Answering inbound calls is defense: the call is coming anyway and you are trying not to miss it. Outbound is offense — you decide who you call, when, and why.
The voice agent's ability to place calls over WhatsApp opens exactly that side up. But like any new channel it has a wrong use: dialing your way through a random number list. Meta blocks that by policy, and even without the policy it would not produce results; a customer bothered on WhatsApp blocks you, and you never win that number back.
The seven scenarios below are the cases where the call is valuable to the customer too.
1. Appointment reminders
The highest-return scenario, because the saving is directly measurable. Clinics, service centers, beauty salons, consultancies — a no-show costs you both that slot's revenue and the customer who could have filled it.
A written reminder may go unread. A call both reminds and completes the reschedule or cancellation in the same conversation. The agent can ask "you have an appointment tomorrow at 2pm, will you be coming?" and update the calendar based on the answer.
2. Quote and pricing follow-up
The quote went out, no reply came back. The task sales teams like least, postpone most, and lose the most revenue to.
The agent calls a set number of days after the quote was sent and learns exactly one thing: where is the decision? If the answer is "the price was too high," that reaches the sales team. If it is "I haven't had a chance to look," the follow-up date is updated. If it is "we went with another vendor," the opportunity is closed and the team stops chasing a dead lead.
3. Unreachable leads
A lead who filled in a form, came from an ad, or chatted on your site cannot be reached. In the classic flow that record quietly dies after a few attempts.
The voice agent has a patience advantage here: it retries at the intervals and counts you define, logs every attempt, and holds the conversation when it connects. We cover how to manage unreachable records with channel sequencing in the 11 August post.
4. Order and delivery verification
Wrong address, missing apartment number, undeliverable parcel. This is the silent source of return costs in e-commerce.
On suspect records the agent calls, verifies the address, asks for a delivery time preference and, if needed, updates the record and passes it to the carrier. Given the cost of a single failed delivery, this scenario usually pays for itself in the first week.
5. Abandoned carts and incomplete payments
There are items in the cart and the payment was never completed. A written reminder is the standard fix, but click rates are low.
On high-value carts a call works differently: the agent asks "did something go wrong at the payment step?" The answer is usually a technical blocker, hesitation over shipping cost, or a payment method problem — all three solvable during the conversation. On low-value carts the call cost may not make sense; you need a threshold.
6. Renewals and subscription periods
Contracts, insurance, maintenance agreements, subscriptions. Calling a customer before their renewal date is many times cheaper than winning them back after they leave.
The agent calls ahead of the renewal date, restates the terms, and starts the process if the customer wants to continue. If they do not, it records the reason — the single most valuable input to churn analysis.
7. Post-interaction satisfaction
Service completed, installation done, support ticket closed. Send a survey link and you get low participation; a short call raises both participation and answer quality.
The agent asks two questions: was it resolved, and how was the experience. Negative answers go to the team as an instant notification. The side benefit of this scenario is catching a bad experience before it becomes a public review.
The permission framework: this channel has rules
This is the part not to skip when designing outbound. Initiating a call over WhatsApp is possible within user permission and Meta's business policies; there is no freedom to dial random numbers. Those policies are also updated regularly.
In practice, the safe framework is this:
- Call on top of an existing relationship. Someone with an appointment, an order, a quote, or an open support request.
- Log the permission. The moment they agreed to be called, through which channel and when, should be visible on the record.
- Keep the exit open. Anyone who says "don't call me" must be able to leave the list in one sentence, and that preference must persist.
- Cap the attempts. Calling the same person repeatedly is the fastest way to burn the channel.
- Set time windows. Late evening and early morning calls create irritation no matter how legitimate the reason.
- Check Meta's current documentation. Permission requirements and call initiation conditions can change.
Measurement: how do you know which scenario works
An outbound program that is not measured quickly turns into pointless cost. Four numbers to track per scenario:
- Reach rate. In how many calls did the other side actually talk?
- Completion rate. How many conversations finished the intended job — appointment confirmed, payment completed, renewal signed?
- Value per call. Value of completed business divided by total calls placed.
- Negative reaction rate. How many people were annoyed, blocked you, or asked to be removed?
The fourth number matters more than the other three. A campaign that looks good in the short term while raising your block rate is damaging the channel itself.
If you have decided which scenario to try first, appointment reminders are the fastest start: measurement is clear, the benefit to the customer is beyond dispute, and the risk of designing it badly is low. You can start by hearing the voice agent on the demo page.