8 Minute Read

Conversational AI And Smart Robotics


A Fully Remote Project

Project Summary

Freethinking was approached by Telkom, a prominent South African telecoms provider to rapidly investigate, ideate, design and build working "proof of concept" (POC) digital solutions to challenges they were facing in their call center with customer interactions. This case study details our method for selecting journeys from various personas and then describes each step in our Service Design Discovery and Delivery process for two of these journeys, one detailing a frugal customer and the other a high spending customer.

Problem First

Discovering And Validating Problems To Be Solved

Understanding business objectives:

It is essential to preface customer-centric research with an in-depth understanding of the business objectives.

Telkom’s business objectives were summarized in the following 5 points:

  1. Reduce call center volumes by at least 30%: Proactively resolve issues and direct customers to other channels.
  2. Reduce the cost of serving customers: find intelligent ways to automate.
  3. Improve first time resolution rates: help customers first-time with minimal delay.
  4. Improved customer experiences: deliver reliable customer experience that is frictionless and effective.
  5. Increase self-service: this will directly enable all of the aforementioned business objectives.

These objectives align with Telkom’s marketing concept “YQ” (Why Queue?), which promotes the reduction/eradication of queueing for services.

The Answers We Found

Cumbersome processes in the call centers concerned with the identification and verification of customers create unnecessary pain points before the real problem is addressed. Reliance on a call center agent to solve the most basic problems degrades the customers’ experience, and impacts the efficiency of call center agents.

A Summary Of Issues
  • No self-service capability for customers.
  • 25% of call center volume comes from the category; mobile prepaid, many are routine processes.
  • Average customer hold time for support is 6.5 minutes.
  • Affluent customers already have access to other internet-enabled devices, and expect the same from their telecoms service provider.
  • Telkom does not currently offer smart solutions to connectivity problems.
  • There is no platform where customers can view their LTE data balance.


Agile Collaboration

The project was executed using Agile methodology. Multi-disciplinary teams, including data and analytics experts, automation experts, project managers and developers collaborated, and this collabotation resulted in a cross-pollination of ideas and skills.

Service Design

Service Design is a strategic discipline that enables coordination of front stage customer interactions with backstage systems, processes, technologies and business teams. It is valuable when building complex services containing multiple touchpoints in an omni-channel space. Research and solutioning activities deliver a range of artefacts, including personas, journey maps, problem statements, service blueprints and prototypes.

Research And Discovery

Personas: Understanding Customer Needs

Having gained an understanding of the business objectives, we focused our attention on understanding customer needs through customer research. We created two personas to focus our thinking and ideation.


Journey Selection: Journey Mapping

Building on our personas, we mapped a number of journeys with which they were experiencing trouble.

Due to the rapid delivery required, our team workshopped which journeys to choose and why.

Journey Selection: Selecting User Journeys

Two journeys for further solutioning were identified for each persona, based on their potential to meet the identified business goals.

Journey 1: Frugal Customer

The mobile prepaid category represents almost a quarter of forecasted call volumes for 2020-2021*.

A significant number of these calls relate to the routine loading of prepaid vouchers for frugal customers.

The average hold time on a call with Telkom is 6.5 minutes**. This customer experience can be improved.

Journey 2: High Spending Customer

High spending customers typically have access to multiple internet-enabled devices*.

They are increasingly watching streaming media, a trend amplified by Covid-19 lockdown.

Subscription to streaming video on demand services (SVOD) such as Netflix increased in South Africa to 75.2% in 2020**

There is currently no digital self-service option from Telkom to intelligently assist customers with connectivity issues.

* Telkom RFP Call Centre Technology and SMB Customer Journeys presentation. ** Telkom Call Centre Data 2019-2020

Identifying Pain Points

Having selected and validated two journeys, detailed journey maps were created. These maps place emphasis on behaviours, feelings, thoughts and emotions at each touchpoint.

Journey Selection: As-is Journey – Frugal Customer
Journey Selection: As-is Journey – High Spending Customer

Problem Statement

Crafting A Problem Statement

Having defined our personas, interrogated in detail the pain points felt by customers, staff and the resultant business process inefficiency, we were ready to craft a problem statement for both frugal and high spending customers.

‘As a Telecoms provider, I need to’:
  • Radically reduce call center volumes
  • Reduce the cost of serving customers
  • Improve first time resolution rates
  • Improve customer experience Drive self-service

Getting the detail right is crucial, as each keyword in the problem statement expresses an area of concern which the solution must address. In aggregate, this statement provides the necessary perspective and context to guide, limit and define the project scope.

Problem Statements

Journey 1: Frugal Customer
  • How might we…

    provide frugal customers with an alternative to the call center to recharge airtime vouchers

  • in a way that…

    is smart, predictive and easy-to-use

  • given that…

    …the reality is that many South Africans don’t own a smart device?

Our Design Solution

A way for customers to text a photo of their airtime vouchers to a service that uses OCR and computer vision to automatically extract and load the recharge value.

Journey 2: High Spending Customer
  • How might we…

    provide high spending customers with an alternative to the call centre to query and resolve internet connectivity issues via a variety of internet connected devices

  • in a way that…

    is smart, predictive and easy-to-use

  • given that…

    …these customers expect personalised omni-channel service?

Our Design Solution

A chatbot, available across multiple channels, fuelled by conversational AI that employs predictive data analytics and Robotic Process Automation to provide convenient customised assistance.

Metrics For Success

With the problem statement, we were able to determine our solutions’ success and failure conditions. We measure success objectively – we constructed metrics (and associated project measurement protocols) to correlate with each of the keywords expressed in our problem statement. These metrics cover three areas: accuracy, speed and customer experience.

We expect to see a distinct correlation between these metrics and traditional business measures such as an increase in revenue and customer retention.


Service Blueprints

After mapping out the customer journey, we visualised the interplay between frontstage customer actions and backstage business logic, technology and systems of our solutions on service blueprints.

The proposed solutions underwent a process of detailed design, idea development and results expression on service blueprints. These blueprints depict the interaction between front-stage customer actions and back-stage business logic, technology, and systems as well as Telkom staff interactions.

These Service Blueprints served as a launch pad for user stories and technical architecture diagrams; the fundamentals software building.

Iterate And Refine

Designing The Experience

At the same time, we started to design the experiential details of the to-be (desired state) user journeys by:

  • Mapping out a detailed information architecture and decision tree for the chatbot.
  • Writing conversational dialogue and considering how Natural Language Processing should be integrated into the experience.
  • Designing the look, feel and style of the chatbot interface, while considering how to incorporate Telkom brand elements and graphic cues.

We designed a BI report to track the NPS metric as customers provided us feedback.



Tech teams, working closely with business and design teams, ensured alignment on the technical challenges of constructing a chatbot and the backend automation components.

Robust backend frameworks ensured that the solution was accessible to customers and created an effortless experience.

Journey 1

Easily Loading An Airtime Recharge Voucher

To demonstrate our solution, we printed test vouchers similar to the ones the customer might buy in store. We snapped a photo of the voucher, our service picked it up, did OCR and pulled out the text. It then called a Telkom API, loaded the voucher and texted back to indicate success along with the updated airtime balance.

Journey 2

Getting Back Online After A Data Connectivity Issue

We demoed our chatbot, the backend UIPath RPA frameworks and displays, and the associated Net Promotor Score dashboard (developed with PowerBI). We used the API’s provided by Telkom, and where relevant, we supplemented with our own data to create a rich, interactive and highly personalised assistance experience. Once the bot had helped the customer to get back online using predictive analytics, it proactively suggested they switch to a different data package to prevent future downtime.


In this solution’s delivery is proof of the viability of the Service Design methodology in a project with a rapid delivery requirement.

The two solutions demonstrated radically different, yet pragmatic approaches to employing automation as a means of improving business efficiency and reducing customer pain points. This is achieved by eliminating the need for interaction with a call center.

Key Features

Working Chatbot Fuelled By Predictive Analytics And Conversational AI

Our chatbot solution offers more than a simple decision tree. It uses natural language processing to engage with customers in a personalised conversation.

Once the chatbot helped customers get back online, it proactively suggested they switch to a different data package to prevent future downtime using predictive analytics. We used the API’s provided by Telkom, and where relevant, we supplemented with our own data to create a rich, interactive and highly personalised assistance experience. A nuanced, sophisticated process sure to impress the tech-savvy customer.

Net Promotor Score BI Dashboard

We built an interactive dashboard to allow the business to dynamically monitor customer engagement over time.

Working Solution To Easily Load An Airtime Voucher By Taking A Photo

To demonstrate our solution, we printed dummy vouchers similar to the ones customers might buy at retail outlets. We snapped a photo of these, our service picked it up, did OCR and pulled out the text. It then called a Telkom API, loaded the voucher and sent a text to indicate success and the customers’ new airtime balance.

Robotics And Automation Monitoring Framework

The backend automation we employed in our solution is monitored and controlled via a series of information-rich dashboards and interfaces.


Rapid Ideation, Brainstorming And Delivery

We proved that it’s possible to employ a full Service Design methodology to a project with a very short turn-around time, and deliver practical working results. Our multi-disciplinary team brought a variety of perspectives to the table and our diverse backgrounds fuelled intense creative collaboration.

Tech Innovation Backed By Customer Insights

Our process begins with empathy. This approach ultimately paid dividends as our focus on customer needs gave us the insight to develop automation solutions for a traditionally underserved customer segment – one that has little access to smart devices.

Realising Business Objectives

Our solutions were informed by Telkom’s broad business objectives and were tailored to meet them.

Methodology For Ongoing Monitoring And Iterative Improvement

Our focus on metrics and the provision of associated monitoring dashboards means that the success of the initiative can be continuously monitored into the future, allowing enhancements to be made based on real customer behavioural data, sentiment analysis and qualitative feedback.