Closes in 10 seconds skip ×

Does Your AI Need A Background Check and a Reference?

By: | June 12, 2019 • 4 min read
Emerging Intelligence columnist John Sumser is the principal analyst at HRExaminer. He researches the impact of data, analytics, AI and associated ethical issues on the workplace. John works with vendors and HR departments to identify problems, define solutions and clarify the narrative. He recently spoke at the HR Tech Conference, when HRExaminer’s Second Annual Index of Intelligent Software in HR was published, and is slated to speak at the HR Festival Asia in Singapore, May 8 through 9. He can be emailed at [email protected]

In June, I start seriously looking forward to the annual HR Technology Conference in October. There is a rhythm to the year that seems to begin and end with the encampment in Las Vegas, considered the industry’s town hall. We get together and show each other what we’ve learned and created over the past 12 months.

For most of the 2010s, the story was the increasing importance of data in the human resource profession. We have all been learning how to effectively incorporate data into our decision-making. At first, the hype focused on big data. Then it was people analytics. This year, don’t be surprised if everyone at the conference is telling you about their artificial intelligence.

It’s essential to notice how the story stays the same even when the language is changing. Big data, people analytics and artificial intelligence are really just pieces of one thing. More data plus more processing yields better understanding. How you categorize it matters much less than how you evaluate it and how you use it.

Advertisement




Currently, the story is about all the ways machine intelligence is creeping into every decision we make. Like ivy on stone walls, the stuff is everywhere. Because they have tons of data, enterprise providers are embedding intelligence in every aspect of their suites.

It’s like having a flock of new workers offering insights, recommendations, forecasts and predictions about every aspect of our processes and outputs. We are introducing these fast-learning but young and inexperienced workers into our organizations. The good news is that they learn quickly—the bad news is that they are more emphatic about their output than the work actually deserves.

Meanwhile, industry-leading HR departments are quickly deploying teams of data scientists to make sense of all the new information and recommendations. These early stages of HR as a science involve investigating correlations between every imaginable piece of data for clues to workforce productivity, predictions of employee behavior, selection suggestions and the ways HR drives business results.

In short, algorithms, models, correlations, predictions, recommendations and forecasts are coming towards us in a blaze of potential insight. The key question is whether or not you can trust what you see. Do a ton of machine processing and math actually improve on historical efforts? Does the underlying model adequately describe the reality it predicts?

The answer, unsurprisingly, is that nobody can tell at first. Although a machine’s output has a certain ring of authority, you’re really dealing with a new worker. As with all junior employees, trust should be earned over time.

Advertisement




With new digital workers, all you really have is the vendor’s assurance that its work will be adequate. It’s highly unlikely a vendor will be willing to be responsible for mistakes that digital tools make. So, in order to use them, you have to cautiously and carefully move from skepticism to trust. It’s the same journey you take with any new employee.

Here are five principles for understanding and building trust with your new digital employee:

  1. Understand What You’re Getting: Deeply review every single algorithm, model, process automation, prediction or recommendation function before releasing it into the “wilds” of your organization. Understand the results before you start using them. Build a review board that considers ethics, legal questions, diversity consequence, cultural implications and business-outcome impacts. Make the review process public and transparent.
  2. Measure and Test: Measure the output of each action initiated or suggested by the machine employee. Review the measures daily at first. Settle into an oversight rhythm based on the importance and frequency of the function. Over time, you are looking for variances that impact your organization.
  3. Anticipate Replacement: Plan for the fact that all digital employees wear out and need refreshment. For each model, algorithm, prediction, recommendation or process automation, have an ongoing process that is constantly preparing for inevitable failure while searching for improvement in the quality of digital performance.
  4. Monitor Outcomes: At an aggregate level, your digital employees will have a system-level impact. It will show in engagement levels, turnover rates, organizational productivity, communications effectiveness and other organization-level indicators. Monitor these things in that light.
  5. Track and Record Development: Build and maintain a central library that documents the workings of each digital employee, its output, replacement strategies, maintenance actions, training data and re-engineering assistance. Digital employees require job aides and descriptions that are better than those you give your human employees because they are much harder to replace.

Adding digital workers to the workforce requires new forms of organizational management. These are the minimum starting points.

 

 

 

More from HRE