Building AI Tools That Researchers Actually Want: Lessons from a Year and a Half of Rapid Innovation

By Conveo

• article

• AI
• Artificial Intelligence
• Qualitative Research
• Conversational AI
• Video Research
• AI Interviews

When I first met the Conveo founders, the product was what you’d call minimal, definitely not viable yet. But what it could already do made a big impression on me. That early glimpse into possibility, combined with my decades of experience in the insights world, convinced me to embark on this AI-powered entrepreneurship journey.

Looking back at my former life as co-founder and managing partner of Human8 (formerly InSites Consulting), I’ve always been intrigued by transitions – from postal surveys to telephone, telephone to online, and the emergence of digital communities. There were already hints of automation and AI in those days. We experimented with chatbots moderating communities, so I always felt this was an avenue the industry was heading toward.

But building AI tools that researchers actually want? That’s proved to be both more challenging and more rewarding than I expected. With that said, the last year and a half have brought many lessons that I’d like to share.

This article covers part of the The Founders & Leaders Series podcast episode 4. Listen to the full episode here:

Lesson 1: Innovate with Relevance

Here’s something I learned the hard way at Human8, and it applies even more in the AI space: develop products that people actually want, not just what they think they need. Innovate with relevance.

Sometimes I see innovations in our industry and I think, “Yeah, it’s nice, it’s a great idea, but do people really want it?”. We’ve built features at Conveo and presented them to clients, only to receive lukewarm reactions. The key is not to pursue those features just because they’re technologically impressive.

If a client doesn’t want it, don’t build it. Otherwise, you’re making products and services that collect digital dust.

Lesson 2: You Need Both Worlds In-House

One critical lesson from watching agencies try to outsource development: you can’t separate the research expertise from the technical capability. Engineering talent must be paired with market research knowledge, and you can’t have one without the other.

Agencies might outsource development to get something started quickly, but tech evolves so rapidly that you need to iterate fast and compound your knowledge in-house. The reverse is also true, pure tech companies need to infuse their platforms with deep market research understanding.

This is why having methodologists who understand rigour and quality insights is essential. When AI technology spits out something that doesn’t make sense, someone needs to recognise that immediately. When it produces an excellent analytical framework, someone needs to spot that too and translate it into actionable features.

Lesson 3: Learn from Your Mistakes (and Let Others Make Theirs)

One of my key learnings is that you need to learn from your mistakes, as they are essential to your success. I’m not at an age or experience level where I have all the answers either.

In fact, I’ve always enjoyed learning from people like Hendrik (Conveo’s Co-Founder and Chief Product Officer). He’s much younger, but incredibly smart. Learning from younger people keeps me going and sharp. I was always at my best when the people around me were intrinsically smarter than I was. That dynamic keeps everyone elevated. Talent is also a huge element in this equation.

Lesson 4: Develop New Skills for a New Era

There’s an extra layer that’s emerged in AI research: learning to “speak prompt”. When you’re sitting with AI on a huge pile of answers, extracting insights and combining them effectively requires a different approach. It’s not rocket science, but it’s not the traditional Boolean way of asking questions either.

This prompt literacy, whether for extracting data or instructing AI platforms to ask questions in specific ways, is a skill we all need to develop. It’s not necessarily product management or tech, but it is about learning how to use new technology effectively.

Lesson 5: Bridge the Translation Gap

My role often involves being the translator between client needs and technical capabilities. I love sitting in front of clients with our developers – I can lean back and let the client explain exactly how they work. The developers are eager to listen, which is the first step, but then, within weeks, there’s often a solution.

We speak the language because we understand both worlds. We have the technology capability, but we also understand the rigour required for quality insights. Being very customer-oriented isn’t just marketing speak, it’s the core of successful AI research tool development.

The Reality

Here’s the honest truth: even with all our AI capabilities, it’s always beneficial to have an expert review the output, interpret it, and provide guidance. Sometimes it boils down to small tweaks, but you’re operating at a much higher level.

In the end, people solve problems, not tech. Technology helps you reach solutions quicker than you would traditionally, as it exposes you to more and better solutions, but you still need to make the decisions.

Building AI tools that researchers actually want means understanding that we’re not replacing human expertise, we’re amplifying it. And that makes all the difference.

This article covers part of the The Founders & Leaders Series podcast episode 4. Listen to the full episode here:

Author

Learn more about