The Four Principles of Creating with AI in the Workplace
This article is an adapted version of a talk delivered at AIFest 2018 in Montreal.
Statistics Canada found that the average Canadian works roughly 40 hours a week. It may not seem like much, but even with public holidays, vacation time, and an expected retirement age of 65, that’s 124,800 hours spent at work. Assuming an average life expectancy of 80, that’s roughly 43 years or more than 50% of our lives. We spend a lot of time at work. How can AI make those 43 years of our lives spent at work more meaningful, safer, more productive, and less mundane?
The best AI experiences in the workplace are those where technology acts on an employee’s behalf. And through our experiences working with Fortune 500 organizations to create smarter interfaces that do just that, we have established 4 key principles for getting started with AI in the workplace.
Principle 1: Understand Your People
“AI needs to accept human behaviour the way it is, not the way we would wish it to be”. — Don Norman, author The Design of Everyday Things and director, The Design Lab at UCSD
Before we can accept human behaviour and beliefs, we need to know what they are. We’ve seen organizations invest significant resources into implementing AI at work, only to find their AI projects fail in the long run. Knowing your users is a critical first step in implementing AI. And it’s the only way to ensure that AI initiatives will be adopted, accepted, and successful.
This goes beyond simply understanding your users’ practical goals and extends to their opinions of artificial intelligence. Once you understand your employees’ attitudes toward AI you can plan and design accordingly. Address their fears, beliefs, concerns, and perceptions. Manage their expectations through good communication and training.
Employees worry that AI will disrupt their daily work habits and patterns. They worry about disruptions to their relationships with colleagues and clients. They worry about new approaches to measuring their performance. They worry a lot about reduction of their roles at best, job elimination at worst. In essence, they’re worried that they will lose a sense of control over their work and its outcomes.
But a recent Capgemini study on the future of work found that AI at work will actually increase job opportunities and improve efficiency and service. This is why communication and transparency are an equally important part of understanding your people.
Organizations must do a better job of being open about their AI strategy. Increased transparency around roles that will evolve, tasks that will be replaced, and any retraining initiatives will help ease the transition. Not allowing your employees to look inside the blackbox of AI only contributes to the cycle of fear.
“Data” as a topic warrants a separate discussion, but we can’t discuss AI without at least acknowledging the spectre of data and how it gets used. And it’s a particularly sticky topic in workplace settings. Every employee will have a different perspective on privacy.
Understanding these differences improves an organization’s chance of driving engagement and efficiency. For example, how do generational differences impact acceptance of new technologies? How do culture or socio-economic factors change people’s perspectives? And what contextual factors — such as the impact of a continuously evolving workplace — alter feelings around AI?
It isn’t enough to assume you understand your user when it comes to building a product. You have to find opportunities to have in-depth, qualitative conversations with employees across a range of regions, tenures, and levels of seniority. Then, to corroborate these findings, you can conduct large-scale surveys to get a sense of their scope.
Principle 2: Explore New Problem Spaces
Practicing design thinking ensures you don’t get caught up in the hype storm surrounding
technology. It can prevent you from creating solutions to problems no one has.That being said, there’s also a healthy balance to be had Sometimes when we don’t have a frame of reference for what’s possible, we develop tunnel vision and skip over problems we didn’t think could be solved.
This can happen in design even when AI isn’t in the picture. Our cognitive bias, driven by current experiences of technology (for example, mobile devices), limits our creative thinking about the application of other forms (like voice interaction). To deal with this, we apply integrative thinking. Integrative thinking asks us to hold two opposing ideas in our heads and reconcile them without abandoning one or the other. We often employ this approach when technological potential and human needs intersect.
The first step in our process is to learn more about the potential of a technology. Understand what types of problems AI is good at solving. This boils down to understanding AI as a design material.
Once you feel clear on the things AI can do in your context, put that list away and review your research. Understand where your employees have the greatest pains. Finally, populate a matrix that intersects those pain points with the AI capabilities you’ve uncovered. By doing this you’ll generate unexpected solutions to problems that you might not have considered solving had the potential of AI not been a lens.
In other words, we can look at the intersections of technology capabilities and employee needs to generate solutions to problems we may not have felt comfortable addressing in the past. And because data plays a key role in AI implementations, it’s a key criterion for narrowing down an area of focus once we’ve generated ideas. In a workplace context, both structured and unstructured data may be widely available and accessible. Leverage it and find greater success.
Principle 3: Pick a Good Partner
Many companies want to make use of AI, but aren’t ready to invest billions in the technology. Leveraging what already exists through an AI technology partner is likely the best course of action for most of us.
So what exactly do you leverage by working with a technology partner? Technology partners bring quality, value, speed, and simplicity to an implementation. And they reduce risk.
Partners offer a host of benefits. They offer the ability to integrate more than one specialized technology into an application. This allows your organization to benefit from the wealth of research and industry knowledge the partner has. They also bring diversity of thought to projects, initiatives, and teams. They may have access to more computing power, so you don’t have to invest in new infrastructure to crunch your data. In addition, many partners have pre-trained models for common use cases, which saves you from training AI models from scratch.
Technology partners also typically manage maintenance and updates. This can be a huge cost savings for organizations. Five years ago, off the shelf technology was synonymous with “conventional” or “inflexible”. Today it’s a very different landscape. Even with off the shelf technology, customization is the standard. Different plans, modules, widgets, and extensions are available that you can leverage and select per your organization’s needs. We now have the ability to start with off the shelf and build off of it to create something completely custom.
The final advantage of partners is talent leverage. AI implementations require talent that is hard to find. In the labour market, we have a skills shortage of AI/ML scientists, engineers, architects, visualization designers… The numbers are growing but it is still slow, steady and fairly limited.
Think critically about your AI partner ecosystem — an integrated network that you can build long-term, mutually-beneficial relationships with to help you reach and exceed your goals. Consider how both AI and your partners can impact your company culture. Does the partner align with you on values? There must be a strong human, business, and technology alignment to succeed.
What does your AI governance look like? What format will it take? AI is all about data, and in the context of workplace employee data is sensitive and confidential. Does the partner fulfill your security and data storage requirements? Has the partner proven that their technology can actually work for your use cases?
Lastly, please don’t pick partners that offer products with subpar user experience. It’s one of the biggest problems with workplace software and it will kill your adoption rate. Keep usability and accessibility a priority.
Principle 4: Start Small & Involve Your People
This final principle is less specific to AI implementations, but important to call out because of the very real human impact of AI. As you bring AI into familiar processes, finding a way to shore up trust for those it impacts in your organization will be crucial to its success. The good news is it’s not too hard to establish trust with your people.
To start, apply Lean Startup thinking. Run small, parallel initiatives that fit into a broader strategy. Institutional and social structures will make or break a new technology implementation and making several small changes can help limit the impact of inevitable early failures (and help acclimatize your teams to AI at the same time).
As part of introducing these early implementations, study user behaviours and usage patterns. Don’t just look at adoption, which is a broad metric to explore. Instead, focus on studying set up, monitoring, and takeover scenarios. This is where most human-machine interaction is likely to occur with systems that are either fully autonomous or agentive.
As an aside, we’ve built these key touchpoints into a Smart Things Canvas for planning the specifics of AI feature sets, synthesizing numerous sources and structuring the canvas around ideas from Chris Noessel’s Designing Agentive Technology.
To keep things organized and strategic, link these initiatives to the broader employee journey. Let the success of your experiments dictate where you invest more on the road to achieving your broader vision. And don’t be scared off if these early initiatives fail — learn from them to move forward intelligently. Build upon the most successful pilots to gradually achieve your broader vision.
It’s also important to get ahead of any misconceptions and treat AI as a change management initiative, not a simple software change or new form factor. More than likely, your first AI implementation will simply surface extra insights or automate portions of a human process. That’s ok. Remember, you’re starting small before scaling up to additional automation or across various channels and devices.
Finally, and perhaps most crucially, define and promote a strong feedback loop before your application is live. Make your employees feel like they’re not just a focus group, but active participants in the evolution of your AI implementation. Use feedback to understand the pace of change your employees are comfortable with and the depth of explanation they expect. And keep those feedback loops open long after the application is live. Don’t stop listening.
AI can be a daunting topic if you’re not a technologist and workplaces can seem daunting and bureaucratic if you’re not in HR. With people spending more than half their lives at work, we owe it to them to create workplace experiences that both assist in their jobs and feel good to use. By following these four principles, you can help ensure your AI implementations will be successful.
Start by understanding your people. Explore attitudes around AI and uncover human needs before committing to any one plan. Next, explore new problem spaces. Use integrative thinking and layer in data to direct your efforts to help ensure the choice you make is the best choice for your team and their needs. Once you know the direction you’re heading, pick a good partner. Save time, money, and risk by leveraging what already exists. And look for partners in unlikely places. Finally, start small and apply Lean Startup thinking to get the project going and be sure to involve your people and listen to them. It’s the only way to be sure you’ll get buy-in. And when you involve your teams, it makes it easier to manage change and the inevitable hurdles it brings.
Interested in finding out what AI can do for your workplace? Drop us a line to discuss your next project.