While attending a recent Salesforce.com Basecamp for Customer Service Pros conference here, I was particularly taken by a keynoter’s observations on our collective progress toward incorporating AI into business software like CRM.
Later in the conference, while viewing some of the product demonstrations, wherever the word “AI” appeared onscreen, I or someone would poke their hand up to inquire and we were almost uniformly answered with some variant of “it’s coming”.
For all the hype around AI, the nearest we have come is in programming our systems to model best practice, then prompt users in performing patterned workflows that hew to that programmed best practice. Call it computer-assisted pattern recognition.
This is not AI, but it’s a start. Using DARPA’s definitions of the Three Waves of AI, it appears many of us are still broadly in the early “First Wave of AI”, in which human-composed (hand-cobbled) workflow rules guide software users and customers through choice architectures with prescribed, rule-based workflow steps. We can perceive patterns, then use reasoned judgment to either follow the pattern or justify exceptions. If exceptions to policy or rules are frequent, then the policy is adjusted to reflect the pattern of practice. Policy and practice are close partners in the perpetual dance of future alignment.
AI Wave Number One: Handcrafted Knowledge
Calling it AI is, in my unvarnished humble opinion, a bit flamboyant. In truth, our “Artificial Intelligence” is stuck in the previous century, at DARPA’s “Wave Number One”, where a PC could recognize our basic input patterns (if x occurs, y is likely to follow), and reflect those patterns in tools like spellcheckers, templates and workflow systems, tuned over time by user experience. Recent advances in cheap processing, distributed data and voracious coding have improved matters to where a user can create workflows in a codeless, drag and drop fashion. Twining together all those 5,000-plus pieces of commercial marketing software is everyone’s grail quest – and there’s even an app store for it. Hello, Zapier.
Today, our systems can support our ability to perceive and derive value by improving our reasoning and judgment, spotting trends and drawing inferences based on historical or near-real-time data flows. This is where the largest untapped opportunity looms for organizations to achieve savings through efficiency by tuning their tech stack in sales, marketing and service. People and time do not scale, whereas a system can instantly scale to distribute workflows and data interpretation to any number of customer facing people, and even extend that capability into the hands of customers themselves.
Is your business technology supporting you in this way? If not, consider yourself a laggart in danger of losing big. Put simply, in life there are 3 types of people: those who make things happen, those who watch things happen, and those who wonder what happened.
Today, thanks to cheap computing, massive data “blooms”, and distributed networks, we can now amass, consume, configure and present interactive display reports on top of large datasets to help us understand it in self-driven configurable dashboards. Get some! We can help.
AI Wave Number Two: Statistical Learning
What’s coming? Looking again at the DARPA definitions of the 3 Stages of AI, we can next expect to see engineers creating statistical models for specific problems and training systems to solve those problems, once again using big data as the source material for the training exercise. Even this stage, however, has its limits. For example: Showing a computer thousands of cat photos can eventually train it to recognize a cat with high accuracy – not flawlessly, but reasonably well. Consider, however, that a 3-year-old child can recognize a cat flawlessly after only meeting the family cat and the neighbor’s cat, and looking at a sketch drawing of a cat, and will point to the cat cartoon and say it’s a cat. Thanks, brain!
Computers, meanwhile, face challenges in recognizing a handwritten number 8. The myriad of writing styles, speeds and implements confounds the problem. This diversity of inputs and human approaches is the biggest challenge to UI developers. Confusion over the validity of our databases is often caused by uncertainty about what the user intended to do or say when they input their data.
A slight 1% inaccuracy of input today can result in an outsized unreliable output. This is also the stuff of internet memes and fake news. Anybody can publish a single tweet to a vast global audience. The pace at which all that published-rubbish (“pubbish”?) speeds past us confounds our efforts to filter and validate truth. Our resulting, collective judgment errors can result in an outsize misinterpretation of fringe views as central guidance. Absent a moral compass, a distorted maniacal map could lead many, unaware, off an ethical cliff. Upshot: to trust your data, you need to regularly audit. Shameless plug: Fan Foundry excels at this.
Wave Number Two will take some time to get right. I’d give it a decade or two to reach prime time.
The Third Wave of AI: Contextual Adaptation
In this future (certainly not the present), systems can reliably explain real life phenomena. They can perceive, learn, reason and even abstract. They can predict success or failure. They can understand why or why not. They can know why you made a mistake, when to trust your judgment, and can guide you on that optimal path of interpretation and judgment.
The big challenge here is for us to surrender our trust to a cyborg partner. For now, though, it’s a bit out of reach. To quote the articulate supercomputer HAL from the movie 2001: A Space Odyssey: “Sorry, Dave, I can’t let you do that.” Codicil: “Not yet, anyway”.
What’s your Sales, Marketing and Service challenge? Does it involve people, processes and technology? Perhaps we can help.