Opus Research Intelligent Assistants Conference 2017

Day 1 Recap

This week at the Intercontinental Mark Hopkins Hotel in San Francisco was taking place the Intelligent Assistants Conference, organized by Opus Research, an analysts firm that focuses primarily on conversational commerce, and the merging of intelligent assistants, contact center automation, biometrics technologies, and more.

Leading technology vendors in the intelligent assistants space were attending and showcasing their products and solutions, and many organizations presented case studies. Here is a summary of the first day of the conference.

Seeing IAs as (digital) employees

The day began with a keynote from Joe Gagnon, chief customer officer at Aspect Software. In addition to giving a very inspirational talk (he rightfully pointed out that we too often limit ourselves by believing we can’t do things), Joe talked a lot about the implications of considering intelligent assistants as “digital employees”. Like any other employee, they need to be trained and they improve over time. And this has implications for the workforce already in place. Of course, these digital employees never get sick, they work 24/7 without complaining, never take vacations, etc. But they won’t replace human employees anytime soon, instead they let human employees concentrate on more complex, less repetitive tasks, boosting their overall engagement. That’s what their chatbot Edward for the Radisson Hotel in London demonstrated.

No digital lipstick

This was followed by a case study from Singapore’s DBS digibank. The bank has started its digital transformation back in 2010. They completely embraced the transformation, making it a core strategic decision, ensuring this was not just a cosmetic change (“digital lipstick” to borrow their words). They mapped hundreds of customer journeys, adopted customer-centered design, and agile delivery. Using Kasisto’s Kai technology, they now serve 1 million customers in India. Of those customers, 80% were acquired and serviced without a single human involved (although they have live agent support on the chat channel).

An interesting comment was made by Kasisto that banks are moving away from Facebook Messenger because FB was not keeping its promises regarding security. That’s something we heard from other banks as well during the conference.

Authentication enables intelligent assistants

The third session was a panel on authentication challenges for intelligent/virtual assistants. Panelists from Nuance, Evernym and Acuant pointed out that authentication becomes a customer experience challenge, but that we should not trade one for the other. This session closed on a very interesting question from the audience: how do we authenticate bots to make sure they represent their brands? Nuance claimed they have a “biometrics”-like solution to that, based on the fact that bots also have patterns of communication they can recognize.

Personality is the UX

During the next panel on branding for bots, it was stressed that personas are not just important, they are critical to achieve business objectives (would you hire someone who represents your brand badly?). Hence the claim that “Personality is the UX”, which is pretty strong. Also, the conversational aspect provides great data: by conversing with a brand, customers express what they like and what they don’t.

This ended by a controversial claim by Wally Brill from Google, that metabots (Google Home, Alexa, and the like) are the future, that they will be no need for proprietary bot platforms. But for this to happen, metabot platforms will have to provide a way to customize the persona.


Intercontinental Hotel Group (IHG) presented a case study on their use of Amelia, the intelligent assistant technology from IPsoft. They showed how they re-engineered their service and support interaction model. Like DBS digibank, technology was not chosen for the sake it, but as a strategic approach. Although the system was deployed only in March 2017, adoption is picking up quickly, and 70% of level 3 (most costly) requests are now properly sent and resolved by level 0 (IAs). And real agents/employees who still have the knowledge are used in a “crowd-sourcing” manner to train the bot.

IA Awards

The team from Opus Research then presented the winners of the 2017 IA Awards. They got 14 submissions globally, and were impressed by the maturity of the industry, both from a scale perspective (100,000s of conversations per month handled by individual bots, the diversity of use cases), and the fact that they are almost all text-based. And the submissions clearly demonstrated that the use cases are not so much about saving costs than building relationships with customers.

The winners were TD Ameritrade TED, IP Australia Alex, and Ask Fedex.

The promises of AI

In the last session, Judith Spitz, former CIO for IT strategy and planning at Verizon and founding program director of Women in Tech NY (WiTNY), talked about the promises of AI and what we should be paying attention to. To her, AI is simply what the computer can do today, nothing more. Which means big claims by lots of companies are simply not substantiated. And unfortunately, AI today is trained on highly-biased data. She gave several examples with respect to women. (Like word embeddings, neural network models for NLU systems, making deductions like “men are to computer science what women are to home cooker”!)

I liked this lesson from her: carefully engineering the ASR algorithm can reduce the error rate by a few percents, but carefully engineering the UX can reduce errors by 50%.

Now make sure to read my summary of day 2!