What does ‘Good Work’ look like, in the AI era?

The origin essay behind the report, on craft, ‘disembodiment’, and deliberately sharing the act of thinking between people and machines.

Marco RovagnatiCo-founder, Quallie.Ai

Marco Rovagnati · June 2026 · 6 min

When I asked Andy Crysell (founder of CrowdDNA) for advice, he pointed out that “AI research” is overly focused on technical benefits and time savings, ignoring the emotional needs of practitioners and the “craft” of research.

He had a hypothesis, based on personal experience, that AI makes work feel “disembodied” and disconnects practitioners from the thinking when they outsource key parts of it to a machine.

With this in mind, we set out on a quest to speak with researchers about this topic, starting with a big, broad question, “What does Good Work look like, in the AI era?”, and a follow-on: how might the act of thinking be deliberately shared between people and machines to get there?

Andy and I spoke to some of the most interesting voices in research, which led to the Good Work Matters report. In this article I’ll reflect on the main learnings and what they mean for practitioners. To inform it, we conducted semi-structured interviews with 30 professionals across the US and UK, spanning independents and boutique firms through to researchers at large organisations like Expedia, Universal and Ipsos.

Defining Good Work

There are many colourful definitions of Good Work, but three interdependent qualities are ever present. Good Work is actionable and integrated into decisions, it might seem obvious, but there’s a lot of research-for-research’s-sake out there. Good Work goes “viral” inside an organisation: high-quality output gets passed around rather than forgotten in a folder. And Good Work is earned through a process of “crystallisation”, it requires clear thinking, sound judgment and the inherently human ability to sit with ambiguity.

This is not new news, but something shifted in the foundations and toolkit of how people get there. It is now possible to reach polished, sleek, directionally correct insights while skipping the last step of human crystallisation. That’s where the “disembodiment” begins, and the psychological ownership of the researcher slowly slips away, one prompt at a time.

What we learned is that ultimately Good Work requires ownership: someone able to stand behind the thinking and justify conclusions, or curveball questions, with conviction. Whether we like it or not, we inevitably lose the ability to reconstruct the intricate details of the trail when we helicopter to the destination.

Beyond the human vs AI debate

Much of the debate around AI in qualitative research is stuck in false binaries, oscillating between outright rejection driven by fear, and res-tech evangelism that insists automation is the only viable future, often with vested interests on both sides.Andy Crysell, CrowdDNA

In contrast, we found that researchers producing high-quality, strategic insights were comfortable occupying the middle ground. They’re navigating change, but then, they always have been. They’re bringing AI into their practices while staying just as focused on what will actually future-proof them: avoiding a race to the bottom, protecting craft, and defining their own version of the work in the AI era.

A big theme came out of our conversations: that Good Work depends on how deliberately thinking is distributed between humans and machines. The debate is not humans vs machines but a matter of intent, knowing which forms of cognition must remain human, and which can be responsibly delegated to a system without hollowing out the work.

What must remain human

Several researchers used the term “alchemy of insights” to describe something that cannot be outsourced to AI: the process of turning hours of conversation, messy data, past decks, trends, focus groups and ethnographies into not just a coherent point of view, but a transformative story.

The discipline of that alchemy is made of three things, deep reflection on the strategic “so what”; asking what to actually do with the information, and why it’s relevant; and interpretative judgment paired with precise writing skills.

Core principle: protect thinking time

Researchers and strategists doing Good Work delegate the “what” and the “how much” to AI in order to protect time to define the “so what” for clients. They intentionally point AI at the tasks that drain cognitive energy, and never at choosing the insight or writing a deck’s narrative from scratch.

Practitioners described using it to test their own blind spots and challenge their judgment, during or after an analysis session:

I tell AI what I think the main themes, tensions and contradictions in the research are, and ask if there’s more, and whether I missed anything.Interview participant

To speed up evidence finding and clustering:

The best AI tools can now efficiently help you find all the quotes about a specific theme and visualise them in grids, with links back to the video.Interview participant

And to take the grind out of video editing:

Some AI analysis tools now offer text-based video editing and reel creation. I don’t have to spend hours trimming, marking timestamps and editing clips together.Interview participant

Ownership is your added value

In the AI era, our added value sits in our ability to “own” the work. When a curveball question lands in the room, we need to justify our point with conviction, and that’s only possible if we’ve lived and breathed the work. There is a role for AI, but it requires disciplined, intentional use: one where we remain responsible for sense-making and in control of the output.