64% of Customers Want an Escape Hatch From AI, and the Data Shows Why

The Market Is Moving Faster Than Enterprises Can Integrate It

Global AI market spend is projected to grow from 189 billion dollars in 2023 to 4.8 trillion dollars by 2033, according to UN Trade and Development, and businesses are inserting voice AI into customer journeys faster than most can properly integrate it. The stakes of getting this wrong are steep: Qualtrics XM Institute puts the global cost of customers switching brands after a bad experience at 3.8 trillion dollars, ContactBabel’s 2024 UK Contact Centre Decision Maker’s Guide found average call abandonment sitting at 8.4%, and Gartner reported that 64% of customers want the explicit option to avoid AI in customer service altogether.

Customers Aren’t Rejecting AI, They’re Rejecting Bad AI

That last figure is worth sitting with rather than treating as a rejection of automation. Customers are not saying no to AI, they are saying no to bad AI, and the distinction matters enormously for collections and BFSI operations specifically. Voice is the least forgiving channel available. A chatbot exchange can be reread and corrected mid-conversation, but a voice call that confidently delivers a wrong answer about a balance, a payment date, or an account change cannot be undone in the customer’s mind. Salesforce’s State of the AI Connected Customer research found that 72% of customers consider it important to know whether they are speaking with AI or a human, which means the fix is not hiding automation better. It is building it well enough that disclosure does not cost you trust.

Where Voice AI Should Never Guess

The part of this argument most collections teams underweight is the specific list of tasks voice AI should never guess at: identity verification, payments, and account changes. These are exactly the categories a collections voicebot handles constantly, and they are also the categories where a confident wrong answer does the most damage, both to customer trust and to regulatory standing. The practical fix is straightforward but frequently skipped: when a request falls outside what the system can verify, the correct behavior is to say so and hand off, not to generate a plausible-sounding response. Anchoring every regulated answer in a verified knowledge base rather than a model’s general reasoning is the same principle applied to what the system says, not just what it decides to do.

Stability Is What Determines Whether Voice AI Pays Off

The detail that gets lost in most voice AI rollouts is stability. A bot that behaves differently every time a prompt gets tweaked or a new intent gets added creates regressions that show up nowhere on a containment dashboard but land immediately with the caller. For a collections voicebot handling payment conversations across different accents, background noise, and regional phrasing, this means testing cannot be a one-time pre-launch exercise. It has to run continuously against real variation in how people actually speak, or accuracy drifts silently while the metrics still look fine on paper. The teams getting real ROI from voice AI are not the ones celebrating the highest containment numbers. They are the ones tracking whether the system hands off cleanly on the calls it should never have tried to handle alone.

[Read the full report]

Related Post

When 66% of Enterprises Deploy AI Agents Without Human Review

A June 2026 VB Pulse survey of 157 enterprise respondents found that half of organizations have shipped an AI agent or LLM feature that passed internal evaluation and still caused a customer-facing failure. One in four experienced this more than once. Despite that, 66% already permit production deployment without human review or are building toward […]

84% of AI Pilots Never Scale: The Problem Is Not the AI

The Finding Nearly half of enterprises investing heavily in AI are struggling to see meaningful returns, according to Kyndryl’s own Readiness Report, a finding echoed by a separate IBM study showing that just 25% of AI initiatives deliver expected ROI, with only 16% scaling enterprise-wide. In response, Kyndryl has launched Agentic Service Management, a maturity […]

Only 12–16% of Enterprises Turn Their AI Strategy Into Execution

According to SAP’s Michael Ameling, CPO of SAP Business Technology Platform, 81% of organizations now have a detailed AI strategy, yet only 12% to 16% actually reach AI-driven execution. The gap is not about the quality of AI-generated code. It stems from enterprises underestimating what it takes to operationalize that code: data readiness, system integration, […]