The Rise of Self-learning Software (SLS)

Imagine it’s five minutes before a meeting. Your smartwatch, without prompting, sends you key points. While in the meeting, you take notes. Those notes are instantaneously absorbed by the system, then collated with relevant prior meetings, files and communications, in order to better prepare you for the next meeting.

Imagine a set of smart sensors placed throughout a city that feed data to deep-learning systems. The systems then predict where traffic jams might occur and recommend route changes to the trucking fleets that keep inventory moving and businesses running on schedule.

This is the promise of what I’m calling "self-learning software." We've already seen it take root in consumer applications, and in the coming years it will help enterprises work faster and smarter than they ever have before. Its potential to increase productivity is so great, in fact, that it will generate unprecedented growth in enterprise software.

Here are some initial thoughts on what I believe to be the dawn of an exciting and enormous market opportunity.

The promise of self-learning software

The first era of the enterprise software industry was on-premises software, otherwise known as "shrinkwrap" software. The software is installed and runs on dedicated databases and machines on the premises of the person or organization employing the software. This initial period in the industry was dominated by the likes of SAP, Oracle, PeopleSoft and IBM.

With the rise and growth of the web, the need for shared servers and greater connectivity within companies pushed businesses toward software run from remote locations. We’re currently in the midst of this second era of enterprise software, "software as a service," in which agile cloud computing has replaced locally installed applications. Big SaaS players include Workday, Salesforce and NetSuite.

I think that the next great era in enterprise software will be the era of self-learning software (or SLS). Born of the innovations of Big Data and possessed of a new net intelligence layer, self-learning software will have huge impacts on productivity across all departments of an enterprise. The effect will be so seismic that I predict that SLS will double the size of the enterprise software industry by 2030. But, first ...

What exactly is SLS?

I define SLS as enterprise software injected with machine learning — the branch of computer science that explores how to enable computers to learn from and make decisions based on data without explicit programming instructions. The basic building block of self-learning software is the ability for a system to learn based on experience, make inferences from disparate signals, and then take action in response to new or unforeseen events.

With much of today's enterprise software, a person is required to make inferences repeatedly based on any number of data points — validating positive compliance with security protocols, for example. But the complexities of business environments often cause employees to be cautious, or even feel paralyzed, when action is required.

Software that functions more autonomously liberates companies from having to codify the rules of engagement and escalation — meaning, if an employee can reach a conclusion once and then train an SLS machine how it reached such a conclusion, the machine will learn by itself how to arrive at similar outcomes in differing situations. The human team will, thereby, be empowered to act more quickly and confidently, and freed to focus on higher level problem-solving.

A colleague’s 65-year-old immigrant mother only recently discovered the joys of predictive text. Now, when she sends her children short messages on her iPhone, she loves being liberated from the tedium of having to peck out every single letter, or having to consult her Chinese-English dictionary to spell a word.

I bring up this story because one (admittedly oversimplified) way to think of self-learning software is as predictive tasks. Your SLS-powered enterprise software will anticipate rote or foreseeable workplace tasks that need doing and simply auto-complete them for you, or get you several steps closer to finishing them yourself.

One of the best current examples of a predictive computing platform is Google Now. Based on a mix of location, calendar and both Google and email searches, Google Now anticipates what information a person will need next and presents it to them automatically. This type of frictionless experience will enable people to spend less time looking and more time doing.

Another way that self-learning software will improve productivity is in customer-data collection. The reality is that relying on employees to enter data into a CRM is inefficient at best. The data are often input only when the employee has a moment to spare, which is an increasingly rare occurrence for most of us. Self-learning software will solve this problem by mining data and context from mobile devices and the Internet of Things to automatically compile an accurate, detailed portrait of customers’ habits.

And the more data a system collects, the bigger the back-end brain becomes. The cloud can essentially crowdsource data and extrapolate the insights to every other person in the network. Companies will learn from other companies, thereby driving up the intelligence of their SLS platform.

They might be giants

Sometimes, as an investor, you just have the sense that something big is about to happen.

When enterprise software was a fledgling industry in the mid 1990s, it was clear early on that the market opportunity was huge. The market has nearly tripled to $630 billion since then. There are currently 12 times the number of companies buying software than there were just 10 years ago, and each of those companies is spending four times more on software than their 2005 counterparts. But a new generation of workers will create an opportunity that even the most optimistic investors of the ’90s could not have seen coming.

Millennials, the first generation of digital natives, are now entering the job market, and they’ve started to question the fundamentals of current SaaS solutions. They expect more from their applications. They want their inboxes to sort and label emails, the same way they expect Spotify to predict their favorite songs. They’re accustomed to one-click, same-day-delivery shopping with Amazon Prime, and don’t have the patience for the dozens of steps and several days that it sometimes takes to schedule a meeting and book a conference room.

Only open-source machine learning algorithms with the technology to collect more data and market to individual consumers — a.k.a self-learning software — will be able to satisfy those expectations.

So, who will win the next wave of enterprise software? Goliaths like Facebook and Apple will undoubtedly get into the game either directly or through acquisitions. Amazon’s AWS will be a force to be reckoned with for years to come. And with its purchase of LinkedIn, Microsoft now has the social graph, enterprise software expertise and massive resources necessary to be the once and future king.

I think, however, that this is a rare chance to innovate in a giant market where everything is still up for grabs. And, as usual, I’m putting my money, figuratively and literally, on the innovators and entrepreneurs.

Founders of the new wave of self-learning software companies are coming from outside the arena of traditional enterprise software, including many from the consumer web companies. This gives startups an inherent advantage, because machine learning and end-user experience with these applications are part of their DNA. Moreover, while traditional enterprise software developers have to attempt to awkwardly upgrade machines built for a different era, startups can build anew, with smarter, adaptable systems not weighted down by the old technology.

Looking ahead, self-learning software will be the one thing that distinguishes legacy enterprise applications from modern ones. And I predict that in 10 years, every new enterprise application will be self-learning at its core — leading to productivity levels greater than what we've seen from mobile and Big Data combined.