Insight Is Not Enough: The Next Chapter of Design Research

I'm going to be talking about a framework on insights of the function in the future but that does not mean that I'm a clairvoyant or any psychic who can just put her hand on a globe and say what it's gonna look like. This is based on my experience of working with a number of organizations. So a heads up and a disclaimer, it is not a rocket science. It's just that I've sort of collated and consolidated every learning from people into a framework that I see.

Some of the organisations will already be implementing that and some of them will be on the journey and someone will really have to start their journey really. But it still doesn't matter. It's just like as long as we are ready to see research as a practice. So the argument what I'm going to present today is that insights is still matter insights still matters, but insights on its own is no longer enough.

The future of design research is more traceable, more governed, more continuous, and more connected to decisions. And this is not basically a gloomy message for our profession or a discipline or as professionals. It's a sign that the discipline has become important enough to be asked harder questions. The opportunity now is to build the kind of research practice that can answer those questions confidently.

Okay. Let me start with a question which that sounds pretty familiar. Somewhere inside an organization, a number from a research is being quoted. It might be in a board paper, a case study, a funding deck, or a presentation you have never seen. Someone says, our research shows that a customer struggles with this.

And the number travels, the presentation, the quote travels, and the recommendation travels. But then months later, if someone asks you, where did that come from? Not which slide, not which document, not which project. What is the evidence of that number that is being quoted. So the question is small, but it changes the whole standard for research. The future of research begins with the trail that survives the presentation. And talking of trails, a lot of Abrahamic religions, a lot of prophecies that people are still practicing after five thousand years, the confidence in the prophecies often comes from what's the trail of evidence of whatever Jesus, Moses, or Muhammad said at some time.

Whether we want to believe it or not, but this is still the religious scholars of the day, this is how they go about it. Now, I wanna ask you this question and how common is this? That you've seen a finding from a work reappeared months later, slightly detached from its original context. Has ever anyone asked you to trace a number because of its source or it should was it harder for you to even find out? Or source turned out to be something like that.

Final? Really, really final? Please see this one. And I mean, this is seriously funny and I'm guilty of doing that where I've written on the folder, please use this one because this is the latest one. But it's a serious problem underneath. Our evidence often travels further than our documentation. When research says inside the project team, that may be survivable.

But when it moves to executive decisions, regulated services, AI systems, public accountability, it becomes less survivable. So here's the truth that we're going to be talking and this is my framework that I'm going to present or shift that I'm talking about in research as a practice. First is insight is becoming evidence.

The memorable finding still matters, but it has to be traceable. Second, AI is moving from a tool we use into an environment within we operate in. It's affect it affects our participation like Vian talked about, our analysis, and our evidence is produced. Third, the research projects are becoming learning infrastructure. The future is less about individual studies and more about continuous systems of organizational learning. And fourth, the findings are becoming decision led and outcome led. The question is no longer what did we learn?

It's also about what changed because we learnt it. So talking about the first shift, insights to evidence. This is not a rejection of insight. Many of us built our careers on the ability to notice what others missed, turn ambiguity into sins, and help organizations see customers and communities more clearly. But the environment around the research has changed.

The more influence research has, more it is expected to withstand scrutiny. A finding that persuades a roam is useful. A finding that can be reconstructed later is much more powerful. So the question for the next chapter of design research is whether we can produce compelling insights.

We certainly can. The question is whether those insights can survive challenge, reuse, decision making, and the project team when the project team has specifically moved on. The simplest way to think about is on the left hand side is a clear insight, clear, memorable, and easy to quote. 64% struggle to understand eligibility.

That's the line everyone remembers and it is also used and referenced. And this is how it goes. On the right side is the trail behind the insight. Who participated? How they were recruited? Which questions were asked? How the analysis was done? Whether AI assisted the process?

And which decision the finding informed? The insights tell people what we learn. The trail tells when we can trust that. For a long time, design research became very good at producing the thing on the left. The future also asks us to become just as good as preserving things on the right side as well.

The life cycle of an insight that we we just you know, I talked briefly about the familiarity of an insight when it's lost. It starts beautifully. You might have experienced a lot of time. There's an insight based presentation. Great insight. You would have heard of that. Great work. Well done.

And then it goes into executive deck, board paper. Months later, someone asks where the number came from, and a small archaeological expedition begins. And we call that the research archaeology. We search old decks. We search Slack threads, repository tags, half remembered project names. Someone says, I think it was in the version before that.

So the humor because what we're talking about here is so real and we can recognize that in the organization maybe making decisions from the evidence that no longer exists. And that is why traceability is not an admin, it is part of the evidence quality. Now it has also consequences.

In the Australian financial services, design and distribution obligations have made the quality of customer evidence more visible. ASIC has reported more than 80 stop orders since the obligation began, and one detail is especially important for researchers, flawed customer questionnaire, and they were identified as a catalyst for several interim stop orders.

That's a very practical warning. A research instrument can also become an operational risk. The lesson is not that every piece of design research is suddenly a regulatory document. The lesson is that in some sectors, especially where products affect people's money, eligibility, access of vulnerability, the quality traceability of research evidence become more much more important and more pronounced.

A poorly designed question is not always a poor question. Sometimes it's a weak link in the decision system. So this basically leads to an evidence letter, and this is a, you know, sort of a framework, sort of a mini letter I suggest. At the first rung, research is episodic and this is the most common form of research we see.

A study happens, insights are delivered and the project ends. And that's where often basic regular day to day design research looks like. At the second rung, research becomes more repeatable. There are consistent protocols, repositories, and reusable patterns. At the third rung, it becomes more auditable. And that's where I think the research ops become more, you know, come into play. We can reconstruct providence, consent analysis, evidence, and use of decision links. At the fourth rung, it becomes contestable.

People affected by decisions can understand questions and seek review. Now this is not a sort of a maturity risk that every organization has to achieve that. Not every study needs to sit at rung for. A marketing concept test and a hardship eligibility journey do not carry the same consequences. The real leadership question is, what does this evidence need to survive once we know that we can design the right level of rigor? The second shift from my framework is about AI.

Not surprising. Many conversations about AI and research begin in the wrong place. They ask whether AI will replace researchers. That's a fear everyone is carrying here. Or whether research researchers should use AI tools. They are not unimportant questions, but they're too narrow. The bigger change is AI is becoming part of the research environment.

It is in the tools we use. It is in the systems we study. It is increasingly in the way participants express themselves. So the future question is not simply, can I use AI to move faster? It is how do I know what kind of evidence I now have when AI may be present at multiple points in the chain or in the user experience.

Recent Australian digital inclusion index figures shows that AI no longer is a niche behavior, especially high among students and younger adults. Before research, this matters all because AI is now sitting sits quietly inside the organization. Think about the open ended comments, or verbatims as we say, or in an unmoderated diary study or asynchronous task at the end of a day long.

A participant may use AI to turn rough thoughts into a polished language. They're not trying to mislead us. They're simply using an everyday tool, but it changes what we are analyzing. Are we reading the participants lived experience or a model's cleanup version of it? That does not mean we throw that data away.

It means we acknowledge a new form of bias and document how we manage it. So looking at this document here or having a conversation about a very old research problem, a research has five words, AI polishes them, a survey platform summarizes a response, a researcher uses AI to cluster the themes, An executive uses AI to summarize everything all the way from top to bottom, AI led.

Congratulations. We have built a very difficult or efficient game of telephone. Again, the problem is not that AI exists in the chain. The problem is when nobody knows when it enters the chain, what it changed, and who remained accountable for the interpretation. Research has always, involved mediation. We summarize, interpret, and translate.

AI simply makes the mediation faster, less visible, and easier to forget. That's why the future is not AI free research, it is AI aware research. So let me ask another practical question and I would say now, because I've listened to a and I loved it, so I'm referring to you again and again.

Does your organization have a written AI boundary? Not a general policy that says people should use AI responsibly. A research specific boundary that says what is acceptable, what is risky, what must be documented, and where AI must never substitute for human participation or judgment. Some teams have written this down, some have a shared understanding, but nothing findable.

Many are still relying on individual judgment. The problem with the individual judgment is it does not scale. It does not survive staff turnover, procurement, scrutiny, or executive decisions. So a written boundary is not bureaucracy. It's a simple way to protect trust in the evidence. So an AI bounded AI practice gives us three categories.

And again, a small categorization or grouping to understand. There are users where AI can assist, like very easy and simplistic to use, Transcription support, translation support, repository reveal of first pass summaries, provided the source remains traceable. But even translation, and yes, I still use AI, but it could be really funny with some of the languages which are not spoken outside the European or the English speaking world.

Really funny and hilarious translations that I've seen. Then there are uses that require governance. For example, coding support, clustering, prompt based analysis, sense making at scale. These may be useful, but the process needs documentation and human accountability. And then there are very users where we should never normalize AI, and that is synthetic personas replacing real participants, AI generated quotes presented at evidence, or AI only discovery supporting decisions that affect people's lives.

I tell everybody, I'm glad I did a PhD when AI did not exist. So I wrote those 100,000 words by you know, through a lot of hard work, blood and sweat. So the point is not to ban the tool. The point is to draw the boundary before the boundary is tested. A team that can explain its practice will be be much better positioned than a team that says we are careful.

Now the third shift, part of the framework is from projects to learning infrastructure. For many organizations, research still operates as a sequence of projects. A team has a question, a study is commissioned, findings are delivered, and then everybody moves to the next priority. The model will not disappear, but it will always exist and projects will still exist. Excuse me.

But the future of design research cannot only be a cure for individual studies. The issues that we are researching in 2026 are two continuous, trust, access, vulnerability, digital exclusion or inclusion for that matter, AI harm, service performance, long term behavior change.

Those issues do not fit neatly into quarterly project cycles. So research has to become something more durable, a system for continuous learning, not just a service that responds to requests. This is a shift from a calendar to a system, as I suggest. A calendar study tells us what's happening in this quarter.

It is useful, but it's not enough. A learning system asks different questions. What signals are we watching continuously? What evidence would I tell to our tell us to stop, scale, or redesign something completely? What are we monitoring after launch? What are the equity, trust, and harm signals?

This is especially important because many of the important research questions do not end at launch. A service can test well and still fail particularly over time. An AI feature can perform acceptably in evaluation and still create harm in the use. The annual plan tells us what we intend to do. A learning system tells us when the organization is committed to or noticing.

And this is not a European or a Silicon Valley concept. We can see versions of it in Australia as well. Monash University Living Labs program involved more than 18,000 participants or stakeholders in 2023 to 2024 with more than 100 partners, more 100 more than 100 researchers, and more than 1,000 students.

Now this is not a one off shop. It's an infrastructure for learning across complex systems. And I'm not suggesting every organization to adapt Monash model. The point is the operating principle. Ongoing participation, experimentation, and learning becoming part of how the institution works. For design research, this is a future signal.

The durable value is not only the method, it's a system that allows learning to continue after an individual project completes or ends. In Australia, research infrastructure has to meet two conditions. And I'm so glad Bhavan and Chanel talked about their presentation where they did almost like a hybrid or more of in person research.

In Australia, it has to be hybrid by design. A continuous panel recruited only through an online portal can look efficient and well governed, while still missing people who do not have reliable connectivity, private devices, sufficient data, confidence, and digital systems. When I first arrived to Australia, I couldn't believe in Australia, internet is slower than my home country, which is considered a third world country.

It was obviously, it has evolved over the years. I've been here for a good fourteen, fifteen years, but if you look at the regional areas in Australia, there's still a big struggle of technology. So it's basically we have to design systems that are culturally neutral. Research involving First Nations people must take indigenous data sovereignty seriously, including authority to control benefit sharing and culturally legitimate governance.

AI makes this more urgent because data and insights can now be reused, remixed, and summarized at speed. Australian research infrastructure cannot simply be digital first and then inclusive later. Inclusion and cultural governance have to be part of the architecture from the beginning. This is a provenance first workflow I suggest.

It begins with the intent. What decision is this research informing? What risk exists and what may be affected? Then the participation. Who is included? Who may be excluded? And why are they excluded? What support is needed? What consent governs the work? Then capturing analysis, how the work was collected, how it was interpreted, and where did AR automation play a role?

Then evidence packaging. How do we show not only the finding, but also the uncertainty, alternatives, and the limits of the insights that are being produced? And then finally, the decision link. What choice did this evidence come from? And also the monitoring. What happened after the decision? What would trigger us to do the research again?

The research conversation may look familiar. What changes is what survives after the study. And the fourth shift that we are going to be talking from the framework is the findings delivered to decisions and outcomes. I'm hearing more and more, and I read a lot on LinkedIn, and there's so much of conversation around decision, linking to decisions and outcomes. And this is most important for the senior stakeholders, senior leaders, because it connects research to value.

A finding delivered is not same as a decision improved. A beautiful tech is not same as a reduced risk, and a strong code is not same as better access, greater trust, or less harm for the participants. The future asks researchers to stay closer to what happens after the readout. This does not mean pretending research alone can fix all the organizational outcomes or problems.

It means being much clearer about the decisions, research informed, the assumptions it took, the risk it surfaced and the outcome should be monitored. In other words, research needs to be more accountable without becoming too simplistic. This is one of the simplest changes with the biggest implications.

Most repositories capture the study title, method, date, and the research lead. But the missing field is often which decision did this inform. Imagine opening a study record and seeing something like hardship onboarding interviews, but also the decision it informed, the product owner, the risk reviewed, and a follow-up signal to monitor.

That changes from a library of outputs into a memory system for decision. It also changes the conversation with executives. Instead of saying, here's what we found, research can say, here's the decision this research has supported. Here's what changed, and here's what we still are missing and we need to watch, and that's a different level of influence.

Okay. I'll skip this one. Now I want to bring the four shifts back to the room and I wanna ask you a question which one's happening in your organization. So first is One is evidence. People are asking for traceability. How many within your organizations people are asking for traceability? One, two, three, okay.

And then the second one is AI is changing the method are simple. How many times the conversation is how much of AI has been utilized in the process of synthesis or insights writing? Okay. Third is the infrastructure. Research is moving beyond projects into continuous learning. Are you seeing this change that we're not talking about projects?

At least in the year, I was aware of that. And then fourth is the impact. Stakeholders want clear links between research and outcomes. Okay. So this is a point of the framework we are talking about. It gives leaders a way to name what's already changing rather than treating every pressure as a separate, you know, problem.

Once we can name the shift, we can decide what needs to be built next. So here's my forecast for 2026 to 2031. Again, not a psychic or not a clairvoyant and summarizing everything that we have learned. The job title may still be researcher, research lead, strategic design strategist, insights director, but the deliverable audience and the shelf life will change. Evidence will move from persuasive readouts to decision grade evidence that can be reconstructed later.

AI will move from ad hoc use to documented, governed, and bounded practice. Infrastructure will move from project by project to continuous learning systems. And impact will move from satisfaction with findings to evidence of decision change, reduce risk, outcomes monitor. And this is a future I see for design research.

Less like a presentation factory, more like a decision capability. And I see that as a more powerful and serious role. And I seriously believe we are still going to be well employed at least for the next foreseeable future. So don't worry. Okay. And the good news is Okay. The most important thing that I wanna share here Sorry, just need water.

So with the framework that I'm suggesting, the very good news is that it does not require a five year transformation project. There are some of the changes that can be made very easily starting next week. So I would suggest four simple moves. The first one is for evidence, add one field to the repository, which research decision did it inform and where it came from.

For AI, publish the boundary where AI can assist, where it needs governance, and where it must not substitute for real evidence. It will take a little longer than a week or so, but it will still give you a good clarity or headspace what needs to be done. For infrastructure, choose one signal to monitor after the launch. Instead of treating research as a finished ad delivery, and also for impact, link one study to one decision.

One risk or one outcome. Now this is going to be the tricky one because often there's going to be conversations and you know, just pinpointing and highlighting one decision, which one decision. But the moment we start writing, and that's the beauty of writing. I shared this with Anja and I read it on LinkedIn and I loved it, that think like an engineer, share like a poet.

And I feel there's some bit of poetic or poet in all of us as researchers. We express we wanna do the storytelling in a way is people connected with that. So if we just start working on that in terms of the documentation and start coming up with what is one decision, what is the one thing that we want to connect our outcomes that will start to make sense.

So all these small operating changes create new habits. New habits obviously create new expectations and new expectations are how disciplines mature. So the future of design research begins in the very practical places. And this is where I want to leave the argument. Insight is not enough, not because insight has become less valuable, but because the world around has become more consequential.

The future of design research is evidence, judgment, infrastructure and impact. And the responsibility comes with a simple expectation, show you're working, stay close to decisions and measure what changes. And that is it. And I finish early on.