Omer Sagi on applying AI in the smart home

Omer serves Tritone Technologies as VP Products and Business Development. Previously, he headed the
multimillion-euro business of Smart Access & IoT solutions for Assa Abloy EMEA. As an expert in the
fields of IoT and connected homes, Omer shares his wisdom and experience with us.

Do you agree that Connected Home Malfunctions (devices and services) are leaving the WiFi to other links in the Service Chain (Cloud, device, user, etc.)?

Malfunctions are not leaving the WiFi. In fact, WiFi problems are actually increasing due to the swift adoption of connected devices in the home. More devices means more connections and that means more pressure on the home router to supply connectivity and bandwidth—both for communication to the internet as well as for inter-device communication within the home, “domestic M2M”. At the same time, however, the numbers and types of services that all these devices consume is growing and becoming more complex. So, we see an even steeper increase in the numbers of malfunctions at the ends of the Service Delivery Chain—in the devices themselves—and at the top of the chain—in the Cloud where media providers reside (music and movie streaming, gaming, etc.). Dealing with all the potential malfunctions necessitates the ability to see and act across the entire Service Delivery Chain within and beyond the home.

Why do you think that, from the end-user perspective, the ISP is and will always be the usual suspect regarding everything not functioning in the connected home (even in cases where the malfunction’s “real” location is in the device or the Cloud components which are not under their responsibility)?

I hesitate to use the term “always”, but for the foreseeable future, yes, the ISP will be the go-to party for support of the connected home. End-users purchase and install devices from a wealthof vendors at a frenetic pace without checking with their ISPs for compatibility. They simply expect their ISPs to supply all their connectivity needs regardless of device requirements. Whenever there is a problem that affects customer experience (“My game is lagging”, “My movie is stalling”, “My video call is garbled”), the customer’s first suspect is the party that supplies the connectivity—the ISP. This puts ISPs in a dilemma. On one hand, to retain customers, ISPs want to help them enjoy their connected-home experience, so they will do whatever they can to troubleshoot problems wherever found. But on the other hand, the problems are increasingly spreading beyond the borders of the home (to the WAN or the service cloud) where the ISP has little or no visibility let alone problem-detection capabilities. Even within the home, the support game is shifting rapidly. ISPs lack the smart tools to see into the wide range of smart devices and the services they are consuming and certainly cannot deal with device-specific problems. So, in a great many cases, ISP Customer Care agents have to resort to the remedy at hand, expensive router replacement, which often is not the source of the problem at all. This just adds to the nightmare of high support costs and customer frustration that is only growing worse day by day.

Do you accept the idea that AI technologies are key to ISP Customer Care and Customer Experience success?

Not only are AI technologies key, they are essential! Look, there are already billions of connected devices produced by thousands of companies operating in hundreds of millions of homes at this very moment. There is no way that any ISP, or even association of ISPs, can create or obtain an accurate index of all the devices, services, interoperability requirements, software versions, etc. that are in play. This industry is like the Wild West. People buy any device they like for any purpose, configure it and expect it to work. The industry is growing way too fast for ISPs to keep up with all the needs of their rapidly transforming customer base, including essential 24/7 services such as security and healthcare. AI, specifically, machine learning systems, are necessary to deliver the visibility, malfunction detection and analysis of all the connected activity that’s going on in all the connected homes—and in real time! The AI needs to be able to detect problems the moment they happen and to analyze and remedy them pronto before the end user even experiences them.
When the nature of a problem is new to the AI, at the very least, it has to supply a real-time analysis of what has transpired to Customer Care to enable human agents to fix the problem. The days of lengthy support calls where the support agent tells the customer, “Try this and tell me if the green light is blinking,” have to end. It costs too much and takes too much time. Today, Customer Care resorts to an expensive policy of “replace the router”, but, in fact, the majority of returned routers are found to have no problems, so the care is not only costly but ineffective. In the rapidly growing connected- home industry, end-user support activities are going to absolutely crush ISPs if they don’t adopt AI to help them through all this.

Do you think that modern troubleshooting in the ISP world would need AI solutions for cost reduction and customer satisfaction?

In a word, “Yes”. In two words, “Hell, yes!” To be effective for troubleshooting in the ISP world, AI solutions will have to provide 5 vital capabilities:

  1. Visibility. Today, the ISP cannot see what is going on in the connected home. The AI will have to create and maintain real-time topologies not only of the individual devices but of the services they consume.
  2. Detection. The AI must be able to distinguish between normal and abnormal behaviors regardless of device and service, taking into consideration environmental and other fluctuating conditions.
  3. Analysis. Upon detecting any behavioral anomaly, the AI needs to automatically analyze it to discover if it knows this problem or if it is a new problem that must be learned.
  4. Location. Problems can occur anywhere along the Service Delivery Chain within and beyond the home. After the analysis, the AI must determine where in the chain the root cause of the problem resides so that an effective response can be constructed.
  5. Remediation. Ultimately, the AI needs to autonomously fix problems as they occur. This is the only way to guarantee smooth customer experience and satisfaction in most cases. However, since more and more devices and services are pouring into the home, there is always more to learn. So, at the very least, the AI needs to support Customer Care with prompt delivery of its accurate analyses so that human agents can quickly understand and solve problems, slashing costs and contributing to customer satisfaction.If the AI does not deliver all these services, troubleshooting costs will skyrocket and customer satisfaction will plummet.

Thank you, Omer, for sharing your important thoughts on this emerging subject.

Check out the interview with Stacey Higginbotham-  IoT blogger and aficionado