Presented by CIBC
It’s easy to see why businesses would want to experiment with artificial intelligence: it can help identify pain points in a company’s workflow, leading to more efficient supply chains or internal processes, which saves time and money. It also allows workers to offload low-value tasks, such as data entry or the cross-checking of transactions or contracts, giving them time to focus on more creative or higher value work. AI can even accelerate new product development, which may lead to new revenue streams.
In fact, according to a 2017 research study by Accenture, it “could boost average profitability rates by 38% and lead to an economic increase of $14 trillion (USD) by 2035,” particularly in the information and communication, manufacturing and financial services sectors.
“AI is allowing us to move faster, opening the door to insights and capabilities that are simply beyond what we were able to do in the past,” says Terry Hickey, Chief Analytics Officer at CIBC. “Leveraging AI to augment what our employees are doing creates a more powerful workforce and allows companies to scale at a rate that wouldn’t be otherwise possible.”
But that’s only if companies are smart about how they integrate this technology. Here are five common mistakes businesses make when investing in AI—and how yours can avoid them.
1. Choosing the wrong projects
Simply spending money to acquire new technology isn’t enough. Companies need a firm understanding of why they want to incorporate AI into their workflows, what they’re hoping to accomplish and what success will look like. So no, your business likely doesn’t need a deep neural network. Instead, spend time thinking about your core business, and the pain points your employees experience on a regular basis—like, for example, a high no-show rate at a healthcare provider, or a factory early-warning system that can’t distinguish between a malfunction and a planned shutdown.
In short, don’t “spend too much money without considering first what AI can really do and whether it will impact on decisions made,” says Joshua Gans, chief economist at the Creative Destruction Lab at Rotman School of Management.
Equally important: starting small.
“My philosophy on choosing initial AI projects is ‘snacks not meals,’” Hickey says. “Meaning, start with a variety of smaller projects vs. investing in one big initiative. This allows you to see where the best ROI is and what types of AI initiatives work for your organization. The trick to making this work is to make them big enough to get people excited and to show the potential of AI, but small enough that they can be implemented quickly.”
2. Keeping your silos
Access to data is one of the major stumbling blocks that new AI projects can face—and information or organizational silos are one reason why. “Siloed departments prevent AI from having all the information they need and from talking to each other,” Hickey says. These projects require reams of data analysis to analyze, but if it’s not being “shared across all departments, your AI can’t learn anything outside of its sources or know what’s happening in other areas. It simply won’t know the right questions to ask.”
That doesn’t mean companies have to overhaul their entire org charts—but they do have to improve communication between departments, and make sure they’ve created a culture of sharing information.
3. Reinventing your workflow
Although companies likely will need to tweak their processes to accommodate AI, Hickey says this technology is far more effective when it’s integrated into existing workflows.
Companies don’t necessarily have to change everything about the way your business runs in order to integrate AI. “Right now, all of the AI applications [we have] improve existing tasks. You only want to reorganize if you are reengineering the entire organization, which may happen one day but not today,” Gans says.
In fact, it’s sometimes impossible to integrate AI into large or closed systems. In those cases, the system needs to be valuable enough that workers will use it, even if it does operate outside of an existing workflow. Hickey recommends ensuring its dashboards, timely notifications or interfaces that are easy to interact with.
4. Expecting too much
“One of the biggest mistakes [businesses make] is being unrealistic about what AI can do. Too many people try to have AI solve every problem, when in reality, each tool has a specific purpose,” Hickey says. “The real goal should be to optimize the purpose and capabilities of AI to deliver better insights and capabilities for your clients.”
So yes, AI can help target online ads, tag photos, translate or transcribe audio and even anticipate whether a mortgage applicant is likely to repay their loan, but it can’t solve every problem in an organization.
That’s also why companies shouldn’t become too dependent on AI. It still requires human supervision and intervention, so implementing AI is not an excuse to stop monitoring systems or training employees on new skills.
5. Ignoring employee concerns
One of the biggest concerns around AI is its potential to automate people out of jobs. Automation does sometimes lead to redundancies, but it’s more likely workers will need to learn to work with AI rather than be replaced by it—and it’s important to tell them so.
“Being transparent about your AI strategy is crucial to help address employee concerns,” Hickey says. It also helps them “understand the benefits, which include freeing up our team members to focus on higher value work that strengthens client relationships.” For example, AI is great at finding patterns in data, but it doesn’t understand context, something humans excel at.