Applications of AI in North American Telcos

August 3, 2017

There’s little doubt whether artificial intelligence (AI) will change the future of telecommunications—but we don’t have to gaze far into the future to see AI at work. Many North American telecoms have already begun to invest in and deploy various types of AI-powered technology.

With the concept and application of AI being as broad as it is, it’s critical to highlight and examine these practical applications of AI in order to gain actionable business insight for decision-makers to move forward.

Below, you’ll find an overview of how several North American telcos are currently applying technologies that incorporate AI—many of these revolve around machine learning and deep learning research, which studies algorithms that can automatically learn, improve model performance, and make predictions based on data.

This information is largely based on press releases, web pages, presentations, and reports that have been made publicly available. This piece is not all-inclusive but instead gives a summary the major AI developments in North American telecom.


CenturyLink’s notable application of AI is its 2016 investment in an AI-powered sales assistant [1]. Made by Conversica, the automated sales assistant “Angie” sends out close to 1,000 emails per day using natural language processing algorithms and sales data to identify which respondents are likely to be hot leads. The automated sales assistant has been reported as processing an additional 120-400 qualified leads per month, engaging in two-way conversations with leads (before passing them off to human sales reps) to enhance outreach and follow-up while freeing up time for sales reps [2]. Hear the Director and Manager of Marketing Operations at CenturyLink talk about "Angie" in the video below:


This North American giant has invested in AI-fueled technologies across the board. Facilitating AI innovation with the formation of the Comcast Applied Artificial Intelligence Research Team, [4] Comcast is looking at various machine learning and big data technologies to enhance content discovery, network capacity planning, customer service, and more. For many of the following developments, Comcast has used the open-source machine learning platform from H20 [5] in combination with Apache Spark’s large-scale data processing capabilities [6]. Here are a few of the developments, as summarized in a 2016 presentation from Jan Neumann [7], a director at Comcast Labs Washington, D.C.:

Predictive text: uses predictive text algorithms to auto-fill search queries as the user types them in using a remote. Knowledge representations factor in here, as researchers develop ways to structure information in ways that are more compatible with human queries

Voice-controlled TV: natural audio language processing algorithms are being used to enhance voice-controlled TV (which initially launched in 2015 and allows users to control their TV by speaking through the remote)

“Smart Favorites”: combines scene-level user metadata, computer vision, and machine learning to personalize content discovery on an individual user level, not just at the household level

“Trending on Xfinity”: enhances content discovery by predicting which content is most likely to be popular in the next hour, day, etc.

Smart home products and IoT: connected home devices that use machine learning to anticipate a user’s needs

Customer service: using intelligent scoring and machine learning for customer care calls; using machine learning models to automatically predict service outages and repair them before they happen

Predictive caching: predicts what content is likely to be popular at a certain time and prioritizes local caching for the most popular content and times accordingly

Preventing Avoidable Truck Rolls: currently in development, this application of AI will use call center data, predictive analytics, and machine learning to prevent avoidable truck rolls before they occur. This solution will identify and automatically attempt to solve customer calls that shouldn’t require a maintenance worker to fix the problem in person.


IShaw is one of several telos that use Comcast’s cloud-based Xfinity middleware platform [8] That said, Shaw’s BlueSky TV AI-based features include many of the same customer experience measures as Comcast’s XFINITY, such as voice-controlled TV and viewer personalization [9].

An overview from Mr. Shaw:

“With BlueSky TV now available everywhere we offer cable video, we are thrilled to introduce to Western Canadians a revolutionary TV experience made possible by our strategic partnership with Comcast. The BlueSky TV experience is more than just a new guide and set-top-box, it is an elegant system that listens, learns and curates content to provide an exceptional viewing experience. We are optimistic that BlueSky TV combined with WideOpen 150 and flexible TV packages will provide a compelling reason for consumers to stay and switch to Shaw.” [10]


Little information about AI applications at Bell has been made publicly available, but given Bell’s partnership with Nokia as part of Nokia Bell Labs, we imagine that Bell Canada Enterprises may make use of the predictive repair care service presented by Nokia in early 2017 [11].

The predictive service known as MIKA (an acronym for Multipurpose Intuitive Knowledge Assistant) combines augmented knowledge, automated learning, and a knowledge library to predict likely hardware failures and present relevant information, data, and recommendations to engineers, ultimately saving time. MIKA is reportedly capable of predicting failures and recommending replacements as far as two weeks in advance, with up to 95% accuracy [12].


Machine learning and AI comprise a small fraction of the wide range of data-driven endeavors at AT&T Labs Research. By exploring the Labs Research page, you can see how AI and machine learning fit into the bigger picture of cloud technologies, optimization, and big data developments [13].

Mazin Gillbert, the vice president of Advanced Technology at AT&T Labs, indicated the company’s development priorities in an article published on AT&T’s blog. Along with AT&T’s new ECOMP platform, which went into open source in early 2017, Gillbert talked about “extreme automation and optimization”, mentioning data-powered algorithms, model-driven service design, self-healing, and self-learning networks. Gillbert explains that AT&T “can utilize ECOMP and development of machine learning and massive data modeling platforms to support emerging intelligent services.” [14]

AT&T’s software-defined approach is being manifested in their Network 3.0 Indigo, of which they hope to have 75% virtualized by 2020. Indigo appears to be recognized as a main component of their overall strategy concerning big data, AI, and machine learning. Gillbert has also mentioned that the company is focused on using AI to implement predictive maintenance [14].

You might also recognize some of AT&T’s applications of AI as part of the following:

Atticus: in addition to developing a customer service chatbot, the company has rolled out an entertainment chatbot named “Atticus”, announced in late 2016. The bot interacts with users via the Facebook Messenger platform and responds to voice and text, spouting knowledge about actors, movies, TV shows, etc [15]. See the introductory video below:

Flying COW: the Flying Cell On Wings refers to flying drones that function essentially as a cell tower, which is useful for obtaining coverage in areas with limited coverage or in disaster situations. The company is exploring ways to add AI and machine learning technology to the Flying COW for enhanced support and infrastructure maintenance [16]. See in the Flying COW in action below:


Telus offers another example of using machine learning for predictive maintenance. Using service monitoring technology from Splunk [17], Telus monitors the noise rise from cell towers with machine learning algorithms that can identify subtle network incidents that have a high likelihood of affecting service availability. For more detailed information on how this works, watch Splunk’s 2016 video presentation [18].


Rogers now offers its subscribers a type of personal service technician through the MyRogers App on their phones. The so-called “DeviceAid” service is powered by Wysdom, a self-care chatbot developed by CrowdCare (many informational video clips are available on their site, including one that shows how it works with the MyRogers App). Using natural language processing and factoring in aspects of the specific customer context, the chatbot is being used to quickly and accurately answer subscribers’ technical questions and concerns, ideally resolving problems through self-care before a subscriber would need to make a call to a customer care center [19].

Here's a video showing how DeviceAid looks on the subscriber side:


Unlike the example set by Comcast or AT&T, Verizon hasn’t made many public announcements highlighting their use of AI-based technologies. However, in early 2017, they announced the launch of Exponent, a B2B service for global carriers [20]. Included in the suite of services is a Big Data & Artificial Intelligence platform, which is “designed to assist carriers to unlock and monetize their wealth of data through the application of advanced machine learning techniques, deep analytics, and artificial intelligence.” [20]

Similarly, while we haven’t found anything from Verizon declaring their use of AI for predictive maintenance, they do offer it as a service (referred to as condition-based maintenance): “Data analytics can help you predict when machines will need service, or when parts might break down. Then, you can proactively resolve small issues before they turn into costly repairs or become major problems.” [21] Since Verizon is offering these services, we can assume that they are in use within Verizon’s own service as well.


1.Power, Brad. How AI is Streamlining Marketing and Sales. Harvard Business Review. June 12 2017. Retrieved Aug 3 2017.

2.CenturyLink Heats Up Warm Leads for a 20-to-1 ROI with Artificial Intelligence from Conversica. Conversica. Retrieved Aug 2 2017.

3.Big Data Services & Management. CenturyLink. (2017). Retrieved 2 Aug 2017.

4.Research — Comcast Labs, DC. (2017). Comcast Labs, DC. Retrieved 29 July 2017.

5.Operationalizing Machine Learning at Comcast. H2Oai. Accessed 04 Aug 2017.

6.Apache Park —Lightning-Fast Cluster Computing. (2017). Retrieved 03 Aug 2017.

7.Neumann, Jan. How Comcast Uses Data Science and ML to Improve the Customer Experience. Presentation. May 2016. Retrieved 01 Aug 2017.

8.Sosiak, Mia. Shaw BlueSky TV offers first voice-controlled video box in Canada. Global News. 11 Jan 2017. Retrieved 30 July 2017.

9.Blue Sky TV—The Future of Television. Shaw Communications. (2107) Accessed 30 July 2017.

10.Shaw Announces Second Quarter Results. Shaw Newsroom. 12 April 2017. Accessed 01 Aug 2017.

11.Nokia Bell Labs. Nokia. Nokia launches MIKA - the first digital assistant customized for telecommunications operators. 27 Jan 2107. Accessed 30 July 2017.

12.Nokia launches MIKA – the first digital assistant customized for telecommunications operators. Telecom Reseller. 27 Jan 2017. Accessed 02 Aug 2017.

13.AT&T Labs Research — Inventing the Science Behind the Service. (2017). Retrieved 03 Aug 2017.

14.What’s Next at AT&T Labs? AI Set to Revolutionize the Network | AT&T. (2016). Retrieved 31 July 2017.

15.Meet Atticus: the Entertainment Chatbot from AT&T. AT&T Intellectual Property. (2017) Retrieved Aug 4 2017.

16.Walker, Jon. Industrial Uses of Drones – 5 Current Business Applications. TechEmergence. 7 June 2017. Accessed 01 Aug 2017.

17.Machine Learning | Splunk Enterprise | Splunk. (2017). Splunk. Retrieved 03 Aug 2017.

18.Rombough, Jason and Modestino, Matthew. Building a Smarter Strategy for Alarms with Splunk Machine Learning Toolkit. Online presentation. 2016. Accessed 03 Aug 2017.

19.Videos | CrowdCare. Wysdom Introduction, Rogers DeviceAid, and How To Use The Device Aid Feature In The MyRogers App. Video clips. Accessed 04 Aug 2017.

20.Verizon Launches Exponent, a New Technology and Business Venture Designed to Accelerate Growth for Global Carriers. 23 Feb 2017. Retrieved 04 Aug 2017.

21.Smart Monitoring & Maintenance | Condition-Based Maintenance. 2017. Accessed Aug 04 2017.

©2010 - 2018 by Tribal Interactive Inc.