• 03-31-2021

  • 8 min read

  • Anna Prist

Who You Gonna Call? Outbound Calls 101

This article was originally published at Chatbots Life.

When a business considers a voice bot simply to do cold calls the way most sales reps do, things may turn very nasty. Usually, clients slam the phone and swear off useless bots. And they probably should, because nearly 50% of U.S. mobile traffic are scam calls. You know, all that Apple tech support, fake payment reminders, etc.

But is it possible to build up a voice bot for outbound calls that would be of benefit for businesses and their customers too? How to determine the purpose of the call and not drop a clanger when picking the builder? Well, let’s get this sorted out.

Don’t call me again! Looking for a use case

Telemarketing is not the most obvious use case for a chatbot. Moreover, it’s not the best one. And the issue is not that robots can’t sell — they do. It’s just that sales require:

  • Special expertise (both vendor and client)
  • NLU (Natural Language Understanding) tech deep dive
  • Solid experience in dialog processing with AI
  • Logs precision
  • Hypothesis testing

SMBs use outbound calls for the day-to-day goals — to save resources and time on outbound communications and to widen customer service. A bot here can automate tasks related to marketing, delivery arrangements, logistics and maintenance, HR services, upsale processes. So which use cases are encountered in practice?

Order confirmations, reminders

How long does it take for a receptionist in a small dental clinic to call all tomorrow’s clients with a reminder? For instance, there are 4 units in this clinic, and 40 visitors per day. A receptionist spends 3–4 minutes for each of them. While a bot can ring up 40 clients in 5 minutes — and all that with report generation and info entering into a database.

Another quite high-demand use case here — big event reminders. Informing, like, 300 event participants would take a whole day. And a bot can do that in a few minutes.

NPS (Net Promoter Score) surveys

Customer satisfaction surveys are very popular, but it’s a tiresome process both for a phone agent and a client. It is wearisome to rate services on a scale of 1 to 10, it is irksome to ask these questions the whole day, and it is almost impossible to withstand this survey using a push-button bot.

NLU engine is the thing that distinguishes an intelligent chatbot from a simple script. An AI-powered bot with an NLU engine makes it possible for a person to talk to a bot in a free-and-easy form. For instance, in case if instead of a 1–10 number you’d say “service sucked”, a bot would understand your answer and kindly ask what made it such a bad experience.

HR services

When a company hires massively, a bot can take upon oneself a primary screening. It would use concrete characteristics or ring-up candidates to specify the date and time for an interview.

What’s more, a bot can conduct first-round interviews with low-skilled employees. It can introduce the company, working environment, and corporate culture. A bot can answer model questions about position, duties, and salary. It can take polls and tests, and keep the results within the internal system. It can do all kinds of staff.


A bot is a helping hand when working with a customer base. Especially when a client uses your service with unfailing regularity. All those healthy food services, contact lenses for the next month, car tires on-season, etc.

A bot is proactive. It would call your client, remind that winter is coming and it’s just about the time to change tires. It would ask the vehicle model and wheels, specify the time of your convenience and offer you a discount. And these actions will retain your customers.

When choosing a case, align oneself with the chief criterion — a call situation is inevitable for a customer. They should know you might call them anyway, and as long as they do, it is beside the point whether they talk to a bot or a person.

Wait, why do you even call?

Today businesses can use any channel. And the choice between a call and a message here totally depends on the situation. It might be the client’s preferment (in case they asked not to call because their kids are sleeping), or a time of the day (you are sleeping while there’s a middle of the morning in Europe), or the importance of the message (a client is asleep, but what if they’re going to miss their discount?)

Texting only wins when your information is hard to comprehend aurally — numbers, percentage terms, complicated conditions, and questions that require closer attention to security data.

Besides, a text message in a messenger does not guarantee feedback and business cannot be sure that their client was informed indeed.

Outbound calls for SMBs are good not just because it saves time to communication. A bot can do different tasks and it won’t forget anything — your clients will get the information on time, while you and your employees can carry out more complex and interesting tasks.

Don’t hang up! A chapter about UX

Robocalls. People hate robocalls! In the US robocalls become pandemic ­– according to YouMail, this September Americans received 4.5 billion robocalls and there would probably be a total of 60 billion calls to U.S. consumers this year. Just to compare: 48 Billion robocalls were made in 2018 with 37% of them were driven by scams. In 2017 there were 30.5 Billion calls.

Still, people do not block all the unknown numbers because they believe they might miss something really important. Even if they know it’s a robocall.

So, an honest business with a need to reach its customers has to:

  1. Find the right use case for a bot
  2. Cut through the flood of spam
  3. Not to lose customers’ trust and loyalty with a poor scenario

First off, you got to remember that an AI-powered intelligent bot is very different from a robocall bot with prerecorded speech.

You can actually build up a dialog with an intelligent bot

And the more though-out your bot’s UX, the more successful your outbound calls would be.

Here are some helpful hints on proper user experience:

  • Try to break the tension with a welcome remark that would immediately answer the question “why the hell they’re calling me?”
  • Avoid bureaucratic language. Let your bot speak in a free and friendly human voice (you may choose the style that suits your business). This speech is easier to comprehend aurally, and people react more interested. For instance, do not start with the phrase “Good afternoon Sir/Ma’am, this is an automated message from a florist shop Seed Bed. The purpose of this call is to get your opinion on our service quality”. Keep it simple: “Hi! This is robot Blossom from your favorite florist shop Seed Bed. You recently bought a nice bouquet in our shop and we wonder whether your experience was good”.
  • A voice bot is not supposed to ask for too much information from a client, it just specifies details. Ask closed yes-no questions — this way you will lead a client to the next move. But, if you want to ask your client to give a detailed answer, therefore it will be saved as a comment. This comment is not supposed to sway the whole dialog script. For instance, during the latter stages of an NPS survey, you may ask your client to express themselves freely — let them say what they love and hate about the service provided. This answer will be saved and it won’t affect the bot’s further actions. Now a bot has to show some respect — thank your client for the response and continue with questions.

Note that legislation in some countries prohibits recordings without sanction. In case you still want to do this in a country like that, it is recommended to apply user confirmation request for recording

  • Outwatch the length of the utterances and the duration of the talk. Test your scenario on yourself and your work fellows. People do not want to spend a lot of time on a phone call when talking to strangers. They’ve picked up a phone — they need to know the reason for this call, they want the problem to be resolved, and they are willing to get back to their business.
  • Don’t hesitate to notice that a person you are calling is talking to a bot. Our experience in contact centers automating shows that people talk to bots in a clear and forgiving way, they also speak in short. But to make your speech synthesis sound more human, use the voice tone settings — some platforms have this feature.
  • Be aware, a client might ask you to call back later or they might want to call you back themselves when it’s more convenient for them. So, offer that opportunity! A bot should apologize for the disturbance, and the platform has to register the reminder to call the client back later.

Checklist. What should you pay extra attention to

There are dozens of products for independent voice bot development that do not require coding skills. It seems like they are easy to understand, but the devil is in the details.

We’ve made a list of these details to keep you from making common mistakes.

Tariffing. How does it work

Sometimes vendors do not point out that calls are charged by the minute. Although it’s not the most beneficial option — the more flexible tariffing is, the less costly outbound calls are.

Aimylogic interface

What’s included in the tariff

Usually, vendors work with a fee per minute and the prices are pretty much the same worldwide. These fees include taxes, workplaces, and other OpEx characteristics. Note that real phone agents might have an uneven distribution of work ­– you have to pay for work anyway, even when there are no calls at all.

When using bots, you pay as you go. Besides, the cost of a minute is much cheaper. But here’s the catch: this alluring low cost doesn’t include basic tech and tools. Make sure to specify whether the tariff includes ASR (Automatic Speech Recognition) and TTS (and Text-To-Speech), logs playback, phone connection, and phone number acquisition.

Integrating external IT systems

A bot can upload name and other data from the CRM system, it can base the info into the customer data sheet during the call, and it can create necessary entities in CRM systems (leads, dealings). It gives you deeper automation and greater virtue from the outbound call.

The flexibility of a solution

Not all users are willing to take a call from an unknown number. Just deal with it. But if the platform you are using makes it possible to bring in other channels (e.g., messengers), you will definitely reach out to your customers.

It is convenient to adapt the phone scenario to a text bot in a messenger in case you want your NPS survey to embrace your customer base to the maximum.

The product connection charge

Check over the first connection charge — when you are paying for the product, even if you are not even using it. In case it happens to be, it should be reasonable and beneficial.

Availability of ready-to-use solutions

It’s very handy when a product has business case templates (delivery bot, NPS bot, etc.) — they do simplify the work. If a product has pre-trained scenario blocks, that can provide a smart and quick-witted bot with minimal effort — that’s even better.

These blocks are the built-in intents that can be used as bricks to build up a scenario: agreement intents, call to agent transfer, goodbyes, responses to expletives. These bricks occur in any dialog — it’s cool when you don’t have to do this entire job by yourself.

Scenario development. Tech capacities

Out-of-the-box cases are fine unless you want something more. Some products make it possible to create scenarios using different programming languages.

As a result, users can employ more complex cases. So, you can teach your bot to specify the appropriate date and time for a medical office visit, automatically check whether it’s available, and quickly reserve it.


Okay, the dialing is over, 30 clients said YES, 10 asked to call back later, 20 said NO, and the other 20 said NO in a rude manner. How do you figure that out?

The platform should be able to show stats for every call — when it started, when it ended and with what effect

Also, you could benefit from the opportunity to listen to the call and save the recording to the external resource (like Google sheets) where you would see the call transcription. All analytics should be compiled online, so that if anything crops up you would notice the friction points in a scenario.

Sure, many say outbound calls are created in 3 minutes and you need no specialized knowledge to do it. Nevertheless, this knowledge is essential and 3 minutes is not enough (in case you really want to do it right).

Decided to make your first outbound call? Find out whether the vendor lets in on training courses and product tutorials. Study out how the tech support works — it might be very helpful in the beginning.

Besides, there may rise issues in the scenario itself, speech recognition, telephony, and write-offs. But if you have a trustworthy second-guessed partner, your performance will be strong and your customers will be happy.

Sign up for our newsletter

Get the latest news and updates, fresh analytics and trends, expert interviews and success stories, numbers of the month, and more.

Join our partner network!

Please tell us about yourself and we’ll get back as soon as we can.



Work email

Work phone

Company name

Your question

Please fill in the form
How can we help you?

Please fill out the form, type your message and we’ll get back to you as soon as we can.



Work email

Work phone

Company name

Your question

Entre em Contato
Você tem perguntas?

Por favor, preencha o formulário e nós retornaremos assim que pudermos.



E-mail de trabalho

Telefone do trabalho

Nome da empresa

Sua pergunta

Thank you for reaching out!

We appreciate you contacting Just AI and will get back to you as soon as we can.

Obrigado por estender a mão!

Agradecemos o seu contato e entraremos em contato o mais rápido possível.

Thank you for reaching out!

We appreciate you contacting Just AI and will get back to you as soon as we can.