Every dynamic variable in voice agent needs a fallback.

Follow me on LinkedIn - AI, GA4, BigQuery

I added a small line to a voice agent prompt last week. It looks like this.

**Fallback:** If the value above shows as a literal placeholder, 
say "Let me check that in a moment — first, can I get your details?" 
and continue with the next uncollected field. Do not return to the time question.

The line sits underneath the section where the agent's current date and time are injected via a Retell dynamic variable.

Current local date/time in Florida: {{current_time_America/New_York}}

The variable resolves at runtime to something like "Monday, May 18th, 2:30pm."

Most of the time.

Sometimes, for reasons that are not always obvious, the variable does not resolve. The placeholder stays in the prompt as a literal string. The agent reads it as part of the prompt text and tries to speak it aloud.

Without the fallback, the agent says something like "the current time is curly brace curly brace current underscore time underscore America slash New York curly brace curly brace." Which, to put it generously, is not a great moment in customer service.

With the fallback, the agent deflects briefly and moves on. The call continues. The caller never knows anything broke.

What dynamic variables are, and why they fail.

Voice agent platforms like Retell let you embed dynamic variables in your prompt. At runtime, the platform replaces these variables with real values.

Current date and time, caller information from the CRM, custom data passed in from the dialler, and so on.

The standard syntax is two curly braces around a variable name. The platform reads the prompt, finds the variables, substitutes the values, and passes the resolved text to the language model.

Substitution can fail for several reasons.

The variable name might be misspelt.
The platform's variable resolver might be temporarily unavailable.
The value the variable points to might not exist for this particular call.
The version of the platform's runtime might have changed in a way that breaks specific variables.

When substitution fails, the curly braces stay in the prompt. The language model treats them as regular text, to be read aloud or interpreted literally.

This is the worst possible failure mode. The agent does not refuse to respond. It does not throw an error. It just confidently reads gibberish to your caller.

Every dynamic variable in a production prompt is a potential failure point.

The substitution is silent when it works and silent when it fails. The agent has no built-in way to determine whether the value it reads is a real value or a literal placeholder.

The fallback pattern fixes this. For every dynamic variable that drives critical behaviour, include a short instruction telling the agent what to do if the variable does not resolve. Use a deflection that matches the prompt's existing tone. Tell the agent to continue with the next step.

Most prompts I review lack fallbacks for dynamic variables. The assumption is that the platform will always resolve them correctly. When the platform fails, the agent fails publicly to the caller.

Every place your prompt assumes a runtime value, ask yourself what happens if that value is not there. If the answer is "the agent says something embarrassing to the caller," you need a fallback.

The fallbacks themselves are short. Usually one or two lines. They cost almost nothing in tokens. But they cost a great deal, where the agent reads curly braces aloud to a confused caller.

Every dynamic variable in voice agent needs a fallback.

What dynamic variables are, and why they fail.

Substitution can fail for several reasons.

Every dynamic variable in a production prompt is a potential failure point.

Other Articles on Voice AI.

Himanshu Sharma