PhoneĀ a call centre and the person on the end of the line will often try to sell you something. Now it seems this upselling is fuelled by artificial intelligence.
, an AI firm based in Washington DC, claims to be able to pair call-centre agents with the people they are most likely to be able to influence, based on a prediction of how both will act during the call.
Many large firms around the world, including Virgin Media andĀ Sky, use Afinitiās software to manage millions of calls a year. ItĀ also works with healthcare providers, banks and insurance companies, and claims to have increased revenues across all of itsĀ clients by 4.3 per cent, or many billions of dollars.
Advertisement
The AI adapts to the needs of a particular company, says Afiniti CEO Zia Chishti. For an internet provider, it aims to reduce the chance of a customer cancelling their contract. Or if someone owes money to a bank, the AI aims to increase the amount they repay.
The systemĀ draws on whatever data each company has about callers. This can include how long someone has been a customer, what they pay per month, when they last called and how that call went. A telecoms company might also know about a customerās internet and TV use.
All your data
This is merged with data bought from information brokers such as Experian and Acxiom, which can include details about where someone grew up and went to school, where they work, their credit rating and any big purchases they have made.
Agents are classified based on the outcomes of their previous calls. The software even adapts to changes in the weather that could affect peopleās behaviour.
To measure its impact, the software turns itself off for short periods throughout the day. Afiniti is then paid a percentage ofĀ the revenue increase its system provides when running.
Because the match-making takes place in a fraction of a second, callers donāt notice it. ButĀ Michael Veale at University College London thinks that callers should be informed, at least in the European Union.
He notes that the use of certainĀ personal information is particularly restricted under the European General Data Protection Regulation (GDPR). If companies want to infer your ethnic background, religion or political views, for example, they are meant to get your explicit consent, he says.
Even if personal information such as ethnic background isnāt used explicitly, there is still a risk that an AI is inferring it from other data and using it to make a decision. āMany companies seem to believe inferring sensitive, private data is different under the law from collecting it,ā says Veale. āData protection regulators increasingly believe and vocally argue that it is not.ā
The software is similar to a recommendation system, says Nello Cristianini at the University of Bristol, UK, but it may go further. āIt is not always clear where the recommendation stops and the nudge starts,ā he says. āI am not sure that current laws have incorporated this new situation.ā
Afiniti insists that its data use isĀ above board. āWe take our data and security-related obligations seriously, and believe we are fullyĀ GDPR compliant,ā says a spokesperson for the company.
The firm now wants to use its technique to plan home visits by technicians. āIn someoneās home, youāre face to face and the power of persuasion is much greater,ā says Chishti. āThatās a very powerful psychological domain for us to optimise.ā
This article appeared in print under the headline āAI is deciding who answers your phone callā
