Home Analysis Service providers should build their own digital assistant ecosystems and skills

Service providers should build their own digital assistant ecosystems and skills

Share on

Service providers offering Pay TV and broadband could introduce their own digital assistants, complete with microphone and speaker, create their own operator-controlled AI/voice environment (with automatic speech recognition and natural language understanding) and provide their own set of ‘skills’. A householder could then say: ‘Okay Service Provider, what is my Wi-Fi password?’, or ‘Okay Service Provider, what new movies are on soon,’ or ‘What is my current bill’, or ‘Okay Service Provider, why does my Wi-Fi suck?’ The assistant could be in the form of a dedicated device, or more likely a three-functions-in-one unit that combines digital assistant, Wi-Fi extender and IoT hub features. One option is to use the set-top box – including STBs that are already in homes if they have the right wireless technologies onboard.

These possibilities were outlined recently by Charles Cheevers, CTO for Customer Premise Equipment at ARRIS, when he also revealed that his company is in discussions with a number of service providers about their digital assistant options. ARRIS, which develops set-top boxes gateways, routers and Wi-Fi extenders, among other things, has prototypes for 300 ‘skills’ that a service provider may be able to offer. Cable operator CEOs have been particularly impressed with the ‘skill’ that lets consumers declare: ‘Okay Service Provider, upgrade my broadband’!

Although Google and Amazon are driving the retail market for assistants, the Pay TV and broadband provider and vendor community has, or has access to, the combined skillset to create alternative voice/AI ecosystems. Nuance has what amounts to an ‘independent’ suite of technologies for speech recognition and natural language understanding, and the company has AI capabilities, as one example.

The biggest operators could probably develop their own voice/assistant solutions. Comcast won an Emmy Award for Technology and Engineering recently for the Xfinity X1 Voice Remote and the technology behind it. The solution uses a cloud-based platform with natural language processing and harnesses machine learning to understand what customers mean when they say certain words and phrases. It is not a giant leap to link these capabilities (which already cater for television navigation commands) through to some general service-related skills/apps.

Service providers much smaller than Comcast could share the ‘assistant’ ambition. SoftAtHome, which creates software for customer premise equipment, has been demonstrating an operator-centric voice assistant solution, complete with its own conversational software and an assistant personality called Maestro (‘Maestro, turn the lights down’). You can read more about that here. ARRIS is developing apps to show the service provider community the functions that will be possible. One of these is to run a speed test on the broadband.

The applications a service provider offers could become quite sophisticated, combining multiple actions under a single consumer command. Cheevers offers the example where you tell the assistant to lock up for the night. Harnessing its IoT intelligence and smart home capabilities, the operator could instigate a sequence that includes automatically bolting doors, confirming they are secure and showing security camera feeds from around the outside of the house on the television set.

Building the skills is not a barrier for this market, Cheevers says. One reason is that 10% of the skills will probably account for 90% of all consumer requests. You can build out a long-tail of apps from there.

Cheevers believes there is an opportunity for service providers to run multiple assistant ecosystems on the same device, which is supplied by them. Alexa, Google Assistant and maybe even Apple’s Siri could be present. Cheevers thinks consumers could then engage these different ecosystems for different tasks: Amazon for shopping, and Google for search, perhaps. The service provider assistant ecosystem, with its own entry command (e.g. ‘Okay Service Provider’) sits alongside them.

Having multiple assistants on the same device is a potential differentiator, Cheevers reckons, since Amazon is unlikely to make Google Assistant available on an Echo or Dot, and Google is unlikely to put Alexa on the Google Home hub. And on the whole, the skills that Amazon, Google and Apple provide are non-competitive to a service provider, so can be comfortably accommodated on an operator device.

Consumers are increasingly likely to have one of these ecosystems on a retail device, anyway. (Amazon’s Echo, Echo Show and Dot, Apple’s HomePod, Google Home and Google Home Mini are all dedicated speaker/assistant hubs, while Siri (Apple’s assistant) is on Apple TV and Google Assistant is included with Android TV, for example. Cheevers argues that if people are going to interact with these systems anyway, it may as well be on an operator device.

There is also the option – as a parallel or separate strategy – to build service provider skills (apps) that are made available with Alexa or Google Assistant. Amazon has its own equivalent of an app-store with third-party skills. Spotify, news organisations and newspapers are among those offering skills today. Third-parties are slowly filling the apps catalogue that goes with Google Assistant alongside the many functions provided by Google itself.

Cheevers thinks there is a limit to what a service provider will want to put inside third-party assistant ecosystems today, however. Google and Amazon are currently in the ‘maybe friend, maybe foe’ zone, he reckons, although he expects that over time these companies will become closer partners in the way that Netflix has. He thinks operators should avoid encouraging consumers to share what he deems sensitive information with third-party assistants, even when commanding them to launch a service provider skill.

In the case of Google Assistant, when you ask it to perform a task, it hands over to the skill/app and what happens inside a non-Google skill is then invisible to them. But Cheevers is concerned with what happens before that point on third-party assistants.

As a generic example, if there was a third-party assistant ecosystem called Charlie, and a consumer says, ‘Charlie, tell service provider my Wi-Fi sucks’ then, in theory, Charlie could gain an insight into how good a job the service provider is doing with its Wi-Fi, over time, Cheevers worries. Tasks associated with faults and self-healing should therefore be exclusive to the operator assistant, in his view.

[We asked Amazon and Google for technical details about whether the information they acquire during requests to a third-party skill/app (or in statements that precede or follow a request associated with a specific skill/app) are recorded or quickly deleted. We also asked them for a comment generally about the views expressed by Cheevers above. Google pointed us to their guidelines for ‘Data security & privacy on Google Home’. There is not much there to illuminate the matter, although one section says: ‘Does Google store my conversation history with external services? Yes, we save your conversations to make our services faster, smarter, and more useful to you, and also to help us protect you’.

Even if third-party assistant ecosystems do record the data somewhere, it does not mean it will ever be looked at again, let alone analysed. Service providers will make up their own minds whether Cheevers’ ‘safety-first’ advice is necessary.]

There is another reason for operators to have their own assistant ecosystem on their own device, anyway. This gives an operator greater control over how many people the assistant reaches. You still need to ship new CPE (like a Wi-Fi extender with assistant onboard) or download an assistant capability to existing set-top boxes (which is possible, Cheevers says, subject to the technology onboard). Even so, you are not entirely reliant on the penetration of Echo and Google Home devices (e.g.) to get your skills into homes.

Cheevers sums up the kind of multi-tiered assistant strategy an operator could follow. “There seems to be a mutually beneficial solution to service providers and Alexa and Google Smart Assistant platforms in particular, where the service provider can offer a combined service from Alexa and Google Assistant as well as their own specific skills. For the service provider, these skills can be implemented to work with specific Amazon and Google Assistant devices.”

He continues: “When you are dealing with more sensitive information like customer support or self-healing skills, these [functions] can be performed directly as a service provider skill using separate ASR [automatic speech recognition] and NLU (natural language understanding) implementations.”

Cheevers believes the devices that an operator supplies to subscribers tend to be physically located where people are – with set-top boxes in living rooms or kitchens and Wi-Fi extenders usually located near the middle of the house. With an increasing focus on Wi-Fi quality, extenders are a product more service providers will be shipping in future. ARRIS has been showing operators options where an extender/assistant comes with a choice of high-end speaker (suitable as a music centre) and more modest speakers suited mainly to voice (like on the Amazon Dot).

A set-top box could potentially leverage the speakers inside the television and a microphone inside the new generation of mic-enabled remote controls. Using an STB as the assistant has the advantage of providing visual feedback (something Amazon has introduced with Echo Show). Equally, it has the distinct disadvantage that it is not always on, or may need to be woken from a sleep mode. Having an operator assistant on both a dedicated operator device and the STB/remote (including via a software update) may be the answer.

ARRIS sees tremendous value in making diagnostics, self-healing and customer care functions part of a voice-enabled assistant environment. “We think they will reduce customer care costs, with less people phoning the operator,” Cheevers says.

As an example of how this could play out, a customer might declare, ‘Okay Service provider, I have a problem’. It replies, ‘Welcome to problem solving. What is your problem?’ A dialogue then continues with the bot, which can instigate checks and launch healing processes. It might present your home network as a diagram on the television set. The bot could confirm: ‘Fixing iPhone connection’.

The bot could make a decision to move a phone or tablet to another access point if this will improve the Wi-Fi/service quality for video streaming. ARRIS has written a ‘skill’ that will access service provider home network performance data and decide if someone needs a Wi-Fi extender. The bot could tell the subscriber: “A Wi-Fi extender will improve reception. We are posting a Wi-Fi extender to you.”

Photo shows ARRIS VAP4300 Hero Wi-Fi extender


Share on