Late to the party, Google's cloud leans on machine learning and AI to capture enterprise market

To get a perspective on Google’s cloud services in India, talking to Debashis Roy, vice president-IT of CESC, the Sanjiv Goenka-owned power distribution company, helps.
A couple of years ago, every company in the $2.7 billion RP-Sanjiv Goenka Group had a different domain name for emails, lacking uniformity. Roy migrated all employee emails — 6,500 of them — to Google Cloud. For a legacy company, which traces its roots to pre-Independence India, having information stored in servers in remote destinations owned by Google, in different parts of the world, was a big culture shift.
After putting emails in the cloud, Roy has taken the next step. He is running two pilots using Google’s machine learning (ML) and artificial intelligence (AI) services to improve customer satisfaction, reduce operating time, and boost revenue.
Take the process of collecting electric meter readings prior to cash collection from customers. It was mostly a manual data-entry job. In one of the pilots, Roy has given meter readers Android devices and installed a proprietary app, which allows them to either key in the reading, or speak into the phone, or just take a picture. In the backend, Roy explains, the software decodes the text, voice or picture, and generates the bill. And, what’s more, the meter reader can speak into the phone in his native language.

In another pilot, Roy’s team is using Google’s ML and AI tools to detect faults in the power supply network. For example, if the same person registers a similar complaint twice, and faces the problem the third time, the software (with the help of ML) will directly prompt the engineer to the problem area instead of going through the process of investigation. “That will reduce power cuts, and in turn help in increasing our revenues. It will also improve our customer experience,” says Roy.
Soon, these pilots will be implemented across the company, into CESC’s daily operations. But, none of the programming, or the AI or ML tasks will happen on CESC’s IT infrastructure. Rather the space, the tools and the services are being rented from Google on a per-minute billing basis. The higher the use, the higher the rent, which proportionally increases with the number of services being rented at any particular time.

Google Cloud — Last but not least?

Although Google’s entry into the fast-growing (and immensely profitable) domain of cloud storage and computing services has been somewhat late, which has allowed competitors, especially Amazon Web Services (AWS) to take a big lead, Google’s cloud services are being touted by the company as a game-changer because of its built-in AI and Machine Learning capabilities.
While AWS and Microsoft (Azure) offer renting and processing data on the cloud, Google provides a lot more to its enterprise customers, which includes hearing, listening, translation, sifting through tons of data to perform tasks, using ML and AI. It also launched a new AI chipset during its recent Google I/O conference, which it has been using internally to run its own AI and ML services. The new chip not only runs neural networks (a computer that is like the human brain and nervous system, and is ‘self-learning’), it also trains them — capabilities that Amazon and Microsoft do not have at present. Even Intel has been struggling to launch its AI-enabled processor.

To be fair, Amazon and Microsoft allow graphics processor-maker nVidia chips to power AI and ML projects for their clients, but none of them have bespoke chips like Google does.

Google’s processor, too, is available as a cloud service, which means the processors cannot be bought off the shelves and installed in individual office servers. Rather these chipsets will be placed in the servers and be used as a cloud rental service to run ML offerings. Google had developed these chips to run its own AI bots.
“If you have billions and billions of (GB of) data, and you want to run that and train the model, we do a great job there. That differentiates us,” says Mohit Pande, country head, India, about Google Cloud.

Google is bringing all its cloud services to India. Reason: In the next four years, the public cloud business in India will more than double. According to research firm Gartner, the market size will grow from $1.32 billion in 2016 to $4.10 billion in 2020.
Globally, Google still trails Amazon and Microsoft. Amazon Web Services continues to be the largest with 37.1% market share, Microsoft Azure 28.4%, and Google 16.5%. India numbers aren’t available, but experts said it more or less follows a similar pattern.

Using AI and machine learning effectively

So, are AI and ML Google’s biggest weapons to transform the enterprise cloud services business and fend itself from competition?
Both ML and AI are required for three things, explains Pande: when a lot of data is required to be processed; need of algorithms to built a framework for ML; and a lot of processing power. Google internally handles all of those — it has seven businesses, including Gmail, Maps and YouTube, with more than a billion users each.
Google does not only want to make its own services better by using AI and ML, it wants small, medium and large companies to rent its AI and ML platforms to run projects.
In India, an insurance company that can’t be named is using Google cloud with ML capabilities to estimate damages for accident claims through pictures, instead of sending a surveyor. Google’s vision API  figures out the damage and generates a quote.
Customer service teams are using Google’s cloud-based services to understand the mood of the caller, if he/she is angry, sad, happy or irritated, using natural language API. These data-sets can be further used for internal training purposes.
In other cases, Pande explains that companies can use video intelligence APIs to search and tag videos automatically. “The software understands if you upload a clipping… if the picture is of a pizza delivery point, or a man in a red shirt, or on a yellow scooter and it will give you the ten thousand tags and it will take you exactly to the position where you want to be,” says Pande, quoting the example of a restaurant using the cloud services to streamline home-delivery of food.
Financial companies are using Google’s cloud services for fraud and theft detection, healthcare companies are using it for genome research, oil and mining companies for geological explorations, and retail and e-commerce companies are using it to figure out the next logical purchase for customers.
Some of this is being used by online real estate company NoBroker to identify rooms, apartments, find the owner’s credentials, and figure out the condition of the property. “No one goes to check these properties. Google scans through the images, names the bedroom as a bedroom. We don’t engage any human being to do this,” says Akhil Gupta, CEO and founder of NoBroker.
Google has helped Gupta reduce 12% of its staff and engage them in other work. He also says that even if the company expands into 50 or 100 cities, he will not have to hire people to do this job.
Most of Google’s services come with a hint of AI and ML. Even the basic Gmail has been upgraded with new features of automated replies, depending your mailing behaviour. Google Maps has AI and ML layers. For example, an insurance company can decide on a customer’s premium based on where she stays in Mumbai. Using Google Maps, and adding AI and ML to it, companies will be able to figure out flood-prone areas and assess damage.

“Amazon and Microsoft help in moving businesses into the cloud, they don’t have AI and ML as services. Google is focussing on futuristic technology, like making humans immortal…,” says Hardik Parekh, CEO of Searce, a Google implementation partner.
Analysts, too, feel that Google has changed the game in serving companies. “AI and ML are core platform capabilities in all future apps… TensorFlow is emerging as the favourite among system integrators as it is really one of the few neural networks that work. Others (Amazon and Microsoft) are trying but are still behind,” says R ‘Ray’ Wang, principal analyst and CEO of Silicon Valley-based consultancy and advisory firm, Constellation Research. TensorFlow is Google’s Machine Learning software.
“TensorFlow is emerging as the favourite among system integrators as it is really one of the few neural networks that work. Others (Amazon and Microsoft) are trying but are still behind” — R ‘Ray’ Wang, CEO, Constellation Research
Google’s focus towards companies changed only when Diane Greene, founder and CEO of VMware (until 2008) joined Google, and beefed up the enterprise business. “Google thought enterprise was for people 50 years and above, which was pitiful,” Wang adds.
In the last two years, Google has streamlined its cloud computing business, and has integrated AI and ML in all its services, explains Pushkaraksh Shanbhag, senior research manager (Asia Pacific) for IDC’s Services and Cloud Research Group. “Google was late in identifying the market, but right now the cloud services market is just a price war. This is a starting point, and in the next two years it will be all about differentiation,” he says.
Google’s original mission was to organise the world’s information and make it universally accessible to all. The enterprise business is not an exception, and Pande knows that well enough.