Could the new wave of generative AI tools be used by telemarketers to circumvent to UK’s ‘PECR’ rules on automated calling?
Direct telephone ‘cold calls’ are sometimes seen as an intrusive method of contacting prospects, and customers tend to feel particularly aggrieved when they find themselves in an unwanted conversation with a recorded message who can’t even respond to them. As a result, the Privacy and Electronic Communications Regulations (EC Directive) 2003 (PECR) quite understandably prohibited the use of ‘recorded matter’ for direct marketing purposes by means of an ‘automated calling or communication system’, except where the person obtains the consent of the individual they are contacting. It was, for all intents and purposes, a ban on using automated ‘robots’ to make cold calls.
However, where a marketing telephone call is being made by a live voice, that pre-requisite for consent does not exist. So as most marketeers know, hiring someone to call potential customers directly is, legally speaking, an easy way to prospect and to generate leads – callers don’t need to have had any pre-contact with that customer and can rely on bought in marketing lists and customer data they have previously collected from other means. Of course, telephone numbers will still need to be scanned against those listed with the Telephone Preference Service, and any customers who have previously asked not to be contacted, but as a starting point, picking up the phone to a potential customer is a clean, if old fashioned, method of lead initiation.
The difficulty with this marketing strategy lies in the costs that are involved in having a team of people who can do this picking up of the physical phones. While you can automate the process to some degree, companies still find themselves with the often significant overheads involved in employing, managing and training a team of people do carry out the task.
It is no wonder then that innovators around the world have come up with a potential solution – leveraging the power of generative AI (i.e. the type of AI that powers services like Chat GPT) to enable voice calls to be run by AI driven software. Caleb Maddix, Co-Founder of Air.ai announced the launch of AIR on July 15th on Twitter, posting a demo of a sales conversation between a potential customer who had gone through a purchase journey but hadn’t, in the end, purchased a product. The AIR software reaches out to the customer and engages in a sales pitch to determine why the customer hadn’t bought the product, and proposing some solutions to encourage them to purchase.
The customer does not appear to either notice or care that they are in conversation with AI software rather than a human being, and interacts as they would with a flesh and blood operator. Notably, the AI software mirrors the tone of language and mannerisms exhibited by the customer, using phrases like ‘yeah no worries’ and ‘I feel you my man’, in response to the customer’s phrases such as ‘I’m not interested man’ and ‘uh fine shoot what’s up?’.
Its not hard to see the benefits of these kinds of tools for lead generation businesses: increased volumes of calls, where you can rely on a bot to use the full range of sales techniques available, tailor their language, accent and level of formality to match that of the customer, and therefore deliver a high quality, personalised sales pitch, all for a fraction of the price.
But can these tools work in the UK, or has PECR been drafted with sufficient ‘foresight’ to block these innovators before they can thrive in the UK market? Are these tools ‘automated calling or communication systems’ which use ‘recorded matter’ to communicate with potential customers.
Regulation 19(4) defines an automated calling system as ‘a system which is capable of (a) automatically initiating a sequence of calls to more than one destination in accordance with instructions stored in that system; and (b) transmitting sounds which are not live speech for reception by persons at some or all of the destinations so called’.
So, an automated caller is one that both automatically dials numbers and then transmits something which is not ‘live speech’.
I think, for now, it is not unreasonable to assume that these tools will automatically dial numbers and that the calls will be made based on instructions inputted into the system by the operators of that system. However, as the software develops, it is quite possible that it will be able to take its own initiative about who to call and when – it will develop its own dataset that it relies on to make these choices.
We also need to consider whether or not the voice responses that are generated are ‘live speech’ or ‘recorded messages’. PECR doesn’t provide any definitions for these phrases, so we need to consider them based on their general definitions. It is true to say that the answers the AI bot will give are not pre-recorded statements based on set prompts; the voice is articulating a response to the specific questions and answers given by the customer on the receiving end of the call. It is tailoring its responses to those given by the customer live. However, in order to generate a response, the AI software needs to interpret the words said by the customer by translating them into text, process that text, understand the context find an appropriate response, and then translate that textual answer into speech. The tool will therefore likely be relying on a resource pool of pre-recorded sounds that it will use to translate the response into speech. On balance however, while each sound may be pre-recorded, the full messages provided in response are not, they are generated live in response to the customer.
Consequently, it seems that AIR ai, and other generative voice AI tools, may well succeed in circumventing PECR on this point and could present a tool to revolutionize marketing and lead generation strategies within the UK.