AI Outbound Dialing Explained: Feasibility & Compliance Issues

Generative AI has advanced in leaps and bounds over the past few years, and has opened up a realm of exciting opportunities for contact centre teams.

Using technology such as GPT-4 and other large language models, it’s now possible to have a conversation with a chatbot that almost feels like speaking with a real human.

What’s more, speech synthesis solutions can now generate human voices that sound almost indistinguishable from the real thing. As a result, it’s now possible to have an at least semi-realistic voice conversation with an AI.

For innovative contact center teams looking to improve efficiency, or make large volumes of outbound contact, this raises the question: can you use artificial intelligence to automate the initial part of the conversation with a customer?

In this article, we’ve explored some of the key things you’ll need to consider if you’re exploring using AI voice to help make outbound calls more efficiently.

Compliance issues

One of the biggest issues you will need to overcome when making automated outbound contact is how you’ll ensure compliance with the relevant legislation in the countries you are dialing into.

Automated dialing compliance

AI compliance issues.

Most jurisdictions do not outright ban automated phone calls, but many regions have specific rules in place that you must follow. For example:

  • In the US, under the Telephone Consumer Protection Act (TCPA), telemarketers and debt collectors cannot make automated outbound calls without prior express written consent. For other types of calls, or instances where the customer has provided permission to receive automated contact, the FCC states that there must be the facility available for customers to opt out of receiving future calls throughout the conversation, and the customer must be made aware of this at the beginning of the call. There is also a three-call limit that applies to each residential phone number, meaning that you cannot make more than three automated calls to a given number within a thirty-day period. This is in addition to other restrictions placed on outbound dialing under the TCPA.
  • In the UK, Ofcom states that automated calls must only be made to people who have given permission to be contacted. Also, if the AI has the ability to connect the customer to a human agent, when this occurs, the customer must not wait an unfairly long amount of time for the connection to be put through.
  • In Australia, automated calls do not have specific restrictions, but are subject to other ACMA rules – for example, when the call starts, you must state who you are calling from, and the purpose of the call. Also, under the Spam Act, you must give people an easy way to unsubscribe from further calls.

Before investing in an AI-driven outbound calling workflow, it’s important to consider your outbound dialing compliance obligations as the first step in assessing the viability of the project.

While these regulations might be considered relatively straightforward in countries like Australia, the same isn’t true in the United States, where the FCC has implemented specific rules that you must follow when making AI outbound contact.

Privacy laws and customer data

Apart from compliance issues involved in making automated outbound contact, you will also need to consider how you will comply with privacy legislation more broadly.

You may need to use customer conversations to train the AI, otherwise it might not be able to perform at its best. Without this level of training, the AI might not be prepared to handle the specific nuances of the conversations your business has with its customers.

Some of your compliance obligations might be met by having the AI inform customers that the call will be recorded for quality and training purposes.

However, there are other issues you might need to consider:

  • In some jurisdictions, especially those covered by GDPR, customers have the right for their information to be forgotten. This may not be as simple as deleting their CRM record and their call recording if the AI has already ingested this information.
  • Do you have the legal right to use customer data to train the AI? Using a call recording that has been collected with a disclaimer may be permissible, but using further customer data to improve the quality of the conversations that the AI is having might introduce compliance risk.
  • In some jurisdictions, you might only have the legal right to retain customer data for a specific length of time. If you use customer data to train the AI, and the AI retains this data in its processed state indefinitely, does this introduce compliance risk?

If you are concerned about the compliance issues involved in making automated outbound contact, you may find that AI is better suited to handling inbound calls instead.

If the customer has a clear purpose for calling you, not only will you be able to minimise compliance risk – you might also find that the AI is better suited to handling the task at hand. For example, taking a payment, or retrieving information for the customer, are both tasks that AI will generally perform well at.

Technological and feasibility issues


Apart from compliance issues, there are also a number of technological and feasibility concerns you will need to address before utilising an AI agent for outbound dialing.

  • Latency: how quickly will the AI be able to return a response to a customer? This is especially important when calling internationally. It could be that the person being called will wait three seconds or more for the AI to respond after they finish speaking, which could affect the customer experience.
  • Transcription issues: can the AI reliably understand what the customer is saying, especially given the audio quality of the call? Customers may become frustrated if they have to make an extra effort to enunciate words more clearly for your AI. Even humans sometimes struggle to understand what customers are saying, which is why we still rely heavily on DTMF tones to capture important information.
  • Customer experience concerns: will your brand credibility suffer if customers realize they are speaking to an AI? Depending on your industry, your company’s reputation, and the type of calling you’re doing, some customers may take issue with receiving a “robocall”, no matter how intelligent the virtual agent is.
  • Controlling the AI: what if your virtual agent goes rogue, and does not stick to the messaging you have prescribed it? There have been documented instances where AI chatbots have declared romantic interest in the user, despite having controls in place to prevent this type of behavior, for example.
  • Making the most of your data: with traditional progressive or preview dialing, your agents will have the opportunity to learn about the customer before the call, and formulate their messaging accordingly (especially if you’re using contactSPACE CallGuide workflows). When using AI, how will you take advantage of this customer data? If you have a lot of useful information available about the customer, it’s important that you are still able to get value out of this data when using AI, otherwise your calling performance may suffer.
  • Agent workflows: what tasks will agents be doing to stay productive while AI calls are taking place, that could easily be interrupted with a cold transfer?

You may find that AI outbound dialing does not provide a significant uplift in productivity or other results you want from the call if you still need to have a number of agents on standby ready to connect at a moment’s notice. You will need to consider how many agents you need to deliver a quality customer experience – this will depend on what other work your agents perform, and how often calls need to be connected to a real person.

You will also need to consider how you will hand off conversations to real agents, including in situations where the AI stops functioning correctly. And when the handoff occurs deep into the conversation, how will the human agent be provided the context of the call thus far?

Conclusion: can AI help improve outbound dialing performance?

Uptrending graph.

For certain use cases, AI can potentially help to make outbound dialing at scale more efficient today.

For example, if you want to remind customers to do something, or want to collect some basic information from them (especially where DTMF tones can be used), AI might be able to help you do this more quickly.

However, for telemarketing campaigns in particular, compliance issues can make using automated calling tricky, depending on the region you’re dialing into.

While AI is rapidly becoming more advanced, and will soon be able to handle much more complex customer interactions, the limiting factor for this technology may be how you’ll ensure compliance with laws designed to target robocalling, as well as privacy legislation.

You might find that the compliance risk of deploying AI at scale outweighs the potential cost savings of using this technology to make large volumes of outbound calls.

This is especially true if you will often need to transfer AI calls to real agents, meaning that you will still need to have large numbers of real agents on standby to answer these calls, reducing any productivity or efficiency gains you will see from using AI agents.

contactSPACE is investing in this exciting area of technological innovation. We can help you assess the best technologies to use for AI outbound dialing, in order to get the results you need while minimizing the cost and risks involved.

To learn more, and begin exploring AI-powered outbound, get in touch on our website, or call 1300 360 553 in Australia or +1 (415) 200 3752 in the US to speak to our team.