AI and Quality Monitoring: Cash Investigation points the finger at Banque Postale. Rightly so?

In a recent edition of Cash Investigation, a manager from La Banque Postale (whose customer relations department is headed by Sandrine Beltran) reported that the bank was using AI to rate its telephone advisers. Sometimes with biases linked to their accent. TP Interact, a Teleperformance subsidiary or BU, is said to be the tool deployed and misused by some of the bank's managers. True/False?

CallMiner, Teleperformance, AI, what is the use of analysing conversations on call centres. What's new about AI? 
-In the report, it is alleged that a call centre agent, of foreign origin, is given a bad mark by the software, precisely because the software cannot distinguish that he is dealing with an agent who speaks with an accent. The agent would be given a bad mark even though he was very competent’, says the employee who testified anonymously for a department manager. Deployment of the QM tool, enhanced with AI, is due to begin in 2022. 

-The Cash Investigation report, which can be viewed on Youtube, did not prompt a verbal response from the management of La Banque Postale. Instead, the bank replied in writing that the solution deployed since 2023 and 2024 had only been used to train staff, not to rate them. La Banque Postale refused to be interviewed but replied by email: ‘In 2024, La Banque Postale deployed Quality Monitoring software, using artificial intelligence to analyse only 10% of telephone interactions between customer representatives (...) and their customers. (...) It is only used for the day-to-day training of call centre customer representatives, to improve the remote handling of customers and their satisfaction.’

AI and its use in speech analytics and conversational agents are not well known, so they give rise to a number of fears and stereotypes: the example given is perhaps marginal and does not allow us to draw any conclusions, but the factors mentioned suggest that the tool was poorly calibrated and piloted, and that there was a lack of corrective interpretation of the ratings.

In the video that follows, featuring François Chevé (now with Zaion), Dominique Yapi, from Groupama's sales and marketing department, and Emmanuel Simonet (now with Accenture), we discover that conversation analysis software has long been used in contact centres, to ensure that the voice of the customer and any frustrations that may arise are taken into account. It's a shame, then, that François Cardona, the director of the Cash Investigation report, based his demonstration on a single case study, without putting it into perspective.

-In 2016, the installation of Workforce Management software at Teleperformance triggered a flurry of negative articles in the major national media (TF1, Europe 1, etc.), testifying to their very relative knowledge of this industry and the tools deployed and used in it. 

We were led to believe that bathroom breaks were deliberately limited because they were timed.

TP Interact, CallMiner, Sandrine Beltran
Teleperformance has been using and deploying this type of tool with its customers for a long time, and it can also be used to assess the emotions suggested by conversations. The world number 1 relies on CallMiner, a well-known software tool. Various French software tools can also be used to carry out the same quality control operations, sometimes taking into account the emotional tone of a conversation (such as Callity)

La Banque Postale employs 3,000 agents in its call centres in France, which receive 12 million calls a year. In its insurance division, it has equipped itself with a conversational bot, Lana.

En-Contact devoted a profile to the 3639 coach, Sandrine Beltran, who joins from Société Générale. She has just been appointed as a director of the AFRC, as mentioned in our latest mercato section.

True/False
To sum up, and to answer the question of the veracity of this report: there is certainly work to be done and areas for improvement at La Banque Postale. But Cash Investigation based its conclusions on the analysis of a single case, without putting into perspective or studying the many benefits of QM and speech analytics in many contact centres.    

The editors of En-Contact.