The Future Of AI Will Be Stacked

Editor’s note: Rick Collins is the president for enterprise business at Next IT, creators of the Alme virtual assistant platform. Prior to joining Next IT, he served as director of business development for Microsoft.

We are entering an exciting period for artificial intelligence. We’re seeing more consumer impacting developments and breakthroughs in AI technology than ever before. And as Nova Spivack recently argued, it’s reasonable to expect that major players like Apple, IBM, Google and Microsoft, among others, will lead a fierce consolidation effort for the AI market over the next five years.

Indeed, it has already begun.

No doubt these efforts will produce some amazing innovation, and I’m excited to see how Siri, Google Now, Cortana, Watson, and even new entrants like Viv continue to progress and transform our interactions with technology.

However, I also believe these all-encompassing virtual personal assistant projects (VPA), and the race to create the one AI to rule them all, are driving the AI experience in the wrong direction. If we follow that path, we’re potentially risking important advancements for both consumers and companies alike.

In considering how AI will develop into a critical technology for businesses, the idea of the comprehensive VPA seems less and less likely, and even unproductive to achieving the kind of substantial impact that it could have for the modern enterprise. Deeply integrated AI has the potential to transform businesses and industries, from the interface level of the customer experience to the breadth of institutional knowledge that employees are able to access and build upon.

The key to any effective AI deployment for businesses, however, requires a sophistication and expertise that is specific to the industry and the company. We call that domain knowledge. The broad intelligence that general-purpose VPAs possess is unlikely to be able to integrate with, or properly serve, enterprise needs. Yet much like the cloud computing market, the most important advancements in AI will be propelled by the enterprise and how major companies are able to deploy the technology across the world’s largest, most complex systems. We should keep this in mind as we consider how AI should and will develop.

The Lessons of Siri, and the Expectations of “Her”

Apple’s deployment of Siri, and the release last year of “Her,” have probably done more to define our current public perception and expectations for AI than anything else, for better or worse.

While Apple, Google, Microsoft and others race to realize those ScarJo expectations in one form or another, we still have a long way to go before we are even close to the simplest simulation of a ubiquitous, global technology. Apple’s Siri should serve as an important lesson, perhaps even warning, for these efforts, though.

As even Siri’s creators have noted, it wasn’t originally developed to be an encompassing VPA, but rather the kind of strategically deployed, domain-specific assistant for functions around entertainment, scheduling, and information. The ambition of Apple to boil the ocean with Siri, to present it as the kind of “everything-assistant” and interface actually backfired for them and led to a lot of the technology’s significant breakthroughs and achievements being overshadowed by unreasonable expectations.

Siri did, however, set the direction for AI and VPAs, with an emphasis on the personal computing market. Cortana has now taken up the challenge directly.

The problem with this, and other AI plays that would seek to be a singular and comprehensive point of interaction for customers, is that we are drastically reducing VPAs’ effectiveness, their transformative and meaningful capabilities by spreading them too thinly across functions and domains. No VPA can be all things to all users.

If, as Spivack opines, we are entering a period of significant consolidation for the AI market, the resulting risk is that it impedes VPA advancement in terms of depth of intelligence in favor of breadth of deployments. It’s the classic endgame of vendor consolidation.

stacked-ai1For companies looking to plug into and develop upon AI technology, the drive to consolidate into a single platform — whether it be Watson, a Siri or Cortana mobile AI, or Google Now experience — doesn’t just threaten to lock them into an ecosystem, it threatens their brand. For an AI to truly function at the enterprise level, it will require such deep integration with company knowledge, and even proprietary information and data, that the experience necessitates control of the VPA by the company. This isn’t cut-and-dried back-end software, this stuff impacts every aspect of the customer experience. Outsourcing the experience is too big a risk, regardless of any upside from standardizing on a platform.

It’s not simply about retaining the customer experience that a brand has defined, though it is essential; it’s about how your company expertise is best leveraged, communicated, and put to work for the customer. A general-purpose VPA simply isn’t designed to do that.

Aside from the functional dilution that results from attempts to create an encompassing VPA, for brands and enterprises, the stakes are even higher.

Even if domain expertise could be preserved within the one-size-fits-all assistant, we quickly enter into problems of who would own and be responsible for customer data, or data surrounding the experiences. There are no industry-standards for privacy and security in that scenario. So what options will the enterprise have?

Developing the Enterprise AI Stack

What we need instead, and what I think serves as the most realistic and effective development of the B2B VPA market, is a full AI stack that companies can begin building and piecing together to best serve their unique needs and their customers. Just as we think about companies building their CRM or marketing stacks, we’ll begin to see enterprise companies building their AI stacks.

Right now, this is not how the enterprise AI market works. If a company wants to develop a robust VPA that is deeply integrated with their processes, brand, and systems of record, their options are limited. They will reach out to an AI platform provider — whether our Alme platform at Next IT, IBM’s Watson, or SRI — and we work to develop a VPA very specific to their needs and goals. It’s design and development from the ground up, built on the intelligence of the proprietary AI platform of course, but architected for a particular company.

The B2B VPA market is still very specialized, so we have the luxury of working deeply and in very white glove ways with companies that want to develop and integrate full VPA capabilities. But as B2B AI scales to become more prevalent, and more of a priority for enterprise, this single platform and AI provider approach becomes less tenable or even desirable for the companies. The future of enterprise AI will be the AI stack.

But what does this enterprise AI stack look like?

Obviously it must work consistently across every channel and device, which can be accomplished by layering NLP on top of existing services in order to deliver an ‘omnichannel’ experience. Domain expertise and systems of record will be effortlessly accessible, connected and cross-correlated. It may even have components for integration with customers’ own VPAs, but must have the agility and capabilities to transfer experiences over to companies’ own systems when required. There are also aspects such as the branded VPA personality, response channel flow and navigation, and personalization through customer data and history to consider. All of this is completely under the control of the enterprise as it compiles its own stacks.

We can even take this one step further and imagine how various enterprise assistants will begin working together. This won’t happen under a single VPA system, but rather by companies’ AI stacks being able to communicate, route and transfer customer service inquiries from the Internet of Things to the appropriate device channel, to coordinate between a patient and their doctor and their insurer, or even between your bank and your investment portfolio.

Customer experience is everything between a provider and a user. Customers expect your applications and systems can communicate with others to accomplish key tasks when they need immediate assistance. And so the future of big picture AI is an AI stack that can play nice with other stacks, serve the customer experience, and do it all while enriching your operations through data.  It must preserve the deep domain expertise and branded customer experiences in which enterprise companies have already so heavily invested.

Consolidation of various AI components, solutions and services could be counter-productive to allowing companies to build their stacks, or give rise to niche services designed to fulfill specific AI stack functions. Either way, as the enterprise begins building its AI stacks, we need to shift the conversation from one of universal VPAs to singularly focused solutions that can be customized for specific business needs.

Images: Bryce Durbin