An interesting fact appeared in April when IBM CEO Ginni Rometty spoke at CNBC’s @Work Talent + HR Summit in New York. Rometty was with Jon Fortt when she revealed how IBM’s artificial intelligence (AI) tool has helped IBM with staff retention.
IBM is developing a suite of AI-powered tools for human resources, and this one looks at predictive attrition. Apparently, its analytical software can crunch data sets and predict which employees may leave the company. The next step is to pre-empt an employee’s departure, which saves the company the expense of training and on-boarding a new employee.
It identifies, with 95% accuracy, that you’re showing signs that you’re ready to leave the company. Management or HR could have a conversation with you about added compensation or other substantive changes in order to convince you not to leave. As a footnote, IBM employs around 350,000 people. It also gets around 8,000 resumes (CVs) a day! Rometty claimed that the AI has saved IBM nearly $300 million in retention costs, so far.
While the software is convincing some people to continue working at IBM when they might have been thinking of leaving, it has decreased the number of staff employed by HR. Rometty explained that by using the software, IBM has been able to cut the size of its HR department by 30% globally.
Transparency Is Crucial
Rometty said that being transparent with individual employees about their career path is an issue that many companies still fail to address. And it’s an issue that’s going to become more critical. So, in addition, the software can be used to predict an employee’s future performance, and it can do that with 96% accuracy. Using performance-tracking programs that use data on an employees’ projects, skills and weaknesses, the software can help decide who should be promoted while generally increasing efficiency.
“If you have a skill that is not needed for the future and is abundant in the market and does not fit a strategy my company needs, you are not in a good square to stay inside of,” said Rometty. “I really believe in being transparent about where skills are.”
By better understanding data patterns and adjacent skills, IBM AI can zero in on an individual’s strengths. In turn, this can enable a manager to direct an employee to future opportunities they may not have seen using traditional methods. “We found manager surveys were not accurate,” Rometty said, referring to formal skills assessments. “Managers are subjective in ratings. We can infer and be more accurate from data.”
Discontinuing Unfair Performance Reviews
And it seems IBM has now discontinued annual performance reviews, and not just because they take up expensive staff hours for no observable gain. Their purpose is unclear, and they can be unfair. IBM assesses employees on their skills growth as part of their quarterly feedback check-ins with management.
The software is not only used internally by IBM but is also for sale to other companies. IBM HR now has a patent for its predictive attrition program, which was developed with Watson to predict employee flight risk and prescribe actions for managers to engage employees. And other companies can start to enjoy the benefits.
IBM believes that the future of work is one where a machine can understand the individual employee better than an HR department currently can—making it clear that this is in no way a slur on the performance of HR employees. The suggestion is that the software can look at millions of data points in order to make a decision, whereas an HR specialist can’t do that.
A Great Step Forward, but Not Without Caveats
Like all AI systems, it seems like a great step forward. And taking steps to retain staff that might be thinking of moving on is a clear way for a company to save money and retain the expertise in house.
However, I do wonder about predicting future performance. Does it take into account short-term issues that may affect a person’s performance and attitude to work? For example, a person now has twins in the house and may be very tired for the next few months—affecting their performance in the short term. Or, does it consider that a person may have a parent ill in hospital and how much that can impact on the amount of work they get through in the short term. These are things that managers can likely catch in a way that an AI system can’t.
Accepting those caveats, all-in-all, this seems like a very interesting step forward.