AI Accelerator: Matt Reed, Director, Product Design

In this installment, we talk to Matt Reed, Director, Product Design, who has watched AI move from a graduate-school curiosity to something that quietly rebuilt his day-to-day work and who argues the hardest problem left isn’t intelligence, it’s focus.

The AI Accelerator series is an ongoing collection of conversations with people across Confluence who are at the forefront of how we build and use AI.

From the intelligent features embedded in our solutions to the tools colleagues increasingly leverage to get their work done, each installment hears directly from people across teams and disciplines. Find out what’s working, what isn’t, and what they’re still figuring out. Ground-level perspectives from the people living it every day.

Q: You have been studying and following AI for a while. What has the build-up looked like from your perspective?

A: I first seriously studied Neural Networks in Graduate School twenty years ago, before they were popular, and experimented with rudimentary language modelling. But at the time it really wasn’t practical for most businesses and engineering teams to integrate AI of any kind into their products. The breakthroughs that helped popularize and scale Deep Learning were a decade away.

I followed the field casually afterward, and vicariously through colleagues in the industry. In that sense it was a slow build, but there was a moment where everything changed and that was spam filters.

There actually was a time before we had spam filters, and I was alive in those times. Of course, their arrival didn’t completely stop junk mail, but especially in the beginning they made a huge difference. What really excited me was that the algorithms behind them weren’t esoteric or difficult; rather they were ones I could easily understand, ones I could build myself. This AI wasn’t something reserved for lab coat wearing PhDs. I could easily imagine this type of AI improving the user experience for products that I was building at work.

AI became a practical tool in my kit. For instance, I designed one of Confluence’s first machine learning features when we shipped intelligent payment categorization in our NXT Expense product in around 2014.

Q: What more recently has excited you in the wave of progress we’ve experienced?

A: Specifically for Generative AI, I was very excited when I discovered Open-AI’s GPT-3 API in 2021/2022. It felt less like a technical breakthrough and more like discovering magic. This was before ChatGPT: to do anything useful, you had to know how to code, and you had to pay for the usage. We should note how incredible that is; just five years ago you had to be technical to even access AI and today orders of magnitude more powerful AI is free to use on anyone’s phone!

Just in the last few months AI has developed to the point where even if the technology stopped progressing, frontier models are intelligent enough to be transformative not just in software development. This will be especially true if the cost to generate AI continues to go down. In this new world, intelligence is plentiful, but context and focus remain scarce. So, you could say I am skeptical of natural and artificial intelligence in the absence of context. The good news is that many of the things that help AI with context help humans too: articulating and socializing clear goals, increasing transparency, writing clear standards and processes, reducing needless friction.

There is still something special about in-person collaboration that I haven’t seen replicated online even after years of widespread remote work. You lose a lot when going from an in-person meeting to a video call, and even more when going from a video call to a text chat. I think as long as AI is stuck there in a text chat, there will be a limit to what it can do as a coworker. Recently, voice interaction models have improved, which is great. But I would love there to be a world were AI can frictionlessly draw a solution together with me and my team on a whiteboard.

Q: How are you using AI within the products you build to deliver a better solution to clients?

A: We recently launched Confluence POINT, which is embedded AI that runs within and alongside our solutions. It enables our solutions to do more for our clients, with greater efficiency. We are also working on making it easier for AI agents to be users of our products. A key component is the data foundation. To move fast and to scale, you need a solid foundation. This was true before the AI era, but it is even more important now. This is one reason why AI is a natural complement to Confluence’s products: our products have always been a solid foundation for data, analytics, and reporting.

On my team, we were early adopters in using AI to prototype and communicate potential solutions to business problems, helping deliver solutions faster for our clients. And that’s the point of all this: solving problems.

To move fast and to scale, you need a solid foundation. This was true before the AI era, but it is even more important now. This is one reason why AI is a natural complement to Confluence’s products: our products have always been a solid foundation for data, analytics, and reporting.

Matt Reed, Director, Product Design

Q: How are you using AI typically right now?

A: I use AI extensively both in and outside work. There are AI experts, there really aren’t any AI applications experts. There’s no definitive book yet, or class you can take. This is all too new. I am trying to figure it out for myself so that I can bring what I learn to others or at least be a good example, so they are inspired to try their own things. This means experimenting with the latest tools and with new approaches to using tools.

Whenever you encounter something in your work that is brittle or tedious or manual, you have a choice: continue to do it the way you’ve always done it or change the system and create automation so that the activity is more robust and repeatable. Maximum automation is always preferable, but in the past, it was prohibitively expensive to develop most of the time. One of the things AI does is make creating automation so inexpensive that it is almost always better to change the system, so activities are more robust and repeatable.

At the smaller end, this applies to my day-to-day work. On the larger end, this is a big part of what Confluence does for our clients, and I believe that we will be able to use AI to make it easier for our clients to have more robust and repeatable processes too. At home, I have been attempting more home maintenance projects than ever before. For example, I’ve been getting better at plumbing: ChatGPT is a patient teacher.

Q: What is the AI development you are watching right now – and why?

A: In the past, most of the AI conversation has focused on individual productivity and efficiency. As the tools have gotten better, there are lots of innovation and experiments happening on how to both do more with AI as well as how to scale AI to transform teams and organizations. I think that the structure of teams and the processes and workflows that govern teams will change.

One of my big worries about AI is that we will become more isolated and less aligned because we are inward focused on our own work and our own chat. The opportunity is that we will use AI to be collectively more creative, disciplined, and focused.

Q: How is AI going to change how people approach your profession?

A: It’s a scary time to be starting a career because everything is so new, but for the same reason it’s also an exciting time. It’s also new for everyone. We are all fresh graduates now! The extra advantage graduates have is that they know they are fresh. One of the dangers I see in AI is that those of us who are further into our careers become overconfident. AI can get us a superficial understanding of any topic almost instantly, and AI can build 90% of any product almost instantly. And at the same time, many of our intuitions about how the world works are no longer accurate. We all need to bring a learner’s mindset to our work. We should all be much more ambitious, but at the same time, we should be more intellectually humble.

Q: Has using AI increased or decreased your enjoyment of your profession? Do you miss anything you no longer do?

A: I’m learning more. I’m creating more. I’m collaborating with a more diverse group of people. What could be more enjoyable than that?