IBM is jumping into the digital assistant market with its own voice-activated Watson Assistant, an artificial intelligence-based system that takes a different approach than other players in the market by targeting businesses. Unlike Alexa, Google Assistant, Cortana and Siri, Watson Assistant, which was announced last week, is a white-labeled service that – from the user perspective – runs in the background on the IBM Cloud.
There is no Watson Assistant wake-word, such as “OK, Google” “Hey Siri,” or “Alexa.” Nor are there plans for a Watson-branded device to be sold in stores.
That’s because IBM is selling Watson Assistant directly to businesses rather than consumers, said Bret Greenstein, IBM vice president of Watson IoT. “We are coming at this from a market point of view that is more friendly to enterprise clients,” he said.
The thinking is that IBM Watson can provide the A.I. assistant technology on top of which a company can put its own brand voice, customizing its features for specific uses and deploying it in a variety of connected devices.
From a consumer perspective, conversations will be with a business – be it a retailer, car company or any other type of organization – rather than Watson.
“First and foremost, we tailor to the client’s brand, so even though everything will be powered by Watson, you won’t be talking to Watson, you will be talking to your BMW or whichever companies embed this – you will be talking to them,” Greenstein said.
How does Watson Assistant work?
IBM already sells conversational A.I. products as part of the Watson portfolio; Watson Assistant combines some of the company’s existing technologies. This includes Watson Conversation, a framework for building chatbots and applications that require natural language processing that’s being rebranded as Watson Assistant, and Watson Virtual Agent, which will be retired.
IBM, in its announcement, offered a rundown for developers of the new features within Watson Assistant. (Early access to the service is now available via IBM’s beta program.)
In addition to providing a framework that allows developers to create their own Watson Assistant skills, IBM has already built three vertical-specific skills – for industrial, automotive and hospitality uses – businesses can apply branding to.
A “cognitive profile” can then be built around a user and used to make proactive recommendations based on past behavior and preferences, said Greenstein. For example, a Watson-based car assistant might suggest a coffee break after three hours of driving and then suggest a nearby cafe.
Raúl Castañón-Martínez, a senior analyst at 451 Research, said that targeting industry verticals allows IBM to communicate the benefits of the Watson Assistant to customers more effectively.
“A key challenge for IBM is market education; helping organizations understand the use cases in which A.I. assistants can be relevant for cost reduction, customer service improvement, workflow automation, etc.,” he said.
“[IBM’s] approach, which focuses on narrowly-defined verticals and use cases, is a smart move that can help them address this challenge.”
Brandon Purcell, a senior analyst at Forrester, said that for automotive firms Watson Assistant “makes perfect sense – auto makers like BMW are definitely looking to empower drivers with voice assistants and do not want to be held captive by Amazon or Apple.”
Hospitality, however, will be “a tougher sell,” he said, as many large hotels have already begun equipping rooms with Apple and Amazon devices.
Watson Assistant in use
Other potential customers for Watson Assistant include connected device manufacturer Harman, a subsidiary of Samsung that has created a digital cockpit system fitted into a Maserati GranCabrio, and Royal Bank of Scotland, which will use the A.I. assistant in contact center calls.
Greenstein also pointed to a deployment at Munich Airport, where Watson Assistant is embedded into a Softbank Pepper robot that answers questions and provides guidance for travelers, such as where gates are located and why a flight might be delayed.
“They are not speaking to Watson, they are talking to Munich Airport and they are getting help as they travel,” he said. That makes Watson “very different from the consumer assistants who insist on their own brand.”
However, there could be drawbacks to IBM’s approach. Services are generally harder to sell than products, said Arnold, “so without a physical endpoint, IBM doesn't really have a strong play directly with end customers.”
He said adoption of Watson Assistant will “largely depend on how willing and/or able businesses or enterprises are in rolling this out. With offerings like Echo, the value proposition is intuitive – it's ‘your' Echo, and you use it as you see fit, and if you're a fan of the Amazon brand, you'll probably want to use it right away. IBM doesn't have that kind of end user affinity, so driving adoption is really in the hands of the economic buyers.”
Data privacy is a focus
Another way IBM hopes to differentiate Watson from other A.I. assistants is by emphasizing data privacy.
Greenstein said Watson Assistant offer greater control over how customer data is used. “IBM’s business model is that clients own their own data,” he said. “It is our objective to help our clients get value from that data to learn more about their users, but it is entirely their data, so we don’t monetize or own that data. Data privacy is essential.”
In practice, this means Watson Assistant users can choose whether or not their data is used to train Watson’s machine learning models, the principle being that the more data fed into Watson, the smarter it becomes. IBM claims to be the only provider that allows this choice. “So when you train your model, you can own and protect the insights you have generated,” the company said.
While data privacy is a key issue for businesses and consumers, restricted access to data has its downsides, too.
“Data will determine who wins and loses in the virtual assistant race,” said Purcell. “Alexa and Siri already have enormous user bases, meaning they are continually getting better at understanding the spoken word. IBM does not have the same magnitude of training data to hone its speech recognition skills.”
He argued that the privacy selling point could also be a “significant drawback, in that it will impede Watson Assistant from learning at the rate of its competitors.”
There are also questions as to what degree IBM’s focus on data privacy will slow Watson Assistant’s uptake. Jon Arnold, at J Arnold & Associates, said that security “seems to be a real strong suit for IBM, but that may not yet be a big selling point or adoption driver.
“It certainly will be for enterprises and businesses in order to protect their data, but probably not as much for end users,” Arnold said, though he noted that attitudes may be shifting, particularly in the wake of the recent privacy controversy involving Facebook and Cambridge Analytica.
Multiple voice assistants can co-exist
With Watson Assistant, IBM is joining a raft of more established smart assistant platforms. Greenstein said that various consumer A.I. assistants already on the market, and individual companies creating their own branded tools, “we are going to have many assistants in our lives.”
So while Alexa, Siri and others may have greater awareness among the public, Watson Assistant will have a place too, albeit less visibly.
Greenstein said that a car owner “may still want to talk to a consumer-facing [voice assistant] in their car, but the full car experience and the car ownership and relationship should be with the brands that are selling the car to you.”
He added that while some people may doubt there is room for multiple A.I. assistants, “I also think we couldn’t afford to have just one.”
Castañón-Martínez said that the market for AI assistants is still emerging and there are not that many assistants currently available. While the number is likely to grow, “in many cases, they will be embedded within the user interface, and will not have a distinctive personality like Siri, Alexa or Cortana.”
There still remains the question of interoperability between the various assistants. Greenstein said that will depend on consumer demand.
“I believe they will [integrate],” he said. “There is still a lot to do, because most of those companies don’t really want to open up in that way. But from a consumer point of view, they are going to expect this in the same way that we have mobile apps from multiple companies on our phones…. Some of those companies compete with each other, but the apps are available because the consumers demand.”