Chatbot Pricing: How new models reduce enterprise risk
An emerging technology
Conversational AI platforms – known as virtual assistants or chatbots – represent a promising technology that is already projected to cut business expenses by as much as $8 billion in less than five years. However, despite the transformative revolution and the promise it brings, chatbots are not mature enough yet for businesses to fully rely on them to perform all tasks.
The pricing challenge
Still, forward-thinking businesses – from small startups to large enterprises – recognize the potential benefits of chatbot development. The decision to move forward is typically determined based on a number of factors, including target audience, required activities, specific business needs, and often most importantly, price.
Setting a budget for a new technology is always delicate, but for customer-facing conversational AI platforms, it is even more challenging as the risks are great. Will the technology be accurate or will the bot’s errors make the company look unprofessional? Will the chatbot be effective or will customers need to be transferred to a human agent anyway? How long will it take until the customers accept the new technology? What if the technology is too effective and the associated costs are higher than expected?
Cost vs. value
The potential financial value of implementing virtual assistants is clear. After all, a standard interaction with a chatbot generally costs less than $1. If that interaction can replace a live communication with the call center, which costs $10-12 on average, or help avoid a technician dispatch, which could cost $150, the savings are obvious.
However, the efficiency of conversational AI platforms will not be seamless from the start. If out of 10 interactions with chatbots, four customers terminate the chat in frustration, three are escalated to a live agent, two offer the customer erroneous information, and one is successfully solved by the bot, then the end results do not justify the means.
The cost involved with taking repeat calls or lowered NPS is simply not worth the amount saved on the few interactions the bot is able to handle independently.
This cost-value analysis is being studied by CX and customer service departments in call centers around the world. What is the optimal model to use when adopting chatbot technologies that will ensure minimal financial risk?
Common chatbot pricing models
Today’s chatbot developers typically offer a pricing model that encompasses each of the following three components:
Development and implementation fee – Companies pay a hefty one-time charge that includes project management, UX design, software customization, testing, infrastructure setup, implementation, and training the model’s algorithms.
Subscription-based fee – Companies pay a recurring monthly fee which entitles them to access resources such as storage, engines, updates, and support. Fees vary based on number of active users, hosting requirements, chatbot complexity and 3d party integrations needed.
Consumption-based fee – Companies pay a sliding-scale fee based on the number of sessions that the virtual assistant has with customers. Per session fees range from a few cents to $2 each. While the first two components are fixed, this element is dynamic.
By combining elements of these three models, a win-win situation should emerge: the vendor ensures that the development and integration costs are paid even in the event that the project is suspended or the technology is not adopted, and the company benefits from a flexible pricing model that is easier to swallow.
Snags in common models
In practice, companies find it difficult to reach the optimal balance between these three pricing elements, and even harder to evaluate in advance the bottom-line cost of conversational AI technology. Businesses – in all stages of maturity – are being asked to dedicate budgets in a reality where technology is not mature enough, and its adoption and potential ROI – at least for the short term – are unknown.
For example, some enterprises find it hard to justify an ‘implementation fee’ before knowing whether the chatbot will be adopted successfully. On the other hand, some enterprises are wary of locking themselves in to a long-term recurring subscription.
Still others prefer to avoid the consumption-based model with its potential to escalate costs unexpectedly if adoption is faster than anticipated, leading them to become victims of their own success. Similarly, they fear high consumption costs for sessions that are only partially-successful or even worse, totally unsuccessful.
Alternative models for chatbots pricing: Performance-based pricing
In light of the challenges involved with the common pricing models, new innovative pricing models have emerged that focus more on ensuring the enterprise’s success. These new models make it possible for businesses to more easily digest the cost of the chatbot due to minimized risk. Here are some examples of success-based models that have been making a splash in the market:
Successful consumption model – This model is similar to the traditional consumption-based model where the customer pays a fixed price per session with a bot, but with an important twist. In this innovative model, the customer is charged only for successful sessions, with success defined through a number of variable options.
For example, if the conversation reaches a point where the customer receives a thank you page, or a survey link, to ensure that the customer has not bounced out the session before completion.
Contingency model – In this model, success is tied directly with the value generated from the conversation. For example, an enterprise will pay only when a technician dispatch or a call to a human agent was avoided. Another scenario is when success is tied with customer satisfaction – if the NPS score or survey results are positive, then the enterprise pays.
This model is becoming more and more common in the BPO domain, where the entire industry is shifting to success-based invoicing. If contingencies become the standard in agent outsourcing, it may be only a matter of time until it becomes a common model for virtual assistants as well.
Pay per box – With this model, the enterprise pays a set fee based on the number of products or services sold to customers. For example, for every $50 set top box a telecommunications company sells, it will pay the bot vendor $1 to support the device, regardless of how many sessions are generated. In this sense, the bot becomes part of the product package – right alongside the user manual.
In a service-based industry, for example, a utility may pay a set price per subscriber to its premium service.
Eyes open to the future
Assessing the financial value of conversational AI platforms – an emerging and promising technology – is complicated because it’s relatively new, and its success and maturity have yet to be fully proven.
While a number of pricing models are currently available, all entail a risk on the part of the enterprise. The key to facilitating the widespread adoption of virtual assistants is to formulate an optimal pricing structure that will support the enterprise in achieving success with the technology.