What should your machines do? with Stefanie Hutka
09/24/2026
As organizations race to automate with AI, the most important question may not be what machines can do, but what they should do. Lou talks with cognitive neuroscientist and UX researcher Stef Hutka, author of the forthcoming Rosenfeld Media book What Your Machines Should Do: The Science and Strategy of Human-Centered Automation, about how teams can make more thoughtful decisions about automation.
Stef introduces the Autonomy Decision Matrix, a framework that asks teams to evaluate automation along two dimensions: Could you automate it? and Should you automate it? Using examples from augmented reality and AI-powered customer service, she explains why technical feasibility alone isn’t enough — and how shared frameworks can help leaders and practitioners make better decisions together.
The conversation ultimately makes the case for “humanist automation”: using technology to extend human capabilities rather than simply replace people. Stef will continue exploring that idea as a curator of the human-centered tooling track at Rosenfeld Media’s upcoming Shift UX 2026, where she and the speakers will examine how our tools are reshaping not only products, but organizations, communities, and society.
What you’ll learn from this episode
- Why “could we automate?” and “should we automate?” are different questions
- How the Autonomy Decision Matrix guides automation decisions
- Why augmentation can be more valuable than replacing people
- How shared frameworks help leaders and practitioners discuss AI
- What “humanist automation” means in practice
- How designers can think beyond today’s tools toward longer-term consequences
Q&A with Stefanie Hutka
Q: What is the most important question organizations should ask about AI and automation?
A: The central question is not, “What can our machines do?” It is, “What should our machines do?” That distinction forces organizations to look beyond technical capability. It asks what people gain or lose, what the business is actually trying to achieve, and whether automation improves the entire system—not just a narrow efficiency metric.
Q: What does human-centered automation mean?
A: Human-centered automation is technology designed to enhance human judgment, capability, and agency rather than remove people by default. Sometimes that means fully automating a routine task. Often, it means creating a productive partnership: the machine handles repetitive work, retrieves information, or identifies patterns, while people retain responsibility for interpretation, empathy, creativity, exceptions, and consequential decisions.
Q: What is the difference between “could we automate this?” and “should we automate this?”
A: “Could” is about technical and operational feasibility. Do we have a model that works reliably, good enough data, usable integrations, and the organizational readiness to operate the system?
“Should” is about value, risk, and consequence. Will automation improve the customer or employee experience? Does it protect trust? What judgment, learning, accountability, or human connection might disappear if a machine takes over?
Organizations need both answers before they deploy AI. A technically possible AI application is not automatically a useful, responsible, or strategically sound one.
Q: Can AI make work better—not just faster?
A: It can, but only if organizations use the time and capacity it creates well. AI can reduce repetitive work, help people access information, and support exploration. That is valuable when it gives people more room for problem-solving, judgment, creativity, and relationships.
But faster output is not inherently better output. If AI simply creates more work, more content to review, or more pressure to produce, it can intensify work rather than improve it.
Q: What should leaders do before implementing AI automation?
A: Start with the problem, not the technology. Be clear about the human or business outcome you want to improve, then examine the workflow and the larger system around it. Ask who benefits, what success would look like, what can go wrong, and where people need to remain in control. If a process is broken because of poor priorities, weak handoffs, or misaligned incentives, AI may accelerate the problem rather than solve it. The goal is not to automate everything; it is to make an intentional choice about how people and machines should work together.
About our guest
Stefanie Hutka, PHD, is a UX practitioner, author, and educator. She is the founder and head of design research at Sendfull, a boutique consultancy for emerging technology teams. She also teaches UX, systems thinking, and vibe coding at UC Berkeley.
Previously, Stefanie led design research and strategy for 0-to-1 product launches at Adobe, Meta Reality Labs, and DAQRI. She holds an ARCT Diploma in Violin Performance from the Royal Conservatory of Music and a PhD in Psychology from the University of Toronto, specializing in cognitive neuroscience.
Stef Hutka is also the author of What Your Machines Should Do: The Science and Strategy of Human-Centered Automation.
Quick reference guide
0:18 – Meet Stef
3:30 – Augmented reality – should we?
5:23 – The reality of the cost of augmented reality suits
7:17 – Augmenting human capability or replacing it?
8:47 – A basic framework, the Autonomy Decision Matrix, for a thoughtful, less risky approach
12:16 – Early conversations in the automation space are important
14:29 – Bloomberg LP
15:46 – Fleshing out the Autonomy Decision Matrix – a case study with Klarna
20:50 – The people who should be having these ADM conversations
22:48 – Humanist automation
24:57 – Steph’s role in Shift UX 2026
29:20 – Stef’s gift for listeners
