Leader Spotlight: The difference between refinement and delight, with Brian Crofts

Brian Crofts is Chief Product Officer at BambooHR, a cloud-based human resources information system. He began his career at Intuit, where he helped create the original mobile strategy and roadmap for the PTG business unit. Brian was CPO at Namely, an HR, benefits, and payroll platform, before assuming the role of CPO at Pendo.io. Before his current position at BambooHR, he also served as CPO at AuditBoard, a global platform for connected risk, transforming audit, risk, and compliance.

In our conversation, Brian talks about how insights make the difference between mediocre, expected refinements to a product and something that truly delights a customer. He shares his thoughts on AI advancements and product development, including the importance of human intuition and experience. Brian also discusses how products need to differentiate themselves to break out of the “sea of sameness” in SaaS.

---

Breaking out of the ‘sea of sameness’

You’re seeing a recurring issue that you referred to as the ‘lost art of the insight.’ What do you mean by that, and why do you think it’s happening?

I’ll start by talking about why insights are so important. Product managers, designers, engineers, and anyone who works on products knows when you hit something meaningful. Something clicks — you know your customer well, so when you develop a feature or product that goes beyond their expectations, it’s like you read their minds.

These kinds of moments happen to all of us in our personal lives as well. When you try to describe products that you love using, it usually boils down to the product reading your mind on what you wanted to do next. It anticipated the next step and went beyond your expectations.

The point here is that getting real insights is that critical differentiator. It’s the difference between the expected refinement of a product and a great product that delights. We saw this difference first-hand during COVID because we had to stop visiting our customers. My product team still interviewed customers, of course, but we missed those in-person, observational insights that show us a user’s true happy path.

In fact, most of these insights pop up when the team debriefs in person afterward. “Hey, did you see that? What did you think about it? I wonder why this is.” Then, we follow up and ask questions. Learning about customers is one thing, and that helps us validate what to build. But it takes work to get to a durable insight that we can build on for years to create a sense of delight.

You mentioned observing users’ happy paths. Could you expand on what that would look like during a user interview at BambooHR?

Let’s say we’re interviewing somebody for our product, which they use for their applicant tracking system. When we watch them in person, we can see that they might go between two other tools and also look at their monitor for the sticky note that has their username and password. When we’re doing a Zoom interview, we miss that opportunity to observe how they flow while doing their job.

Also, when you’re doing user interviews, people want to tell you what you want to hear — or what they think you want to hear — because it’s easier. So, instead of prompting them and saying, “Can you show me how you receive an update on this candidate?” you can watch them go through the flow in real life. That’s very different, because in the first scenario, they often leave out some of the details, and the details are where you can create a truly differentiated experience.

What impact do you see a lack of insight and conversation have on modern products? Is it that they are not anticipatory in the way that a user would want them to be?

My take is that all of those details together make the difference between mediocre, expected refinements to a product and something that really delights a customer. These points of discovery and insights that you get lead up to a very differentiated product. Now that SaaS is a huge and popular industry, almost every category within it has a “sea of sameness” problem. For example, when you look at a G2 grid now, it’s often not even useful.

Ten years ago, you could see the difference between newer but more innovative products and older, legacy leaders. Now, every software starts to feel the same. The challenge that every product leader currently faces is getting away from the sea of sameness and creating products that customers love — all in an era where the barrier to entry is getting lower and lower. The products that last will be the ones that have that discipline.

Using AI to get insights faster

Despite wanting specific feedback, you often have to rely on bigger data sets to make product decisions. How do you distinguish between big usage trends and the specific behaviors people say they do in interviews? And how do you make those takeaways function together?

I think we’re entering a newer era now where you don’t need to talk about it and you don’t even need to observe. You can put something in the market quickly by way of a prototype and see what users do. You can do that with larger numbers — track usage, see what people are actually doing, and look at how they’re behaving.

Again, you’re going to miss some of the things that I talked about earlier, so I still push people to talk to their customers and do formal observation work. But that’s just to get to a place where you have a strong enough hypothesis that if you build this, it’ll work.

Before, teams would only have one shot to create a V1 and V2, and they’d be operating on the same premise. But today, with increased cycle times thanks to AI, we can get prototypes out there faster. We can build more divergent prototypes and ship them to larger amounts of people. Now, when engineers build, they can do so with massive confidence and build at a very high fidelity.

Now, with AI, there’s a lot of autonomous product building where people use these tools to drum up their own solutions. How does this change the way that you think about serving a broad user base?

This is where it gets tricky. I love thinking about how we can take the productivity element of AI and apply it to building SaaS products. The concern that people have that SaaS is going to go away. I don’t think that’s the case,

My favorite approach is to apply new technology to things I already know. To me, this means using AI to research faster and at a higher quality for SaaS products. The hope is then to get better insights from that.

AI will never replace talking with customers. Instead, AI should help you create a better thesis so that when you meet with your customers, you’re educated and prepared. It allows that time you have with them to be more effective and meaningful. It’s also great to take all of the pieces of feedback and perspectives from your team and use AI to help you debrief it all. That will ultimately help you develop insights faster.

One of the negative aspects of AI is that people are already getting lazy with it. The product management practice is still the same, but I see people using AI to confirm a belief they have about their customers rather than talking to them directly. I’m seeing a lot of AI-written documentation, and it dilutes the person’s actual thoughts about the subject. We’re going to see more debriefs, write-ups, and white papers than ever before thanks to AI, but the question is, are they going to be any good?

Saving space for intuition and experience

Where do you go with that sort of understanding? AI can help people produce a lot of low-value content, so within your realm of work, how do you try to stay away from that?

There’s still a real place in the world for intuition and experience. There’s a lot of value in building an application just based on us being human, as well as working for and building for other humans.

The other day, I got a debrief from someone on my team about some research, and they’d used an AI tool to pull it together. And I thought, “Man, I don’t want to look at any of this. What do you think? What’s your intuition?” It’s great that we can ask questions and get answers faster, but the only way AI is going to make you better at your job is if you use it to get to a place that you could never get to before. You need to apply human intuition, judgement, ideas, and pattern recognition. Otherwise, it’s just noise.

You mentioned that before AI took off, teams would only have one shot to create a V1 product. How does the ability for teams to ship and test multiple things at once change how you build product components and release them to the market?

It’s still based on great product development principles. One of these is the double diamond approach, where you go broad, go narrow, and then go broad again. We used to have to start at the really low-fidelity research level — we’d think about all the possible things, create an affinity around a couple of areas, and pull on that. Then, we’d narrow that down, take those few things, and go broad again. That principle is great because it allows you not to anchor too quickly on the wrong problem — or solution — too quickly.

Today, you can apply that same kind of approach to actual working products. Let’s say you narrow down to the problem and the space. You can now put out seven distinct ways to solve that problem into the market. A few things could happen from there. One, there might be a clear winner, which usually doesn’t happen. Two, there could be a small handful of them that perform better than the others, so from there, you’d refine those few a bit and go back out again. Three, they could all be scoring the same, which means that you’re not solving a big enough problem.

I think about this with BambooHR and payroll, for example. This is a very deterministic product — we either get it right or we get it wrong. Yes, we want to provide a great payroll experience for admins to run. But the admins don’t care what the experience is like as long as it works every time and it can cover every use case. Because if it doesn’t work or something goes wrong, it goes from nobody caring to everybody caring and being really upset.

AI as an evolution vs. revolution

Do you think AI is ubiquitous now in product? If everyone is incorporating AI at these different levels and in different products, do you reach a point where everything becomes imitation?

A bit, yes — it can be an echo chamber. But also, if you’re rapidly prototyping and people have input in a way they never have before, there’s an argument that you could see more divergence than ever before.

The reality is that there are few companies and people who are true influencers — the ones that do have great product taste. This has been true forever, and they set the bar for what’s cool and what works. That still applies, but today, looking at SaaS apps, many of them look the same.

I’m hoping for fewer monolithic, big “take-over-the-world” products, and more right-for-me, bespoke things that have massive control over my workflows. They’ll take over everything that I don’t want to do in my day and take care of it all for me. In that sense, it’s the most exciting time ever.

What do you think people are getting wrong about AI right now, and what’s going to surprise people most?

There’s still debate as to whether this is a revolution or an evolution. I think it’s going to take a lot longer than some people have been predicting to get to how it’s all going to work in the future. I think the transformational potential is there, but I’m not quite seeing the gains yet. It could take 10–20 years to see the future that some people talk about happening tomorrow.

We’re seeing this right now with all the models out there and with where investments are going. We thought there was this AI race and whoever got to artificial general intelligence first would win and could do everything. But as we’re seeing, there are lots of big players. I’m not seeing any escape velocity. We’re seeing these tools become more incremental in nature, which is good. I love that the newest models aren’t blowing everybody away because it means that it’s more linear than exponential.

When it comes to applying AI to build better products and have more chances to do so, I’m excited about that right now. We’re using AI to ask better questions and do better research. That’s way more applicable than saying “SaaS is dead and AI is going to take away all these jobs.” We’re not seeing that. And I still don’t see a world where that happens. Instead, giving humans these really great tools produces some amazing outputs. When you take away the dystopian view, you see that we’re in a really cool timeline of continuing to build.

What does LogRocket do?

LogRocket’s Galileo AI watches user sessions for you and surfaces the technical and usability issues holding back your web and mobile apps. Understand where your users are struggling by trying it for free at LogRocket.com.