Enhance our advisors’ performance using AI: this is the objective of the partnership between Allo-Media and Webhelp. In this first progress report, we discover the partnership’s challenges and initial successes as told by Mathieu Jougla, director of business development of the Technology Enablement team at Webhelp.
A customer and an advisor talk on the phone: how can we leverage AI to help this agent, in real time?
This operational issue is the basis of the partnership work between Webhelp and Allo-Media, a start-up specialized in speech recognition and supported by The Nest by Webhelp, the customer experience accelerator.
This project, which began in 2018, is based on a common belief that AI should not only be used to automate conversations (as many chatbots do), nor to replace agents (in the short or medium term), but rather to enrich and facilitate interactions between the agent and the client.
In this project called “Self-Augmented Agent”, each partner plays a very distinct role:
- Allo-Media brings its expertise on the transcription and understanding of human language and develops the technological brick,
- Webhelp brings its expertise in customer relationships, the integration of technical solutions into CX IT environments, and its business know-how.
First workshops: collecting requests from agents
It is in October 2018 that the 1st workshop was organized, in the Webhelp call center of Montceau-Les-Mines.
The idea was to listen to a dozen advisors to gather their real needs. The advisors, all volunteers, had varied and representative levels of experience.
Around the table were also several project managers and data scientists from Allo-Media, technological experts from Webhelp, and data science specialists such as Mathieu Caron from gobeyond France (Webhelp).
Very quickly, the agents specified that they certainly do not need an emotional analysis tool that would tell them that “the customer is getting impatient” – they already know it!
They defined their priority need: when the call does not go well, or the customer is not interested, they want to be helped to re-capture their attention and interest.
At the end of the workshop, the Webhelp and Allo-Media teams defined the use cases to which the augmented agent project will have to respond: assistance in selling a product or service, for example.
First technical tests: the transcript is a success
Before building the features of the Self-Augmented Agent, it must first be proven that it is possible to correctly transcribe the conversation with a client.
On a first set of tests, the AI was able to transcribe conversations and display them in real time on the agent’s workstation.
This initial transcript phase is therefore a technical success… but it is only a first step: agents do not need it, as such!
The project can therefore move on to the following 2 objectives:
- Present the right information and advice to agents, knowing that they already have many tools open on their workstations. Hence a work on UX,
- Develop an AI that will be able to detect in the conversation the intended use cases.
In May 2019, the first objective was achieved: the UX of the user interface is being finalized.
Soon the first tests on the v1 of the product
In June-July, the first version of the product will be tested. These tests will take place in a test environment, before moving on to customer tests, 3 to 4 months later.
The use cases tested will initially focus on the sales process, for example:
- At the beginning of the call, the objective is to enter key information, without errors and with less effort: the AI will transcribe the retrieved information into the CRM or check it;
- Based on this initial information, the AI will retrieve and analyse the description of the competing offer to which the customer has subscribed;
- The AI will select or build a relevant counter-argument and forward it to the agent (“you would save €3 a month”, “you would have and additional service”, etc.)
- The AI will supervise the call and check that no step in the sales process has been forgotten by the agent, or that the T&Cs are well stated
3 typical use cases are presented in these videos:
Artificial intelligence at the service of emotional intelligence
In the Self-Augmented Agent project, the objective is clearly not to replace the agent, or to transform it into a “robot”. The goal is to enhance his relational performance, free him from simple tasks so that he can concentrate on important tasks.
Ultimately, Allo-Media and Webhelp want to bring the solution to market, according to terms to be specified soon.
It should also be noted that, in addition, the 2 companies have developed other AI projects:
Feel free to contact me if you are interested in this project or in AI in the field of customer relationship and experience.
Mobile : +33 1 44 40 17 56
Keep reading :