IBM hit the news this week when their CEO Ginny Rometty announced in an interviewthat they have developed an AI tool that can predict with 95% accuracy when employees are likely to leave. The tech giant, which employees 350,000 people globally, claims that this has helped them save $300m that the departures would otherwise have cost them.
Many firms have been pushing similar tools for a while, but few have made such bold and specific claims about the accuracy and financial impact as Rometty this week did.
The implications for employers are huge, and the story picks up on many of the seductive promises made by AI proponents; it suggests that computers can give you rich insights into the mystery of human behaviour, they can do this in advance and better than you can, and they can provide you the tools to act on this, change the future and save lots of money in the process.
So how should leaders respond? There are eight things you should think about to help you decide whether or not to pick up the phone to Rometty and book a meeting.
1. Embrace the technology
It feels like we could be entering Minority Report-style territory here, with computers predicting what your employees can do before they know it themselves, and that might seem barely credible. But the first thing you should do is accept that this technology is, if not here, then just around the corner.
IBM may be one of the first to market predictive technology in this area, but it will soon become common. It’s over two years since Facebook revealed they can tell when you fall in loveand seven since Target spotted a teenage pregnancybefore the girl’s father did.
It’s therefore inevitable that tools to predict employee turnover will become commonplace and pretending otherwise is not an option. Start thinking about this now.
2. Recognise IBM are selling something
IBM is a serious company with a long track record of technology-driven innovation, so it’s not surprising that they might develop a ground-breaking tool like this.
But they are also looking to sell you something. Rometty talks about the impact IBM has seen internally when using this tool rather than overtly marketing it, but consider why they’re talking about this publicly at all. IBM is not an academic body with an impartial interest in AI research; their brand and commercial success – their ongoing existence – are built on making money from AI and that’s why you’re hearing about it.
That doesn’t mean, of course, that the product does not work or their claims are false. But it does mean you should remember where they’re coming from when the phone rings and they want to talk.
3. Understand the algorithm
IBM are cagey about the detail behind the algorithm that they have developed, saying only that “its success comes through analyzing many data points.”
That’s a perfectly understandable response from a company with proprietary technology, but it’s completely unacceptable for you if you’re planning on using this with your employees. If you are going to use AI to inform decisions about how to manage retention risk then you need to be confident that the analysis, assumptions and data used are robust. Trusting a third party to do this without knowing how they do is an abrogation of responsibility.
Not only that, but it risks opening a legal minefield if you can’t demonstrate transparency and objectivity behind your actions, and raises ethical questions too; if you are basing decisions on predictions from algorithms owned by another firm then who in your company is taking accountability for how you manage your people?
4. Have confidence in your data
Those “many data points” that IBM analyse: how accurate are they? There is plenty of research to show that performance ratings are wildly subjective, and many HR systems have other data – career history, job titles, demographic data – that is just plain wrong .
It’s unclear what data points IBM use, but if you’re going to start using a range of data to make predictions about employee behaviour then you need to be very sure that this data is accurate. That means having processes and governance in place to ensure that the data is proven to be robust, it is stored, updated and audited regularly, and it is securely managed.
5. Recognise the potential for abuse
Rometty describes how IBM use retention insights to trigger proactive conversations and tackle the issues that would otherwise have led to departures, and that is undeniably a good thing.
But it is not difficult to imagine how these insights could be misused. You know, for example, that someone is thinking about leaving; how does that impact the bonus, promotion or career opportunity that you might otherwise have given them? Why would you reward someone who is clearly not as committed as they could be?
Or they might be a good performer but their manager just doesn’t like them; perhaps rather than try and make them stay, the manager might turn up the pressure, knowing that a little nudge is all that’s needed to make them jump.
If you are going to use AI insights to change employee behaviour then you need very clear governance and oversight to ensure that you do so correctly.
6. Ensure your managers are equipped to act on the insights
Many people managers already struggle to have an effective discussion with employees about performance and development. If you are expecting them to be able to proactively discuss whether they are thinking about leaving, why that might be, and what both parties can do to prevent it, then you’re raising the bar significantly when it comes to the level of interpersonal skills you’re expecting them to display.
There are three things that you need to do very well for this to work: make sure your managers understand and can explain exactly how the AI works. This is imperative if they are to have confidence in the insights and can pass that confidence on to the employee.
Empower your managers to know what levers they can pull to influence the employee decision; should they be offering more money or supporting people to change roles? How should they decide how important the employee is? At what point should they decide the effort to retain outweighs the risk of departure? There are many variables to consider, and you need to make sure your managers can do this effectively and consistently.
Lastly, you need to hire, train and reward your manages for their interpersonal skills. That should happen anyway, but this is a new and very specific situation that you’re expecting them to deal with. The behavioural skills involved – empathy, listening, building trust, openness – are essential for any good manager, but the context is unique and you need new training and guidance for your managers to be able to handle this.
7. Reassure your employees
How will your employees react if you announce you’re rolling out this software? You can probably guess, but it’s fair to say there will be high levels of nervousness. Even if they buy into the purpose and validity of the tool, people will inevitably speculate about what comes next – AI telling managers who to fire? Being managed by robots?
Unless you already have a very strong culture of trust and transparency with high technology adoption then your employees will be very cautious at best.
This means considering your current culture and employee relations before adopting this sort of technology. It also means complete openness and transparency with staff about how it works and how it will (and won’t) be used. And depending on employee (and union) reactions, it means understanding the risks involved should mistakes happen: if you have fractious industrial relations, for example, the reputational and legal risks could be significant.
8. What can you do without IBM?
Lastly, think about this: IBM are using data points that you most likely already have to predict the behaviour of people who already work for you.
This means that you can almost certainly be doing more than you already are to spot potential retention risks and proactively manage them. Before you invest in new software to predict employee departures, ask yourself if you’re actually investing in technology to do something that you and your managers should be doing as a core part of their jobs.
That’s exactly what you are doing if you adopt this, so consider whether IBM are actually helping you solve the underlying issue – you’re not communicating properly with your people – and whether there is a better solution.
IBM are painting a picture of an exciting new future. It may not be here yet, but you need to start thinking now about how you will equip your company and the people in it to make the most of this new technology. Burying your head in the sand is not an option, but neither is jumping on board and implementing without fully and carefully thinking through all the implications.