Automated Bank Call Disclosures: What Mars Climate Orbiter Taught About Clear Openings

An automated bank call should identify the bank, say that the caller is an automated voice agent, state the purpose of the call, and give the customer a clear way to decline or opt out before asking for information. In Twi or English, those disclosures belong in the opening seconds, when the customer is deciding whether the call deserves their attention.

In September 1999, NASA lost the Mars Climate Orbiter as it approached Mars. The investigation found that software from Lockheed Martin had supplied measurements in pound-seconds while NASA’s navigation system expected newton-seconds. The spacecraft was lost because two connected systems carried different assumptions about a basic piece of information. NASA documented the failure in its Mars Climate Orbiter Mishap Investigation Board report.

A bank call has lower stakes, but the same discipline applies. The customer should not have to infer who is speaking, why the call arrived, or whether the voice is human. State the important information early, in language the person can understand.

Give the caller enough information to make a choice

The first line should name the bank or its authorised service provider. A customer who receives an unexpected call needs a recognisable identity before they can assess the request.

The next line should say that the call is automated. Hiding that fact creates confusion from the start, especially when a natural Twi voice sounds conversational. A good voice experience earns attention by being clear about what it is.

Then state the reason for the call in plain terms. “This automated call is about your account service request” gives the customer more context than a vague request to continue. Avoid asking for account details, PINs, passwords, or one-time codes in the opening. The purpose should come before any request.

If the customer does not want to continue, the agent needs a direct path out: an option to end the call, request a human channel where available, or record an opt-out. That choice needs to work in the same language as the greeting.

Make Twi disclosure understandable, not decorative

Twi language support changes the quality of the disclosure. A customer who is more comfortable in Twi should be able to hear the bank’s identity, automation notice, purpose, and exit option in Twi, rather than receive a polished Twi hello followed by critical information in English.

That requires more than placing a translated sentence at the top of a script. The agent needs to recognise a Twi response, including hesitation, refusal, or a request to stop. It also needs to handle a customer who switches between Twi and English.

Asenda Talk is built around native Twi speech recognition and synthesis, fine-tuned in-house, alongside Twi and English agent configuration. That gives teams a way to evaluate whether the disclosure is understood as spoken, not merely whether the text reads correctly on a screen. The platform remains in early access, so teams should test these opening flows with real review and monitoring before relying on them in customer-facing operations.

Treat consent and refusal as call events

A disclosure only helps if the system can show what happened next. Did the customer continue after hearing the purpose? Did they refuse? Did they ask to be removed from future calls? Did the call end before the disclosure completed?

Those outcomes should be recorded as part of the call lifecycle. Otherwise, a team investigating a complaint has a recording or a summary, but no reliable account of whether the person was told what the call was and how they responded.

Asenda Talk includes consent, opt-out, and audit-trail controls for every call, as well as a telephony lifecycle webhook pipeline for tracking call truth. Those controls matter most when a customer says they never approved contact or never understood what the call was about. The investigation should start with the exact call record, the opening script version, the contact’s consent status, and the resulting opt-out state. [This consent complaint investigation]( /blog/consent-complaint-investigation-what-an-audit-trail-reveals-before-calls-resume-d9e31c91/) shows why that evidence should be available before calls resume.

Build the opening as a controlled decision point

For a bank team, the opening should be treated as a small decision flow:

• Identify the bank or authorised caller.
• Say that the voice is automated.
• Explain the specific reason for the call.
• Offer a clear way to decline, end the call, or opt out.
• Record the customer’s response and preserve the script version used.

The Mars Climate Orbiter failure was traced to a mismatch that should have been explicit at the system boundary. A customer call has its own boundary in the first few seconds. The bank knows why it called; the customer does not. The opening disclosure is where that gap closes, or where trust starts to disappear.

Outbound calling in Asenda Talk is currently gated behind an explicit telephony-provider decision and is not live by default. Until that decision is made, teams can use the platform’s agent configuration and call-control design to prepare, review, and test the disclosure flow before any paid calling is enabled.