Implementing AI in telecoms


CX Network spoke exclusively to Saad Naseer, Senior Manager Products and Platforms at Pakistan Telecoms. As one of the pioneering organisations in the region to implement this system, we wanted to find out the successes and challenges of driving cultural change and how these systems can be used to realise tangible ROI within a business.

What were your motivations for implementing AU into your business strategy? 

We are an incumbent operator in Pakistan and have been operating since 1947. We have a huge legacy network and because of this we also have lots of complaints and issues with the customers.

Our call centre operates 24/7 and has 1,200 employees; but even with this huge number of call centre employees, it remains busy 24 hours a day. This issue called for a strategy to lessen the number of the call centre operators and, at the same time, increase customer satisfaction.

We decided to resolve this issue and found a company called Affinity which provides AI solutions. Their proposal said that after implementation, all our KPIs would improve, including the call holding time and call efficiency. 

The issue was that we have a legacy network in which we were bombarded with a huge number of customer calls which we were not able to manage. This prompted us to implement a solution that can provide us with a better method through which we can satisfy our customers.

What were the key challenges that you faced in integrating AI systems into your business?

We asked Affinity for a proof of concept (POC), and they deployed servers initially at our call centre. The main purpose of the servers was to record all the calls and they had an initial briefing with our operators so that they could intelligently route the calls.

When we integrated the system, an issue was that there is a different solution system deployed compared to the system we use for our older customers. Initially collecting the data takes six months and for company facing such a pressing problem, this is a long time. The vendor needed time to build enough intelligence within the system to make the decisions later on. 

What was the main benefit of implementing AI relating to your customer experience and service? 

We had three objectives in mind when implementing this system. The first was to reduce the mean call holding time; the second was reducing the complaints to calls ratio and third, was to start upselling to the customers.

The first two objectives were easy to implement because they were based on the system and the intelligent routing that could be done. The third objective to upsell customers who were complaining was a considerable a challenge.

AI worked very well for us because the routing was done in such a way that a like-minded customer was routed to a like-minded agent.

Pakistan is a big country with multiple languages and multiple regions, so when people called in there was sometimes a language barrier. Through this system they routed the call to a person who spoke the same language as the customer. This made a huge different to the customer that they can speak in their own language to an agent.

Affinity also did a survey of our call centre operators to find out what kind of activities they like and what kind of hobbies they have. In customer to operator mapping, this was considered and like-minded customers were sent to the like-minded operators. 

We tried to minimise the complaint ratio by putting in an Interactive Voice Response (IVR) system so that fewer calls were transferred to agents. In terms of upselling customers, AI also performed very well and we have seen very good results in all three KPIs.

How have you managed to realise tangible ROI through AI? 

In Pakistan, we were the first company to use this solution, so we were understandably sceptical at first. There are two challenges when you are implementing AI into an organisation.

Firstly there is a risk as to whether the system will generate ROI if the company has invested money into it. Secondly, the call centre operators play a very crucial role in the success of the system and we realised that some of the staff were very worried that if the system performed well their job could be in danger. We had to mitigate this issue by explaining to staff that the system was not here to take away jobs but rather to complement them.

We decided to have a revenue share with the vendor, and agreed with them that if we got incremental revenue from their solution, we would share a percentage. For the first six months we had three hours were we had this solution enabled, and then the three hours when the system was off and manual routing was done. This enabled us to compare both the results and measure the difference and we found that the difference was very noticeable.

How did you ensure that your business relationship with Affinity was mutually beneficial both sides? 

This depended on the results that we had in the initial six months when we were trialling the system. The vendor was very confident that they would have very good results because they had deployed this solution across the world, mostly in the European and American markets. After the six month trial period we were both confident that the system would bring results and be beneficial for both parties. 

Where do you see the future of AI within your organisation? 

I can definitely see the further implementation of AI within Pakistan Telecoms; in this case it was only implemented in the call centre operations. We have now decided to extend that to our CRM and real-time offering.

Based on the customer strength and based on customer viewing history, we can immediately pitch new offers to the customer rather than waiting for the customer decide what they want to buy. 

We are in the process of making an RFP for a scenario where the customer is interested in watching Netflix most of the time. We’ll pitch directly to that customer saying if you subscribe for this much of time, you will get this much discount. We also want to integrate AI with our omni channels.