Better decisions, not just faster ones
In a workplace increasingly shaped by AI, speed is often treated as the biggest prize.
Renée Richardson Gosline, Senior Lecturer at MIT Sloan School of Management and lead of Massachusetts Institute of Technology’s (MIT’s) Human-First AI Lab says some of the most useful AI systems may be the ones that also use friction deliberately to reduce risks in human-AI systems.
It’s a concept she describes as “beneficial friction”: points in a workflow where people are prompted to think more carefully, test assumptions and engage with a decision, rather than simply accepting an AI-generated answer.
Beneficial friction can take different forms, Dr Gosline said. In practice, it could mean designing a workflow so people have a chance to test both their own assumptions and the AI’s recommendation before making a decision.
The aim is to reduce the risk of people defaulting to the easiest answer, especially when speed may come at the expense of accuracy, judgement or learning.
Dr Gosline said better decisions mattered more than decisions that were simply “faster,” particularly where AI-enabled convenience could push people towards poor outcomes.
The risk of going too far either way with AI
Dr Gosline said the challenge for organisations was avoiding both blind trust in AI and blanket resistance to it.
“We have people who are AI sceptics and may be missing out on some of the value that be created by these tools,” she said.
“At the same time, we have some folks who talk about AI as if it's a panacea. And even if they're in the middle, they find themselves over-relying all the time because we become habituated to the ease with which AI can help us complete tasks.”
Finding the right balance sits at the heart of the debate about how AI should be embedded into work, she said. Key is whether organisations design the systems, governance and support around them in ways that help people make better decisions.
Keeping people at the centre
Professor Mary-Anne Williams, who was appointed as CommBank’s Chief AI Scientist in May, said the strongest use cases for AI are those that help people make time to do more meaningful, higher-value work while keeping human strengths at the centre.
‘If you can make a human a superhuman by giving them these extraordinary tools, then they can ask bigger and more interesting questions and they can, achieve far more. AI can take of tasks that require data processing faster, but not those tasks that require a deeper understanding of the world or a strong sense of empathy,” she said.
Dr Gosline also pointed to the limits of AI when it comes to original thinking. She said AI can help identify “the mean” - what is typical or already established - but its real value may be in helping people see the conventional path clearly enough to move beyond it.
“Humanity requires us to do things like have faith in other people and believe in what the data may not lead us to conclude.” Having both humans and AI working together will “push the boundaries of what humans can do,” she said.
The value of inconvenience
For Dr Gosline, protecting human judgement also means protecting curiosity.
To guard against over-reliance on AI, she encourages people to “practice acts of inconvenience to engage their critical thinking, their curiosity, their exploration. Not as a time waste, but as something that creates richness, that pays down the road. Without that, will we do it as a species? Where do we go when we kill that kind of curiosity?”.
It’s a message that’s especially relevant as organisations look for ways to scale AI while maintaining trust, accountability and human capability.
Why the research matters
Dr Gosline shared her findings during a session attended by more than 1,200 CommBank employees.
“The turnout reflects a deep curiosity amongst our people about how to best apply AI in their daily work and in their teams without sacrificing human skills,” CommBank Chief Behavioural Scientist Will Mailer said.
CommBank is collaborating with MIT Sloan as part of its broader focus on responsible, human-centred AI, with research exploring how organisations can better manage AI risks, understand human-AI collaboration and design AI-powered tools that support critical thinking and decision-making.
CommBank Chief Technology Officer Rodrigo Castillo said:
“Insights from the Dr Gosline’s presentation provide us with greater understanding of how people interact with AI technology and insights into how we can continue developing a human-centered future workforce.
In 2024, we established our Human-Centered AI unit which supports a research-backed approach to AI. Bringing experts together across technical, business and social science disciplines is an essential part of how we approach the technology responsibly and deliver better outcomes for our people, customers and communities.”
