What Happened
Nearly seven in ten financial institutions have shut down an AI chatbot after deployment, according to new research from Sinch, which surveyed more than 500 financial services companies. The failures span everything from routine payment confirmations to high-stakes fraud alerts. What the data makes clear is this: the technology rarely caused the collapse. The organizations did.
The Communication Angle
Here is the lesson, stated plainly: you cannot automate a conversation you have never bothered to master yourself.
Financial institutions rushed into AI deployment treating it like a software upgrade. Plug it in, turn it on, watch it work. But communication is not a feature you install. It is a set of deliberate choices about tone, timing, word selection, and trust. When you skip those choices and hand the job to a machine, the machine faithfully replicates your confusion at scale.
Think about what these chatbots were being asked to handle. Fraud alerts. Account holds. Sensitive financial disclosures. These are not neutral messages. Every single one of them lands on a customer who is either worried, frustrated, or afraid. The communication challenge is not just accuracy. It is reading the emotional temperature of the moment and responding to it. That requires a clear, intentional voice that someone in the organization had to design deliberately. Most did not. They uploaded a product description and called it a persona.
The fix is not to pull the plug, which is what 69% of these institutions did. The fix is to do the communication work before deployment. That means sitting down and defining exactly what your organization sounds like under pressure. Not in a brand guidelines document that uses words like "empathetic" and "customer-centric." In actual sentences. What do you say when a customer's account is frozen? What is the first sentence? What does the second sentence accomplish? If you cannot answer those questions with a human representative, you absolutely cannot answer them with a bot.
The institutions that kept their AI chatbots running successfully are not sitting on better technology. They are sitting on better communication architecture. They mapped the conversations first. They tested tone. They built in escalation paths so customers could reach a human the moment the interaction required judgment rather than information. That is not an AI strategy. That is a communication strategy that happens to use AI.
This is exactly the kind of scenario I break down in Say It Right Every Time. The chapter on high-stakes messaging gives you a framework for designing communication that holds up under pressure, specifically for moments when the reader is already anxious before they finish the first sentence. The principle is simple: clarity earns trust faster than warmth does. Financial institutions forgot that. Their chatbots reminded them.
Never Be Lost
for Words Again
Get word‑for‑word scripts for the conversations that shape your life, from job interviews and negotiations to difficult talks with family and partners, so you always know exactly what to say and how to say it.
Never Be Lost
for Words Again
Get word‑for‑word scripts for the conversations that shape your life, from job interviews and negotiations to difficult talks with family and partners, so you always know exactly what to say and how to say it.
Key Takeaway
Before you automate any customer communication, write out the five most difficult messages your organization sends. Word for word, as they should actually read. If you cannot write them clearly and confidently for a human representative to deliver, you are not ready to automate them. Get the human version right first. The machine will follow.
