end up sending a message back to the user related to deductibles.

Progressive does not have a demonstration video showing how Flo works, and there are no case studies available for the solution.

Progressive claims that Microsoft helped them build their chatbot, which emulates their popular TV mascot in addition to understanding customer questions and responding quickly. Flo uses simple language but adds in wit where appropriate, which may engage customers and tie into Progressive’s marketing. Progressive built the Flo Chatbot using Microsoft Azure Bot Service and LUIS. The company found that the software facilitates updating the bot and its responses without needing to write complex code.

The APIs also provide the ability for the chatbot to expand its database with each customer interaction, so the quality of responses improves over time. According to Microsoft, Progressive updated Flo’s models over 75 times in the first four months of activity to help the chatbot continue to improve itself and customer interactions.

Allstate Business Insurance

Allstate Business Insurance uses a chatbot called Allstate Business Insurance Expert (ABIE), which it claims can help small business owners by answering initial questions such as “what is a deductible?” and “how does the insurance claims process work?” In addition, the company claims their chatbot can improve the relationship between the agent and the customer using what appears to be natural language processing.

We can infer the machine learning model behind the software was trained on thousands of questions from insurance agents involving anything from policy pricing to claims. This text data would have been labeled under categories such as policy-related questions or claims-related questions, for example. The labeled text data would then be run through the software’s machine learning algorithm. This would have trained the algorithm to discern the chains of text that a human might interpret as a policy-related question or a claims-related question, for example, as displayed in a text message.

Allstate claims its ABIE chatbot is located in the business insurance section of their website, where users can pose business insurance questions to it. Then, the algorithm behind the chatbot runs through the question and provides a response out of data from a structured content platform called easyDITA. A structured content platform allows companies to publish content to multiple sources at the same time and label data within that content under specific categories.

This can also be done with data points within the same piece of content. Developed by Jorsek, easyDITA enables Allstate to single source all of it’s business insurance information for the chatbot. The company publishes all of it’s data to easyDITA so that the chatbot only has to “search” in one place for every purpose. This could allow the chatbot to produce answers for every customer interaction.

Allstate does not have a demonstration video available showing how its software works.

Jorsek claims to have helped Allstate Business Insurance employees access data they required to quote and issue complex business insurance products. The company purportedly accomplished this by teaching the chatbot to pick up more data from customer interactions.

First, Allstate developed a content strategy that focused on what information employees needed to know the most. The team then added keywords and contextual data that allowed easyDITA’s search engine to “understand” the question and deliver a message with relevant results.

For example, they may have added “rate” as a keyword and statistics on the most frequently asked questions regarding the word “rate” or “rates” as metadata so the chatbot has more information when a customer asks, “what kinds of accidents increase the rate of my deductible?” According to Jorsek, Allstate was able to handle 25,000 support tickets a month with the chatbot.


GEICO offers a chatbot called Kate, which they claim can help customers get accurate answers and specific responses to insurance inquiries using what appears to be natural language processing.

We can infer the machine learning model behind the software was trained on thousands of customer service questions and audio clips from support calls involving [insurance topics such as deductibles and roadside assistance. This text and audio data would have been labeled by the type of question being asked, such as deductible questions and roadside assistance questions. The labeled text data and audio data would then be run through the software’s machine learning algorithm.

This would have trained the algorithm to discern the sequences and patterns of 1’s and 0’s that, to the human brain, might be interpreted as either a deductible or roadside assistance question as displayed in text and recorded audio.

GEICO claims users can message Kate via the GEICO mobile app with either text or voice to pose insurance questions. The process in which the machine learning system takes user data and turns it into insight is unclear. The chatbot provides an answer to the user’s question, but can also bring the user to the appropriate section of the GEICO mobile app if it concludes the user just needs to get back to a specific menu.

For example, if a user needs roadside assistance but is having trouble locating the roadside assistance menu within the app, the chatbot will navigate the app to the required menu.

We could not find a demonstration video available for the software. Also, GEICO exclusively uses Kate internally, and so there are no available case studies for the software.


Progress offers a software called Native Chat, which the company claims can help insurance companies save on customer service costs and accelerate marketing campaigns. The system uses natural language processing.

We can infer the machine learning model behind the software was trained on thousands of customer service questions involving basic insurance matters such as billing and changing account information  This text data would have been labeled as either billing-related or account information-related, for example.

The labeled text data would then be run through the software’s machine learning algorithm. This would have trained the algorithm to discern the chains of text that might be interpreted as a billing question or someone needing to update their account information as displayed in a messaging app interface.

Progress claims insurance companies can use Native Chat to create chatbots for their company smartphone apps. Customers can make requests to the chatbot after they install the app.

Below is a short 2-minute video demonstrating how Native Chat works:

That said, Progress does not provide any statistics reporting on results the software might generate for insurance companies.

Progress also lists Cigna, Nationwide, and Dr.Shterev Hospital as some of their past clients.

Yogesh Gupta is CEO at Progress. He holds an MS in Computer science from the University of Wisconsin-Madison. Previously, Gupta served as President and CEO at Kaseya.

Takeaways for Business Leaders in Insurance

Microsoft Azure may allow businesses to create chatbots that not only solve interactions quickly but may also be able to retain a brand’s voice and style. Like Flo, insurance chatbots could reinforce marketing strategies in tandem with pleasing the customer. Allstate’s chatbot highlights easyDITA’s capacity for single-sourced information and immediate responses, which may be an asset for any chatbot built to cut down customer interaction times and maximize customer satisfaction.

Even though GEICO is a trusted company, their lack of available information or success stories with their chatbot Kate does not allow us to verify that their chatbot in fact uses machine learning, although it appears it might.

Progressive’s Flo chatbot seems to have the most traction. It received over 15,000 customer questions by March of 2018, less than half a year after the chatbot became available. Also, their chatbot was built by Microsoft Azure, which signals trust in the legitimacy of the AI behind the chatbot. Additionally, the chatbot is linked to the company’s Facebook page with over 4,000,000 followers. These followers post questions to the company Facebook page, like, and share certain content. This is data that can continue to train the algorithm behind the chatbot on how to best respond to customer questions.

Allstate’s ABIE showed the most positive results from our research, stating that it reduced call center interactions by 25,000 calls per month using the service. This could allow the call center to further help people who need more information than the chatbot is able to give.

Insurance companies can expect more possible applications for insurance chatbots in the future, but not necessarily as replacements for their current employees. While it is currently possible to settle claims via insurance chatbot, business leaders should expect to have chatbots as a kind of first line of support for customers, while their employees serve to field more complex tasks.

Business leaders should not expect chatbots to solve every customer interaction, however. Even Progressive’s Flo can fail to understand a question sometimes, but it does connect the customer to a human employee that can offer better help if it does not know the answer to something.

In addition, Progressive’s chatbot is intended for use by employees as well as customers. This is not the case for other chatbots for insurance.

Any chatbot will take some time to integrate into your current system. Microsoft Azure and easyDITA are paid services may require insurance companies purchase a plan with Microsoft or Jorsek that they pay monthly. While it could be beneficial to have a dedicated AI staff that can manage and update these chatbots in house, it is not always necessary. Frequent data updates to chatbots built on these structures, however, will keep them working well and improving.


Header Image Credit: chatbotslife