The algorithm dilemma Gartner didn't predict – and how to avoid it
Gartner’s tagline for their recent Digital Workplace Summit in London was that by 2020:
“Algorithms will positively alter the behavior of over 1 billion global workers”
Thus they will boost employee engagement and business productivity. This optimistic claim stands in stark contrast to dismal global workplace engagement statistics. Gallup’s 2017 State of the Global Workforce provides a damning picture of a largely disengaged workforce.
Western Europe provided the most startling numbers with only 10% of employees actively engaged in their jobs, a figure which “offers insight into why labor productivity in many European countries trails that in the U.S.” Even the US, the report’s leader in workplace engagement, only managed an engagement rate of 31%.
This productivity loss, combined with a higher rate of employee churn due low engagement, can severely impact the bottom line for businesses. The “good news” according to Gallup is that employers “often need look no further than their workforces” if they commit to “management changes that drive continual development”.
In Gartner’s vision of the future, algorithms have a key role to play in turning engagement around and, to be honest, they probably do. AI and Machine Learning are probably going to be provide managers with tools that can provide insight into how their employees feel about their jobs or help automate tedious work.
Technology isn’t always the solution
They are, however, just tools and their success is entirely dependent on how they’re used. Amazon’s recent HR issues provide a perfect example of this – even a company which prides itself on driving technology forwards can end up being guilty of problematic implementations.
Other recent waves of enterprise software provide other examples. Tools like Slack can certainly boost engagement by giving employees channels for informal collaboration. Yet, as Harvard Business Review research suggests, it can just as easily cause “collaborative overload” by making everyone reachable, all the time, everywhere.
The same goes for digitalization: Giving people digital tools to access the information they need for work should benefit productivity, but it can also lead to digital fragmentation. Research from IDC suggests that this happened, with 61% of knowledge workers having to spend time accessing 4 or more systems on a regular basis to “do their job” and 13% needing to access 11 or more systems.
Any technology that has the potential to boost engagement and make people feel happier at work at work can just as easily amplify existing issues with workplace culture. This might sound a bit pessimistic, but it doesn’t mean companies should abandon trying to boost engagement with AI, just as they shouldn’t abandon Slack or cancel their Salesforce or SAP licenses.
Instead, any discussion among CIOs and IT Directors over whether AI has a place in improving engagement needs to start by directly involving the people whose engagement it’s meant to be improving. If machines are to be any use to employees, they need to be designed with humans in mind and implemented through a bottom-up approach.
Start with pain points, end with solutions
What this means in practice is starting with a comprehensive and open analysis of what employees’ pain points actually are and then looking at technology options through the lens of these pain points. The most recent Digital Workplace Awards provides some points on how this might be done, with both winning projects making use of AI for enterprise search and conversation.
One winner described the employee frustrations they found were fairly mundane – having to take a minute and 10 clicks to find a PTO report or cafeteria menu. None of these pain points were especially disastrous, but they add up both in lost time and in workplace frustration. Other found that employees were frustrated with having to install work apps to solve individual problems, or AV systems in meeting rooms that are frustrating to use or never work.
These might sound like small problems, but these kinds of frustrations are often missed or just not fixed. Feedback may only be sought in the context of a performance review, or an exit interview. Customer-facing systems may be improved over internal systems, because of the more obvious ROI. They also aren’t the kind of 'wicked problems' that we envision AI as solving. These aren’t transformative discoveries made through using AI to find hidden patterns in big data, but they are problems that pretty much everyone in an organization will encounter.
An example solution both awards winners found was to make it easier to find information through providing a conversational interface for employees through which they could get answers to workplace problems. This is a pretty limited application of AI, limited to Natural Language Processing and intent recognition, but it went some way to solving the pain points that were raised.
If organizations want AI algorithms to positively impact the behavior of employees they need to start any project with behavior and people first, rather the presupposing a problem and applying technology to fix.