The Internet of Things is generating a fresh round of charts projecting growth and revenue. And for service providers, some analysts are predicting a trillion dollars worth of potential services revenue.

But the projections gloss over some critical dependencies that a future of 50 billion connected devices rests on.

To start with, designing, developing and connecting all these new devices.

Identifying use cases that are viable economically (though “Smart X” is an intriguing start).

More flexible and low-cost forms of ubiquitous connectivity (software-defined networks? 5G?).

Virtualized, distributed applications that can scale up (and down) to meet demand.

But the Internet of Things also represents a fundamental challenge to the conventional operating model of telecom.

Here’s why.

1. IoT is inherently unpredictable.

Telecom has worked incredibly hard to build networks to support specific services: voice, video and well-behaved data. Statistical models ring-fence expected behaviour norms and become the basis of carriers sizing their networks (and ultimately equipment purchasing decisions).

IoT isn’t likely to be as well-behaved.

For one thing, the number of possible applications seems limited only by our imagination. A quick look at the program for a recent Smart X conference references smart cities, smart homes, smart grids but also smart buildings, smart forests…

What sort of demands will such IoT applications place on networks? Not only in terms of capacity but also latency, burstiness, resilience or time-to-repair requirements? And with billions of devices to be deployed, over what timeframe will these devices come on stream, be connected and start generating traffic?

Given the experience with other devices – from the iPhone to tablet to TV streaming services – it would be a brave service provider that insists on certainty over flexibility in IoT.

Traditional telecom is built on predictability so that it can forecast, size networks, and plan buildout on an annual timeframe. But in the IoT world, that will be the exception. With tens of billions expected to be connected within the next few years, service providers will need to shift their emphasis from long plan/build/operate cycles too much faster, automatic and adaptive ones.

2. Latency is a Market Opportunity

Telecom is steeped in the goal of engineering networks to provide the best possible quality of service.


Customers are willing to pay less for lower grades of service. Not everyone buys into Amazon Prime & next-day delivery. The UK still retains a second-class rate for post. While we might pay to fast track at the airport on business, we might be happy to pay less for the privilege of queueing a little longer when travelling in a group.

As for people, so much more so for “things” that are more likely to have a wider range of tolerance to latency. Sure – safety-critical applications like connected cars, monitoring aircraft engines or trains require absolutely the lowest possible latency. But there are a hundred other applications – IoT’s long tail – that don’t.

In an IoT world, there will be many more value points on the curve for carriers to operate at, and potentially to move between. This suggests that it may be more necessary (and more profitable) to be able to reconfigure services in real-time. The idea of optimizing a network against a lower grade of quality of service requirement seems odd but may make great business sense. (In the logistics business, startup Cargomatic’s business model is reselling the odds-and-ends of spare capacity on trucks to people whose delivery just isn’t that time-sensitive.)

3. 10x devices = 10x failures?

Maintaining service to 50bn devices connected is going to be a challenge. And It just won’t be possible to scale today’s NOC-based processes for detection, diagnosis and resolution of failures.

Service providers will need to automate much more of the entire process, ensuring that service can be maintained automatically. This will involve choosing intelligently in real-time from a range of access technologies (3G/4G/LTE and 5G) or a wider range of options that a software-defined network might present.

The scale and complexity point towards Artificial Intelligence as a critical enabler. AI and machine learning will be needed not only to detect failures but also to design and activate alternative configurations needed to mitigate their impact.

4. …and 5G is only part of the solution

5G includes brand new potential for service providers (or enterprises or MVNOs…) to flex networks in new ways. 5G network slicing, in particular, should provide an effective way to separate out the traffic of differing kinds into distinct layers of a mobile network, enabling them to be managed (and monetized) according to different rules and policies.

This will mean taking more, and more frequent, business decisions about how to use this capability to adapt to dynamic conditions. And it would be simplistic to limit that to simply adding more capacity. Connecting up short-lived sporting events, concerts or natural disasters will eventually need to turn down capacity just as quickly as they turned it up.

For example, it might be necessary to change the configuration in real time to maintain service quality within acceptable limits. It might be desirable to change the network in order to reduce the risk of a service failure (say for that slice that supports emergency services), even without an actual outage. The complex, multi-constraint problems presented by 5G’s potential are something that Artificial Intelligence is ideally suited to resolving.

IoT isn’t a technology play. It’s a scalability play.

In general, changing network configurations remains something that requires human intervention. In the IoT world, service providers will need to raise their expectations for how much can be automated, since humans in the chain will decrease speed and agility, and limit scalability. Not ideal with 50bn endpoints out there. These factors may explain why carriers, in general, have been a little slow to jump in on IoT

However, the step change in automation, intelligence, scale – as well as the opening up of massive opportunity – should prompt service providers to recognise the need for timely new technologies such as Artificial Intelligence and Machine Learning to help deliver it.