Narrowing AI: A Useful Context for AI Innovation, Opportunity, and Investment
I was asked recently if there was a meaningful analog to AI – its pervasiveness, its transformative potential, its power – in the annals of technology, and I answered with a straight face: “Yes. The Wheel. The Printing Press. Electricity. The Internet.”
I truly believe AI is going to be that big, if not bigger, because it is not only going to create new efficiencies and help solve a lot of problems (in fact, it is already), it’s going to shake the very foundations of how work itself – and businesses, the orchestrators of work – are organized.
Already hundreds of smart startups are building applications on top of AI to solve specific business problems. This is key. While it’s easy to be seduced by the massive processing and reasoning power of IBM Watson-like AI systems, I’m less enthralled with “AI will cure cancer” homilies than I am by the laser focus of the best of today’s AI solutions and the business value they are creating.
AI needs to be uniquely tuned to specific enterprise capabilities. In other words, with AI, the narrower the application the better.
It is through this lens of narrowness that I’d like to examine three categories of AI applications that stand out today for their high growth potential and current utility:
1. Visual AI
2. HR AI
3. Enterprise ‘Applification’ AI
My fascination with “Visual AI” is rooted in a monumental demographic shift – the number of people already online and the large underserved population that will soon be. According to We Are Social and Hootsuite’s latest Digital 2019 report, nearly 4.4 billion of the world’s 7.5 billion+ people have Internet access, with that number growing by more than 1 million per day since January 2018. As more of the underdeveloped world gains Internet access, expect these numbers to increase.
This means businesses of all kinds need to invest in the digital presentation and delivery of their services, whether informational or commercial, spanning e- and m-commerce. The digital front door to all these services is the UI, or user interface.
Prior to AI, UI quality and compliance testing was done manually. When a global brand developed its e-commerce site, the protocol would be sending it to staff overseas to QA check the site’s look, feel, navigation, and operation.
That may have been good enough when 20 percent of a brand’s customers interacted with it online, but now with 80 percent or more caring about – and depending on – that digital experience, a better quality guarantee is needed. Visual AI can look at the same website and instantly detect a defect with 99 percent accuracy when compared to the model, determining whether a visual interface measures up to human needs, regardless of language. And it can scale, working to ensure the UI is visually perfect across any device, operating system, or screen.
Visual AI also improves upon the current “if you like this, then you might like that” e-commerce norm, a.k.a. behavioral modeling. Say you are looking at a blue shirt on your favorite retailer’s web site. Visual AI sees the shirt for what it is – its color, print, texture, sleeve length, all its characteristics – and populates your view with precise visual equivalents of that shirt.
In short, behavioral click-stream data is being replaced by visual AI techniques, enabling “visual search.” This delivers a more personalized and engaging online experience, resulting in documented conversion-rate increases of 6-10 percent.
Moving from e-commerce to human resources may not seem the likeliest transition, but, remember, when looking at AI narrowly, there are many ways and places in which it can deliver business value. So let’s discuss how AI is making it easier for job seekers and hiring companies to find each other, and in some cases changing the rules of engagement altogether.
One West Coast firm has inverted the model of hiring, putting the candidate in control. Its thesis is if it can provide a platform that attracts the best and brightest data scientists, marketers, and technologists from all around the world, then the Amazons, PayPals, and Capital Ones of the world who are desperately seeking that talent will actively hire from it.
It’s the first marketplace that enables candidates to self-guide and self-progress. It’s also the first marketplace that allows for full geographic transparency of available jobs. AI works constantly behind the scenes to make the product better, and the product is the candidate. So AI detects the delta between a candidate’s capabilities and the hiring company’s needs, proactively suggesting courses and tests for the candidate to complete in order to clear the hurdles to attracting Uber or eBay.
A different kind of HR AI application is turning an East Coast startup into a rising superstar. In this case, the company uses AI to morph the candidate’s experience on the websites of large hiring companies. The system learns as it goes, seeing where a given candidate spends most of his or her time during their site visit, directing them to jobs that candidates with similar interests exhibit.
Here the AI engine is figuring out what makes a given candidate tick based on where they go and how long they spend on the site and customizes their experience accordingly. The more the candidate is on a given site, the more the application learns about them, closing the gap between what they want and what the hiring company needs. The beauty of this system is that it works equally well for a company’s existing workforce, ensuring more qualified employees see in-house opportunities that are right for them before they churn.
In this case, AI has taken the back-office process of recruiting and enhanced it to enable companies to hire and retain the best talent.
Enterprise ‘Applification’ AI
The last AI vector I’d like to discuss focuses on what I call enterprise “applification.”
For decades, applications were rooted in a database paradigm. The database was literally the “base” upon which enterprise applications were built. You bought popular database software, added some screens on top of it, and out came an application that solved a business problem.
Today AI is enabling a different kind of application development, and one untethered from the database, where AI and the data it and other enterprise applications produce will become the new substrate from which a new and exciting class of applications will be built.
For example, one leader in this space takes all the data in a CRM application such as Salesforce and ingests it into an AI layer. It then runs algorithms on this data and predicts customer issues, best-path-of-resolution, and product-support intelligence, resulting in net new applications that enable its clients to better compete in a Subscription Economy where customers will churn unless taken care of with white gloves.
At its heart then, I see AI changing application development altogether, enabling a new class of AI-fueled applications that can supercharge business performance.
* * *
Even as I write this article, startups are innovating with AI in new and exciting ways beyond the scope of what I have just outlined. But that’s the point. Approached narrowly, there is room for many, many upstarts to focus and unleash the power of AI, creating net new markets and business value.
At the same time, not every company in this red-hot category will be a winner. Traditional investment principles still apply – great people, great teams, and great ideas are all needed. And so is great technology. In AI not all algorithms are created equal. No one wants a self-driving car that is just 80 percent accurate.