This artificial intelligence won’t take your job, it will help you do it better

AI-powered tools are everywhere. The challenge lies in deploying them so they actually do some good.

This artificial intelligence won’t take your job, it will help you do it better
[Photo: wutwhanfoto/iStock]

Artificial intelligence (AI) and machine learning are increasingly powering workplace platforms and tools. The sophisticated automation tools have been widely promoted as relieving workers from tasks that are “dirty, dull, or dangerous,” unleashing them to do higher-level work and create. PwC research estimates that AI will contribute $15.7 trillion to the global economy by 2030, driven primarily by productivity gains and AI-fueled product innovation.


In various categories, it’s beginning to deliver on its promise. Financial services companies are using such technologies in ways that range from chatbots that answer basic customer questions to AI-powered platforms that help prevent fraud and money laundering. Human resources (HR) applications help companies sort through resumes, find talent, and even conduct initial interviews. It can be used for maintenance alerts and prevent equipment and vehicle failure in automotive fleets. Purchasing algorithms can help sort through data to make better procurement decisions. In healthcare, promising applications range from robotic surgery to diagnoses of various conditions to AI-powered preauthorizations and other medical certifications.

But realizing the benefits of AI requires thoughtful planning, says Soumendra Mohanty, executive vice president and chief data officer at LTI, a global IT firm. Effective AI applications “elevate the work,” allowing humans to do “higher-order work,” he says. That works best when the technology is implemented the right way.

Build a framework that works

A scattershot approach to AI usually isn’t the right way to make a difference, says Dan Priest, technology strategy leader with PwC’s Strategy&, the strategic consulting arm of PwC. In the past, automation typically came in big rollouts like ERPs that had a very distinct value proposition. “AI is smaller. It gets introduced in fast sprints, in a more decentralized model, and so the companies need to manage that type of automation differently, having some guardrails in place,” he says. Across its various applications, there are some common steps to get it right.

Start with the task—not the job

Effective AI applications start with a business problem rather than a specific role, says Jeanne Meister, founding partner of Future Workplace, an HR advisory and research firm providing insights on the future of learning and working. Collect data on the business problem, then educate business leaders about what you’re trying to solve is the first step to successful AI application. Identify the specific tasks that AI can improve and what you hope the tool or platform will achieve in terms of productivity, efficiency, accuracy, or other goals.

“Why are you doing this? How are you going to move forward, what are you going to do, and how are you going to track it?” she says. “We forget to keep it simple.” That means starting with the problem that needs to be solved and tapping a cross-functional team that can help identify the potential—and possible consequences—of adopting AI in a given area.


Get the data right

AI is only as good as its data. What data does the tool need? What data is necessary to train it? “[I]t’s interesting because a lot of people think the challenge is the algorithm, and yes, those are getting better and better and continually improving as we learn more about this space. But the biggest challenges actually lie in the data,” says Malcolm Silberman, a director and blockchain and AI practice lead with accounting and advisory firm Grant Thornton.

AI applications are, by nature, data-hungry, Mohanty says. Depending on their purpose, they need information about the employee and their performance, as well as environmental and other factors that affect performance, which may require different algorithms.

Once the inputs are identified and refined, he says it’s critical to revisit the data component regularly to check for data and algorithmic biases that can creep into the interaction between the AI tool and the employee. Recently, Amazon abandoned its AI recruitment tool because it may have been eliminating women job candidates, for example.

“AI output needs to be also understood in the context of why it said something and based on what kind of information it has,” Moharty says. “[I]f it is a mission-critical goal, you need to be very, very clear about why the algorithm said something and why it is recommending something. So, that’s the general trend and certain kind of a careful approach that needs to be taken going forward.”

Understand your team

You also need to consider your team and how adoption will be received, says Carrie Duarte, PwC’s Workforce of the Future leader. Employees typically fall into three categories—early adopters who are curious and enthusiastic about the technology; those who are motivated by the positive impact the tool may have on their jobs; and then the reticent who may not like change or are concerned about the time and effort it will take to learn and use the new technology.


Getting employee and cross-functional team input is also critical, she says. If you don’t understand the employee’s job or the potential complications of rolling out the technology, “You can’t really know that you’re improving their day-to-day, and the productivity and the employee experience and engagement,” she says.


When you understand the role and the task, you can also better establish metrics to measure success. “Ask the question, ‘How would we categorize the different activities performed by Jim and Bob or Sally,’ and then the second question that often doesn’t get asked is, ‘What exactly are we trying to solve for?'” says Ravin Jesuthasan, a managing director at Willis Towers Watson and author of Reinventing Jobs: A 4-Step Approach for Applying Automation to Work. Typically, he says, AI is used as a solution in one of four areas:

  • Minimizing errors
  • Minimizing variants in performance
  • Increasing productivity
  • Achieving breakthroughs

Establishing how “success” will be measured up front will help organizations adapt and use the right tools to achieve the desired outcomes, he says.

One thing is certain: AI is here to stay. Whether you are a retail worker or a C-level executive overseeing strategy, the entire range of work and rules and files are going to be augmented with AI, Mohanty says. Ensuring that these tools are properly deployed and monitored is essential to use them optimally.

About the author

Gwen Moran is a writer, editor, and creator of Bloom Anywhere, a website for people who want to move up or move on. She writes about business, leadership, money, and assorted other topics for leading publications and websites