‘MindReader’ Catalyst PoC uses APIs to track customer intent

mindreader catalyst data
MindReader Catalyst on display at Digital Transformation Asia 2018.

To improve customer experience (CX) and increase ARPU, communications service providers (CSPs) traditionally have relied on rather rudimentary tools like surveying customers to determine whether they’re satisfied. More recently, operators have begun using CRM platforms such as Salesforce to collect data about customers’ interactions.

Now a TM Forum Catalyst proof of concept called MindReader is taking this a step further by tracking customers’ intent using application program interfaces (APIs). The idea is to create a more accurate picture of customer experience that can be used to personalize services and create new ones.

Telstra is championing the Catalyst, which is being demonstrated this week at Digital Transformation Asia in Kuala Lumpur. Participants include:

  • CloudSense, supplying intelligent commerce, real-time decisioning and CRM on Salesforce
  • Infosys, providing orchestration, intent intelligence, intent qualification management and APIs
  • Nokia Australia, delivering the service management platform, digital intelligence and intelligent customer agent on Salesforce

Here’s how it works

Mindreader-Catalyst-graphic-325x186@2x

A customer interacts with a CSP (for example, by exploring and purchasing products, using them, experiencing troubles, churning, etc.) through the Cloudsense Intelligent Configure Price Quote (CPQ) system. In this context, the interactions are called ‘episodes’.

The customer is prompted at key points during the interaction for guidance through questions or selections of options through the Cloudsense CPQ graphical user interface. An ‘intent’ is captured at each salient point during the interactions through explicit questions and answers or usage patterns and recorded via the Infosys Intent qualification management, including the episode information and association in the intent management data store (in this case a data lake because of the volume and velocity of data).

The intent is ranked based on parameters such as relevance, frequency and value. The function is performed centrally in the data lake and also by the individual AI components while using the information in the context of episode – for example, Nokia Service Management Platform (SMP) and omnichannel capabilities for assurance episodes, Cloudsense for sales episode, intent qualification management for usage and product evolution episode.

The ‘intent object’, which is created as a result of the first two steps, is used to refine subsequent interactions with the customer for the episode (refinement location depends on the episode type and customer interaction – in this use case context in Cloudsense for next-best offer selection and Nokia SMP for assurance episodes) as well as others, and the most relevant intent is selected to drive next-best actions for offering the customer a personalized experience.

Intent insight is offered as a service internally from the Infosys intent qualification central data lake (and exposed externally as a product/service via the Cloudsense CPQ) so that any system from the network to customer-facing apps may call the service to deliver intelligence and relevance in real time for customer interactions.

Intent as a service is also available for product managers across the enterprise to help them evolve products. Eventually, it could be sold to other businesses as an offering configured in the Cloudsense CPQ.

As part of the project, the team has created two new APIs for intent and episode management, which they plan to contribute to TM Forum as Open APIs. In addition the project makes use of other Open APIs for service catalog and ordering and for activation and configuration. The team also plans to  contribute the intent and episode objects to the TM Forum Information Framework.

Making sure customers get what they want

CSPs have always captured intent and episode data, but they haven’t been able to track or utilize it efficiently because it often gets lost in interaction notes or call detail records. By making the data more accessible and useful, customers will benefit from better, proactive customer experience and new services suited to them specifically.

“We can use this intent information to ensure that we are providing a customer what they actually wanted, which may differ from what they have asked for,” says Andrew Davison, general manager of enterprise product software development at Telstra. In addition, by logging and tracking intent, the operator can retrospectively qualify customers for new services as they become available. This knowledge of what customers have asked for in the past can be used to build sales leads.

“We recognize that as we work towards a technology platform that is a federated set of small, loosely coupled services which we assemble into product offerings, it quickly becomes difficult to trace specific service actions, or intent episodes, initiated by users because there is no way to trace an API call with certainty back to a particular interaction with the customer (whether driven by an online portal, an app or service staff),” Davison adds. “So, we need to be able to propagate a label for the interaction down through the technology federation to make it simple and fast to recall, and profile all network and system interactions that occur as a direct consequence of the customer interaction.”

At Digital Transformation Asia, the team is hoping to get feedback from other CSPs on the new intent and episode management APIs and on the idea for delivering intent as a service. They plan to continue their project at Digital Transformation World in Nice, where they will test the APIs and introduce big data, artificial intelligence and business intelligence into the solution.

“We are planning to be involved in future phases of the work, to build out an end-to-end service prototype which allows us to demonstrate the power of intent propagation in better understanding what is happening in a complex, federated delivery environment without needing to engage data-sifting or probabilistic algorithms,” Davison says.

If you’re interested in participating in the project, please contact Tania Fernandes via tfernandes@tmforum.org.

Dawn Bushaus, managing editor at TM ForumWritten by Dawn Bushaus, managing editor at TM Forum | Original story posted at TM Forum Inform

Be the first to comment

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.