Saturday, 16 December 2017

Artificial Intelligence – endless possibilities for the telecoms operator?

A Knowledge Network article by Stuart Newstead, Director Ellare Ltd
Tuesday 05 December 17

Artificial Intelligence (AI) brings with it the opportunity to uncover deep insights and to revolutionise decision making and business operations. AI also has the potential to create vast black holes for business strategy, investment and analysis. Anything that increases awareness of the issues and questions surrounding AI is therefore to be welcomed. This article summarises a discussion at the 2017 Total Telecom Congress, where I chaired a table discussion on the impact of AI on telecoms operators. Main points 1. The “self-learning” element of AI is critical for telecoms networks 2. AI adoption by telecoms operators is still more talk than action 3. One of the main obstacles to insightful data analysis is inaccurate or incompatible data. Machine-sourced input data can be of great assistance in improving data quality 4…

Artificial Intelligence (AI) brings with it the opportunity to uncover deep insights and to revolutionise decision making and business operations. AI also has the potential to create vast black holes for business strategy, investment and analysis. Anything that increases awareness of the issues and questions surrounding AI is therefore to be welcomed.

This article summarises a discussion at the 2017 Total Telecom Congress, where I chaired a table discussion on the impact of AI on telecoms operators.

Main points
1. The “self-learning” element of AI is critical for telecoms networks
2. AI adoption by telecoms operators is still more talk than action
3. One of the main obstacles to insightful data analysis is inaccurate or incompatible data. Machine-sourced input data can be of great assistance in improving data quality
4. There is a talent gap of data scientists who can also speak the language of business. And such talent as there is does not necessarily see a job in a telecoms operator as the most attractive place to work.

What is AI and what part of AI matters most for telecoms?
A dictionary definition of Artificial Intelligence is:
The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.

Within the context of telecoms, the table felt that the implicit “self-learning” within the definition is the factor that is most critical.

A telecoms operator’s functions can be broken down, at a simple level, into categories of Customer Experience, New Revenue Streams and Operational Efficiency. It is probably the network control within Operational Efficiency ("the self-healing network") that offers the earliest opportunity to use AI effectively. So much data is beyond human analysis; it is algorithmic and dependent on feedback loops; and network volumes and the range of services rise ever upwards. This makes the network control ideally positioned for the feedback loops to become properly “self-learning”, whilst staying within bounds of control and thus avoiding unexpected or unexplainable consequences.

One further observation made was that a lot of the clever longer-term AI and self-learning is likely to be at the network edge (devices, apps). For example, a videoconferencing app that learns to pick out and focus on faces rather than background as realtime end-to-end Quality of Service varies, because it is faces where we concentrate our attention at the other end of the call.

So why aren’t self-healing telecoms networks commonplace?
Telecoms operators often find themselves in the slow lane when it comes to implementing technology advances and benefitting from the added value of new services or of old services delivered more effectively. Table participants expressed concern that this could be happening again:
“The notion of ‘self-healing networks’ was on display at BT Research Labs in England over 20 years ago - complete with an ant colony seeking food using simple rules. AI is still being mainly talked about, rather than done”

One issue voiced was that technical people like to keep their knowledge and expertise close to their chest, so that these doesn't become systemic or widely-shared. Knowledge is power, as the saying goes:
“What's the incentive for someone to pass their knowledge to a machine? Is it a turkey voting for Christmas?”
A more positive view expressed was that AI has the potential to take away a lot of the more repetitive and routine work associated with any role – much like the replacement of mass typing pools by word processing software and the subsequent creation of “knowledge worker” jobs in service, analysis and digital.

Data quality
Some of the data practitioners around the table pointed out that they spend a disproportionate amount of time cleansing the data in any database. This doesn't usually leave much time for deep analysis, no matter how much talent they have in their data science team.

The group felt that machine-sourced data is a massive step forward, even though it is something of an unsung hero. It dramatically increases input data quality, for example by reducing rekeying or reformatting, or by having clearly-chosen preset category choices:

“Machine-sourced data reduces cleansing time, effort and cost. So you can get on and actually analyse it”

The goalposts are constantly shifting, however. One might think that metatags and categorisation would be something that would help maintain data accuracy, storage and access. However, as one participant highlighted:

“We have to think about what data is permanent, eg when the data is a fact of something happening. Other elements, like name - or even gender - can change, so tagging and categorisation is a lot more difficult”

Last but not least, the Europeans around the table stressed that they are having to give a great deal of thought to the impact of GDPR on requirements for the holding of accurate personal data. One area that is very much front of mind is how one best links together different data sources that may hold personal information. And the clock is ticking to GDPR Day on 25th May 2018.

The “talent gap”
The table had a near unanimous view that:

“There's a big talent gap. What is missing are ‘data scientists’ who can bridge between the analytics function and the business need and speak both the language of analysis and the language of business.”

This talent gap is common across industries. What gives telecoms a particular challenge is that other types of firm are often seen as more attractive places to work and to develop oneself. Examples of such firms would be social media, startups and rapidly growing software-based businesses.

And on that open challenge, we ran out of time. My thanks to all the participants at a very full table.

For more information on the strategic issues facing telecoms operators and other organisations in the digital, data or communications sector, contact Stuart Newstead at sn@ellare.net

This article is based on a roundtable discussion at the 2017 Total Telecom Congress. If you are interested in participating in 2018 contact info@totaltele.com

Please enable JavaScript to view the comments powered by Disqus.

Newsletter signup

Quickly get on board and up to date with the telecoms industry