One click too many: When the bot is rewarded for closing the inquiry
The customer relationship management sector is shifting toward a Pay-per-Resolution model. However, aggressive AI automation risks creating conflicts of interest and eroding long-term customer loyalty.
The field of customer relationship management is undergoing a revolution: from assisting human agents to the aspiration of replacing them entirely, and from a SaaS model of payment per user to a payment model based on results. Zendesk has recently been at the center of the discussion with an aggressive push of a new model, under which companies pay between 1.50 and 2 dollars for what is defined as a successful AI resolution. It joins Intercom, which paved the way with an AI agent priced at 0.99 dollars per resolution, and Salesforce, which introduced a model of about 2 dollars for every interaction managed on its platform.
On paper, this is the dream of every CFO. There are no high setup costs, no payment for time when employees are inactive, and the return on investment is clear: you solved the problem — you paid; you didn't solve it — you didn't pay. Data from the research firm Gartner also reinforces this trend, predicting that by 2028, at least 70% of customer journeys in the customer service world will begin with a conversational AI interface. The pressure from management to implement GenAI tools in service and sales has never been higher. But this is exactly where the mine is hidden.
The question that marketing managers should ask is how to define in advance what a successful solution is for the customer, and more importantly — who determines that the inquiry has indeed been resolved. In the world of automated systems, an inquiry can be considered resolved the moment the AI has provided an answer, the customer has not responded for a certain period of time, or has clicked a finish button. In reality, however, at the other end is a customer with expectations, emotions, and limited patience.
What happens when the AI provides a correct answer on paper — for example: "According to company policy, a refund cannot be received after 14 days" — and closes the inquiry? From the AI provider's perspective, the problem is solved. From the customer's perspective, they encountered a bureaucratic wall, experienced alienated service, and left the chat with the feeling that they would not return to the brand.
When the commercial model incentivizes the technology provider to close as many inquiries as possible, a conflict of interest may arise between them and the brand. The bot wants to check a box, while the brand seeks to build a long-term relationship with the customer. Recent studies in the field of customer experience point to a similar phenomenon: organizations that implemented aggressive AI mechanisms improved efficiency metrics, but in many cases saw a decline in customer satisfaction and loyalty. Customers often feel they have been trapped in a "bot loop" that does not allow them to reach a human agent even when the problem is complex.
Customer service and sales are not just a process of transferring information; they are the brand's showcase. This is where trust is built — or eroded. A brand that relies solely on a Pay-per-Resolution model, without human control mechanisms, may lose its most important asset: the relationship with the customer. If the AI solves the problem only apparently, the monthly invoice from the technology provider might be lower, but the price the brand will pay in the long term may be much higher.
Gidi Adlersberg is Head of Business Line at AudioCodes, a company specializing in Voice AI solutions for organizations.
