Chatbot Business Models

Few, far, and unproven…

Major VC funded chatbots may be placed into three categories- Artificial Intelligence (AI) led, Marketing led, or Human Intelligence (HUMINT) led; i.e. the chatbot AMH stack. All categories depend on conversations with human beings, however, each group presents remarkably different business models. In this article we will explore the chatbot landscape and proxies for potential markets and business models for each category.

The Chatbot Landscape

Artificial Intelligence Chatbots:

There are three types of AI chatbot. First, Augmented human communication aims to automatically elaborate on conversation (e.g. hailing an Uber or noting a future meeting in a calendar). Second, Human replacement aims to create chatbots sufficient for fully automated conversation (e.g. Microsoft’s chatbot “Tay”). Third, intelligence services (e.g. Google queries or relationship management) where the chatbot provisions smart services as part of the conversation.

Marketing led Chatbots:

These aim to either be a platform for the dissemination of targeted information (e.g. Facebook), or, for the provision of established services in a more convenient format (e.g. Salesforce data, Amazon prices, and/or Yelp reviews).

HUMINT led Chatbots:

These are the “native” chatbots who’s functionality would be difficult to provide in other formats. They try to provide information on the participants thought processes or opinions in view of their conversations (e.g. user testing, sentiment analysis).

Current state of chatbots

The following is an overview of major VC funded chatbots:

Major VC funded Bots. Source CB Insights and Crunchbase

Few major VC funded chatbots currently charge, and the ones that do are dominated by personal assistant type AI bots. The larger portion of bots are currently free to use with limited monetization models or in private betas providing free or custom pricing. All chatbots websites currently imply a SaaS based service model, though the marketing bots (e.g. Mezi) may also imply an advertising model.

This makes it difficult to understand the various chatbots’ potential business models, and demonstrates that the current business models and markets are unproven.

The best analysis of current in use business models in chatbots comes from Benedict Evans at venture capital firm, A16Z. He focuses primarily on the tempered successes of WeChat and Facebook to conclude:

WeChat uses messaging as a platform but doesn’t push interaction down to the level of ‘dumb text’ — it provides its own social and acquisition model but much of the actual interaction with third party services happens in rich UIs (mostly built as web views, as it happens). And after all, if you’re offering a user a question to which there are only two options, should you tell them ‘you can reply ‘red’ or ‘green’’, or should you give them two buttons within the chat? What if there are five options? Should you perhaps construct some sort of on-screen interface for your users that lays out, graphically, the options? You could call, it, perhaps, a ‘GUI’. You could have ‘links’ that you tap on, that load new ‘pages’… And indeed, if you’ve got your chat bot working, does that need to be in Facebook, or could it be on your own website too? It depends what kind of interactions you’re looking for, and maybe whether you’re solving your own problems or your users’.


Therefore we will use proxies for potential businsess models and profitability for each respective category in view of the potential to disrupt a large profitable current market and company.

Potential business models

We will explore potential consumer and enterprise business models for each category of chatbots in view of a proxy large profitable market and company.

AI Chatbots

a. Human augmentation or replacement: Ezra Galston at Breaking VC does a good job of reviewing potential models in augmenting or replacing humans with AI powered chatbots. The message is replacing people even by a small portion by AI is extrodinarly profitable. The business model primarily falls into the enterprise space given few consumer’s directly employ talent.

Augmenting humans- the average margin for human talent marketplaces is ~20% (e.g. Amazon’s mechanical turk, Upwork). For every $100 of services the marketplace charges, the marketplace keeps $20 and pays talent $80. However with AI augmentation, the marketplace could provide the same $100 in service with less than $80 in input from the talent. Even in cases of single digit increases in efficiency (e.g. 7%), the marketplace would see large increases in profitability (e.g. $26 per $100 transaction yielding ~30% increase in profits).

Replacing humans- customer service agents are a large expense for companies large and small. As of January 1st, 2016 the price-to-earnings ratio of the S&P500 is ~22x, and the average SaaS startup valuation is 4:1 revenue. That means every dollar in customer service expenses removed yields a 4–22X return on investment!

Source: BreakingVC- Economics of Chatbots

b. Intelligence Services- The business model proxies may be split into consumer (i.e. Google) and enterprise (i.e. Salesforce). Google yields the vast majority of its monthly $277 in revenue per user per monthfrom adwords directed to Google search queries from consumers. An AI chatbot that can parse questions and provide advertiser sponsored answers would be in a position to disrupt Google’s $75 Billion in annual revenue business.

Salesforce has revenues of ~$6.7 Billion, ~150K customers, with ~25 users per customer. This yields ~$140 in revenue per user per month. Salesforce provides a variety of customer relationship management software that could be disrupted. Chatbot AI could automatically manage much of the enterprise customer relationship; e.g. automatically routing and escalating imporant events.

Marketing Chatbots

a. Consumer: Facebook would present the greatest direct proxy and market for a marketing chatbot to disrupt. A marketing chatbot in this instance would be a social media platform that would monetize the marketing of other company’s services to a captive user base (i.e. targeted advertizing). Facebook has ~$18 Billion in revenue with monthly revenue per user of ~$60.

It should be noted that Facebook views chatbots as a potentially large market as well through a variety of acquisitions and the development of platforms for Facebook messenger and Whatsapp. Creating a chatbot in this space may be difficult as the networks effects are what create such a high level of value. Facebook is competing to dominate this level of consumer marketing in the chatbot space.

WeChat is the greatest proxy we have to success in the space of a marketing based consumer chatbot success. Wechat makes the vast bulk of its revenue through its messaging based payment service and games. Though WeChat has recently expanded to include paid for advertising, it generates less than ~$1 Billion in advertising revenue compared to ~$7 Billion from payments and games.


It may be worthwhile to focus on building atop of either the Facebook or another platform if we use WeChat in conjunction with Facebook’s marketshare as a proxy of future business models. Both dominate their user base due to network effects, however, both are used the preeminent means of marketing campaigns and customer interaction (e.g. WeChat 84% of public accounts). This creates an opportunity for chatbots that help clients use the platform more effectively to market to consumers.


b. Enterprise: Arbitrage between current non-chatbot services being applied to chatbots. This can be considered much like moving the bookstore experience to the internet, or the chatbot Birdly making Salesforce information available on the Slack platform.

However it is difficult to predict the success or market size of the business model behind this arbitrage. The most relevant data is from Facebook’s head of messenger, David Marcus:

There are now more than 30,000 bots on the platform, Marcus said, up from 11,000 in July. But bots have suffered from a perception that they are not nearly as useful as they were advertised to be earlier this year by tech giants including Facebook, Google, and Microsoft.


HUMINT Chatbots:

The most difficult to find a potential market and business model. HUMINT chatbots derive value from the chat itself. These chatbots provide or record information that would not be readily accessible in other formats.

For example in enterprise, Growbot interjects in chat to allow employees to quickly give “props” to another employee for a task well done in order to improve employee evaluations. Growbot could not exist before chat, as it would have been substantially more burdensome to place a task in desktop software, write an email that would create a disjointed record of performance, or maintain a portfolio of work.

For the consumer model, the inclusion of GIFS natively in Apple’s iMessage and Facebook allow for coding of emotion in greater detail than would be available through AI or syntax analysis of a conversation.

There is a concern that applying older monetization models to new forms of information may be inadequate. Mature SaaS based models may be used for enterprise applications, and advertising/e-commerce models may be applied to consumer applications. However, there is a risk of platform value capture from the SaaS models (e.g. the platform making the data exclusive to its developers), and limited advertising value of new information (e.g. Twitter’s limited value in relation to Facebook).

The most relevant business models to these “new” forms of information may be business that revolve around selling synonymous obscure pieces of information.

Qualtrics (a survey provider) is a representative example of monetizing obscure information that would be difficult to find otherwise. Qualtrics launched a service to provide access to market research as part of their survey licesense.


As Qualtrics co-founder and CEO Ryan Smith explains, for $5,000, the price of a single Qualtrics license, companies will have access to 250,000 use cases and examples of 130,000 different surveys. The exchange will also offer up existing Qualtrics clients as technology partners, to provide any spot help a company needs.

These use cases and surveys make readily available obscure market research information that would have been infeasbile to acquire at a reasonable price. The market alone for market research is $30 billion.

a Qualtrics customer who could never afford to hire an expensive market research company logs in and can access instructional videos and articles about how large companies like IDEO and Bain Consulting used the service. Most people with a basic level of computer savvy can copy these steps and see a survey through to the point where it gets sent out, but if a company gets stuck with the more technical data analysis it can hire other Qualtrics’ customers for one-off bursts of work.


The business models for chatbots is few and far between. There is no dominant chatbot business model, and current SaaS based models are inadequate to describe the chatbot AMH stack business dynamics. Potential models rely on disrupting current operating businesses and markets. AI bots have great potential to increase the value of companies by reducing the need for people. Marketing chatbots can hope to either provide a dominant platform for the distribution of advertisements to consumers, or provision marketing services to other businesses. Finally, HUMINT chatbots may provide valuable obscure data more readily or cheapily than current market leaders.