The Paper List Abena Had to Settle Before Parcels Went to the Wrong Place

Keypad campaigns work best when the caller needs to make one simple, trackable choice. WhatsApp follow-ups work best when the person needs time, a written record, or a link, while conversational voice agents fit calls where questions, language switching, and follow-up decisions change the outcome.

At 4:40 p.m. in Kumasi, Abena was holding a paper list of customers whose delivery details needed confirming before the next morning. A keypad call could ask each person to press 1 to confirm or 2 to speak to someone. But one caller wanted to change the drop-off point, another asked whether a family member could receive the parcel, and a third moved from Twi into English as soon as he tried to spell a landmark.

The list had to be settled before the courier route was planned. If Abena guessed at unclear responses or left them unresolved, parcels could go to the wrong place and customers could spend the next day chasing them. A simple keypad flow would collect clean answers from the easy cases, but it could not safely handle the exceptions piling up at the bottom of her page.

Choose keypad campaigns for fixed decisions

A keypad campaign is a good fit when the question has a small set of valid answers and each answer leads to a known action.

Think of appointment confirmation, event attendance, payment reminders, delivery confirmation, or a request to opt in to a callback. The caller hears a short message in the language they understand, then presses a number. Your team receives a clear result: confirmed, declined, or needs a person.

This format earns its place because it limits ambiguity. It also makes reporting easier. If the real business decision is binary, a long conversation can create more failure points than value.

Before choosing keypad, test the question against three checks:

• Can the caller answer without explaining their situation?
• Does each possible response have an approved next step?
• Would a wrong or unclear answer create material risk?

If the answer to the last question is yes, build a path for human review. A caller who presses the wrong key while trying to change an address should not be treated as a confirmed delivery.

Keep consent and opt-out handling visible in the flow. A campaign result is useful only when you can show who was contacted, what happened, and how future contact preferences were recorded. What Happens When a Contact List Includes Missing Consent and an Opt-Out? explores why that record matters before calls begin.

Use WhatsApp when the customer needs something to keep

WhatsApp works when the next step depends on information the customer should be able to reread. A map pin, order reference, document checklist, payment instructions, product photo, or rescheduling options are easier to act on in writing than through a one-time voice call.

It also gives people breathing room. A parent in traffic may see a follow-up later. A shop owner may need to check stock before replying. Someone who does not want to speak can still respond with a short message.

The mistake is treating WhatsApp as a universal replacement for voice. A written message may sit unopened when the matter is urgent. It can also create a long back-and-forth when the customer has a question that would take 30 seconds to resolve aloud.

For Abena, WhatsApp became the right follow-up for callers who needed to send a location pin or review the delivery reference. It gave them a record they could share with whoever would receive the parcel. It did not solve the calls where the customer needed to explain an exception in Twi and get a useful answer before the route was finalized.

Bring in a conversational voice agent when the call has branches

A conversational voice agent earns its place when the script is only the beginning. The caller may ask a question, correct a detail, switch between Twi and English, or need the system to decide what happens next.

That is where language quality matters. A voice experience that forces a Twi-speaking caller into English can turn a routine support task into a dropped call. The same applies when a caller begins in Twi but needs to say an order reference, name, or technical term in English. What Happens When an Order Reference Switches a Twi Support Call to English? covers the handoff problem in more detail.

Asenda Talk is built for this category of workflow. Teams can configure an agent’s persona, first message, and voice, with in-house Twi speech recognition and synthesis alongside the assistant runtime. Its call lifecycle pipeline is designed to track what happened in a call, while consent, opt-out, and audit records remain part of the operating model.

The platform is in early access. Outbound calling remains gated behind an explicit telephony-provider decision that is not live, so teams should not plan a production outbound rollout around it today. The practical work now is to identify the conversation paths worth testing, define escalation rules, and decide what evidence an operator needs after each call.

Design the handoff before choosing the channel

Abena did not need one workflow for every name on her list. She needed a decision rule.

Confirmed delivery details could go through a keypad campaign. Customers needing a reference or location exchange could receive a WhatsApp follow-up. The cases involving changed instructions, uncertain language, or a question outside the script belonged in a conversation flow with a clear human escalation path.

The next morning, her paper list was shorter for a better reason. Each unresolved customer had a reason beside their name: send location, call back, or route to support. That is the standard to use when selecting a channel. Choose the workflow that leaves the team with the next safe action, not merely a completed contact attempt.