Embracing the Messy Middle of AI

Brooke Sykes is a senior manager of Firefox user research at Mozilla. A dedicated champion for a healthier, more open internet, she works closely with cross-functional leadership across product management, design, and engineering to ensure data-driven insights actively shape product strategy. Beyond research leadership, Brooke is focused on delivering operational excellence, team empowerment, and scale, and regularly evolves organisational practices for the Firefox community. Allison Robins is a mixed-methods researcher with nearly a decade of experience across B2B and B2C contexts, and currently leads research for Mozilla Firefox’s productivity workstream. She’s spent the past few years experimenting with AI in real workflows, separating hype from substance, and learning how to use these tools in ways that sharpen thinking and protect the parts of the craft that matter most. How to AI UXR is supported by Strella (https://www.strella.io/) , an AI-powered customer research platform that partners with you to build, moderate, and synthesise interviews, allowing you to go from question to actionable insights in just a few hours. In This Conversation For many research organisations, AI adoption has moved beyond play and into the trickier work of reshaping people management, practice, and team culture. Leaders are asking teams to move quickly, while individual contributors (ICs) are still trying to understand what good use looks like, how much judgement they should retain, and how their competence will be perceived in a working environment whose norms are…well, not yet normal. In this final episode of How to AI UXR, Brooke Sykes and Allison Robins from Mozilla’s Firefox research team give this, shall we say, “unsettled period” a useful name: the messy middle. Brooke speaks from the manager’s side of that transition. For her, the job isn’t simply to tell a team to adopt AI; it’s to translate a broad mandate to match her team’s routines, personalities, and principles, create space for experimentation and psychological safety, recognise uneven skill levels, and be honest that managers are also learning on the job. Allison brings a practitioner’s vocabulary to one of the most overused phrases in AI work: human in the loop (HITL). Rather than treating HITL as a single activity, she describes three modes of AI use: the self-automator, the centaur, and the cyborg, as outlined in a Harvard Business School paper titled, “Cyborgs, Centaurs and Self-Automators: The Three Modes of Human-GenAI Knowledge Work and Their Implications for Skilling and the Future of Expertise. (https://www.hbs.edu/faculty/Pages/ite...) “ While each mode has its place, none is sufficient on its own. Instead, research judgement lies in knowing when to hand off a bounded task, when to retain authorship while delegating parts of the work, and when to stay in continuous dialogue with AI. Together, Brooke and Allison offer a pragmatic way to think about AI maturity in research. Their aim isn’t to resolve the messy middle or to pretend that they’ve already developed a settled practice. Instead, it’s about being clearer about the risks, more precise about AI evaluations, more tolerant of uncertainty, and better equipped to choose the right kind of human involvement for the work at hand. The How to AI UXR Map This series builds on the insights shared in the How to AI UXR (https://www.theresearchopsreview.com/...) map (https://www.theresearchopsreview.com/...) , a five-page map that charts key trends, helps you pinpoint your AI maturity level, and offers practical, real-world applications you can adapt to your research systems. Download the Map (https://www.theresearchopsreview.com/...) In this episode, we cover: Why “adopt AI” isn’t a strategy, and how research managers must make practical decisions about workflows, experimentation, guardrails, and verification to achieve that goal How uneven AI literacy affects team adoption, with some researchers already using AI as an always-available thought partner while others are still building confidence and fluency Why psychological safety matters when teams are learning AI together, especially when people need room to ask basic questions, make imperfect experiments, and revise their views as the technology develops How Allison frames the central tensions of AI use, including the fact that AI can accelerate individual output while creating more work for others, and can help people build skills while also giving them misplaced confidence Why the social stigma around AI use can leave researchers caught between pressure to use the technology and concern that visible use will make their work seem lazy, superficial, or less credible How the self-automator, centaur, and cyborg modes offer a more useful vocabulary for deciding when to hand work to AI, when to retain strategy and judgement,...