Unlocking the power of AI in HR: Four key principles for success

In the ever-evolving landscape of human resources, artificial intelligence (AI) stands as both a promising ally and a source of trepidation for HR professionals. The potential for enhanced productivity and efficiency is exciting, but there is a concern among employees about how AI might impact their job security, rights, and overall job quality.

“Getting AI right in HR is critical to help introduce employees to an AI-supported workplace. Responsibility needs to be baked in from the get-go if employees are to respect, trust and realise the possibilities of AI,” said Aaron Goldsmid, product lead at Deel.

Best practice change management and communication practices should be applied to ensure employees are part of the transformation journey, including visibility into how their data is being used to train AI models, he said.

“Trust is vital in workplace relationships,” said Goldsmid. “While there is a sense of excitement about the benefits of AI in terms of enhanced productivity and efficiency, wariness persists among employees about how it might diminish their rights, the quality of their job, or even make their roles redundant.”

At Deel, AI is “acting as co-pilot” to employment lawyers and tax advisers, “taking on the repetitive queries and freeing people to tackle the harder questions,” said Goldsmid.

At Deel, they’ve identified four key principles to guide how the HR industry can best adopt AI. “Broadly, the principles aim to maintain human control and trust in AI-guided HR and wider workplace systems,” he said.

Four key principles for HR to best adopt AI

Firstly, HR AI must be “ring-fenced” in how it’s developed and the data it uses, said Goldsmid.

“Staff should have the confidence that HR AI is ring fenced and employee data is safe and secure, and able to see human accountability,” he said.

“AI technologies are intelligent but incredibly naive at the same time, and when dealing with employee data, there is no room for error.”

At Deel, where they hold a lot of highly sensitive information, they test new products extensively in a highly controlled environment, or ‘sandbox’ for each data set so the AI does not access anything it is not meant to. Then they have human experts who are validating every query –key to the second principle: no AI works without a human component, said Goldsmid.

“Because the landscape of global compliance is constantly shifting, HR AI must be developed with human-assisted learning and human experts need to validate every query. We describe this second principle as ensuring that ‘humans are always in the loop.’”

The third principle underscores AI's role as an assistant rather than a replacement. “AI should ‘free up humans’, by which we mean the purpose of AI should be to remove menial chores, freeing up HR pros to concentrate on talent development and growth,” he said..

The final piece of advice is that AI is not the decision-maker, said Goldsmid. “We insist that AI can never decide by itself; it’s the assistant and not the boss. If AI’s role from the outset is designed to augment and support, rather than replace, then the employee experience can only be enhanced.”

Four principles summary

HR AI must:

1. Be ring fenced in how it’s developed, and then in the data it can use when it’s deployed.
2. Have ‘humans in the loop’ - no AI works without a human component. 
3. Free up humans - remove menial chores, freeing up pros to concentrate on talent, development and growth.
4. Never decide. AI is an assistant not the boss. It can take the load, but never make the decision and always be checked