“Smarts,” or the subtle application of machine learning to improve software usability, already powers countless interactions between you and your technology every day. But we’re only at the beginning of the smart collaboration curve.
Data science and machine learning have been dominating the list of fastest-growing career categories for years. In 2017 alone, Linkedin clocked a 9.8x annual rate of increase for members listing “Machine Learning Engineer” as their job title. And according to data from CB Insights, investment in AI startups rose from a few hundred rounds and a few billion dollars in a typical quarter to a staggering 1,245 rounds and $17.3 billion in the second and third quarters of 2019.
There’s certainly a fair amount of hype (and anti-hype) around the potential impact of machine learning. We’ve written before about our philosophy; there’s a lot of ML for ML’s sake out there, flashy features that make for great marketing but have little impact on the daily life of a user. But there’s also a huge opportunity to apply machine learning to alleviate many of the less enjoyable, repetitive parts of our daily lives. Collaboration software is a great example; smarts can make collaboration easier and faster, accelerating work and reducing overhead.
The future of software is smart
For companies that build software, smarts will soon be more than a competitive differentiation, or a way to create great user experiences. Soon, just as it became an expectation for software to come with smooth UI design and an accompanying mobile app, it will be expected that software will be smart. Lacking these capabilities will be as egregious and noticeable as having a clunky and antiquated UI, or having a “mobile app” that’s really just a wrapper on a web browser.
Let’s take autocorrect and autocomplete as examples. Users expect to see autocorrect everywhere. When it fails, it causes instant frustration, even spawning its own “damn you, autocorrect!” meme trend. Autocomplete is relatively new compared to autocorrect, but is already becoming an expectation in more and more types of apps. Soon, even more sophisticated smart functions like intelligent search, notification prioritization, and automated task prioritization will be just as commonplace.
Machine learning that benefits users
The pervasiveness of smarts follows the consumerization of IT – that is, the cycle of technology emerging first in the consumer market and then making its way into businesses. Slack is a great example. As consumer chat apps improved, users became accustomed to easy-to-use chat technology, and came to expect it everywhere.
Machine learning in work management and collaboration software is sure to follow a similar path, and the impact on user expectations of IT is likely to be immense. It’s hard to overstate the impact of machine learning on consumer technology with similar functionality to collaboration software. Twitter’s algorithm-driven feed of tweets, for example, has fundamentally changed the way people connect with each other and share information. The primary functions of the news feed are very closely related to those of collaboration software; the news feed suggests which content to read or watch, who to connect to, and which events to attend and when.
However, anyone with a Twitter account can attest that actual productivity is unlikely to be high on the list of outcomes of a feed-scrolling session. Scanning headlines without actually clicking into many articles is the more likely outcome, which is great for Twitter as it keeps users in the app and looking at ads. And that’s where the impact of machine learning will differ between consumer and business applications.
Machine learning algorithms are designed to achieve a singular goal. The algorithms that power Twitter’s feed are tuned to maximize the value of the ads on the platform, often at the expense of the user. You’re likely to be fed lots of spicy discussion (interspersed with ads, of course), but unlikely to come away with a deeper understanding of a nuanced topic. The goal of collaboration software, on the other hand, is to help teams get their work done better and more quickly. Applying machine learning algorithms towards that goal can create better outcomes for both enterprises and employees. Put another way, a feed of information designed to help achieve goals faster without distraction will learn to surface the most important and relevant information while reducing a user’s exposure to irrelevant documents and notifications that could derail their workflow.
Machine learning is used to personalize products in both cases. In Twitter, it’s personalized ads and personalized connections to the folks you follow and are likely to engage with. In Atlassian, it’s a personalized path to help you get back to your work faster (through Start), and personalized connections to the people in your team you need to collaborate with (through smart mentions).
Invest in smarts now, or fall behind
Past IT consumerization trends, like design and mobile, were relatively gentle on laggards. Organizations that didn’t lead the charge could catch up about as fast as they could afford to hire the talent and build the tech. But catching up on smarts won’t be so easy. Building smart software requires training machine learning algorithms with huge amounts of data – data that can’t be bought or rush-collected. The longer an organization takes to start collecting, structuring, and processing data to train the algorithms they need to power smarts in their software, the further behind they’ll fall and the harder it will be to catch up.
For enterprises assessing technology providers for their potential as long-term partners, a clear-eyed smarts strategy must be on the list of considerations.