Cobotics: a $2,900,000,000,000 Pie — Machines and Humans in Perfect Harmony
“Augmented Intelligence,” or cobotics, AI designed to cooperatively assist — rather than replace — humans, is an exploding market, which Gartner estimates will deliver $2.9 Trillion in business value in 2021. Four comma numbers are typical of GDPs, not commercial markets — Gartner feels that AI’s embryonic stage and profound potential invalidates traditional market sizing.¹
Undaunted, Grand View Research, took a traditional stab at sizing one key segment: Intelligent Virtual Assistants (IVA’s) (which Gartner calls Agents), estimating that spending will grow at 40% CAGR to reach $25 billion in 2025. Their 3 comma estimate aligns with Gartner in pegging IVA as a humongous market — one of the fastest-growing ever.
Such an enormous market, encompassing consumers through Enterprises, naturally has a spectrum of capabilities, and since virtual assistants nominally replicate human jobs, the spectrum of capabilities for IVA software is especially broad — the Bureau of Labor Statistics tracks over 1,000 occupations. Plus, human capabilities vary widely — after all, Jeff Bezos and a dude selling bottled water on the corner are both entrepreneurs.
The entry point to this broad market would be rules-based chatbots, except they aren’t actually AI at all ². The distinction is more than pedantry. IF-THEN bots struggle with ambiguity and language-specific rules, so hand-crafted rule-based systems are error-prone and clunky, which resulted in a chatbot image problem.
Most newer chatbots use AI-based Natural Language Processing (NLP) to reliably understand the written and spoken word, and — if well designed — converse within narrow domains as well as humans. Even so, after years of overhype and unintentionally hilarious failures, the term chatbot elicits more sneers than wows. In fact, Facebook avoids the term in its just-rolled-out Messenger chatbot, calling it Lead Generation, an automated question experience.
For gathering information and routing request, message-based chatbots work well, typically delivering better results than web forms. Everyone hates forms — the abandon rate for gated content is over 80%. It also does’t hurt that executives with checkbook authority have the highest propensity to initiate conversations via messaging.
In human terms, these basic chatbots are like trade show booth babes. They’re better than the empty booth of traditional web sites, but NLP-powered agents are capable of much more.
Everyone knows Siri, Alexa, Cortana, and other general purpose NLP-powered digital assistants (DAs). Like Matt Damon in Good Will Hunting, they’re wicked smart, but — so far — stuck in menial jobs that don’t leverage their formidable capabilities. While major DAs aren’t conversational, Amazon Lex, Google DialogFlow, the underlying AI platforms — can all maintain context, enabling intelligent human-like conversation (so does IBM Watson Assistant). Also, their newest synthetic voices, using deep generative neural networks trained end-to-end to model raw audio waveforms, can fool humans in the right settings.
With NLP-based, voice-enabled chatbots, the next wave of e-commerce is conversational, via smart-speakers and connected cars. Like an Honor Roll teen working at Foot Locker, intelligent agents have amazing potential. Tapping it requires training, experience, and the right opportunity.
This is where Enterprise agents come in, training these same AI-engines to engage in conversation and to execute tasks. With this training, plus appropriate integration, Enterprise virtual assistants are already capable of replacing humans in specific roles, like appointment setting, call screening, or note taking. As an example, 100,000 callers to an auto warranty sales line handled by Vocinity CCAs resulted in the same number of sales as an offshore call center — at 40% lower cost.
Like Pepper Potts (Tony Starks’ Iron Man EA), who ultimately took over as Stark Industries CEO, virtual agents will ultimately do far more that menial tasks — but even menial tasks are would be welcome, In fact, a recent Cisco survey shows Enterprise teams think they’re a godsend — unlike Tony Stark, rank-and-file employees live under a tyranny of tiny tasks. Stuck with all the pesky work that secretaries once did, workers desperately need someone (or something) to step into the administrative void. Cisco, which just released the WebEx virtual meeting assistant, concludes that teams are “pounding the table” for virtual assistants to take on administrivia like:
- taking meeting notes (requested by 60%)
- providing alerts to upcoming meetings (52%)
- checking attendee calendars and scheduling follow-up meetings (50%)
- transcribing the meeting (50%)
- sending follow-up emails (49%)
- distributing action items (43%)
- adding new attendees, even after the meeting has started (38%)
Mad Men’s Peggy Olson is another famous fictional secretary who climbs the corporate ladder, in her case to copywriter. AI agents have already made that leap — Persado’s AI agent, is taking over headline writing duties for JP Morgan Chase.
Last time, I described the AI models which vest agents with super-human perception to discern personality, intent, and sentiment — essentially, artificial intuition. With that, agents can tailor persuasive messaging with perfect linguistic precision, “warm up” sales prospects using verbal, auditory and/or visual psychological priming, and orchestrate hyper-personalized content presentation. Putting it all together, it’s easy to see why Gartner says: “Customer experience represents the majority of business value through 2020, when new revenue takes over to gain prominence for the next five years. Cost reduction, while important, will not be a point of differentiation from most products and with most users.”
Next time we dig into this new paradigm: cobotic selling,
¹ Gartner Forecast: The Business Value of Artificial Intelligence (published March, 2018)
² — AI in the common vernacular, eg machine learning and neural networks https://www.theverge.com/2016/2/29/11133682/deep-learning-ai-explained-machine-learning
