Wassup Mehdi!

Thanks for jumping on. I’ve heard a lot from the team about the acquisition. V excited for the inside scoop today.

Take me back to the start. Why did Cycle exist in the first place, and how did the Atlassian conversation actually begin?

8:00 AM ✓

Thanks for having me! Loved the last JPD interview series, so it’s cool to be here 🙂

The story is essentially this: I was a PM and I was frustrated with the tooling that was available to me, because everyone was talking about being 'customer obsessed' but the customer was nowhere to be found in the stack PMs used day-to-day.

The most important thing in building trust with customers (and internal customer-facing teams) is making sure they feel heard. That when they share feedback, it doesn't enter a black box. That closing the loop was so important, and so hard to do.

So I thought, okay, there's an opportunity to obsess over this problem.

We did that for 6 years. The first 4 were pre-AI - very manual, PMs spending hours tagging insights.

Then AI came and we started shipping the feedback autopilot. Capturing from Slack, JSM, Zendesk, Zoom calls, all into one place. The autopilot extracts insights, links them to ideas, closes the loop automatically when an idea ships.

That's when we hit product market fit. And that's about when Tanguy from JPD reached out.

8:08 AM ✓

In that first discussion, Tanguy was just curious. “Oh, that feedback use case, that's interesting”. No urgency.

Then a year later, he noticed it was becoming the single biggest pain point in JPD customers. They were hacking feedback together themselves, polluting backlogs with ideas that should actually be feedback.

So from the Atlassian perspective: they need this. Do they build it or acquire?

They acquired us because it was effectively a 6-year head start on a problem we'd been obsessing about.

8:11 AM ✓

I heard through the grapevine that Cycle had a few hundred customers vs JPD's 20k+.

Pretty much a rounding error.

So the acquisition probably wasn't really about the customer base. How do you see the acquisition from their pov?

8:13 AM ✓

I think about it like this: in an acquisition you can buy 3 things: the business (customers, ARR), the product, and the team.

In this case Atlassian didn't want the business. It was too small.

It was really the other 2.

When we joined, there wasn't a single shipped AI feature in JPD. We'd been obsessing about AI for 2 years, processing millions of pieces of feedback every month. That know-how, all the thinking that went into building a feedback app the right way - that's what they bought.

And with it, the team that's going to take that know-how and rebuild it within Atlassian.

8:17 AM ✓

When you joined - did you end up sunsetting Cycle? I imagine that would’ve been a hard call. How did that conversation play out?

8:18 AM ✓

Yea we did - probably the hardest part of the journey!

We had customers who'd built entire workflows around Cycle. The reaction to the shutdown told us how much they'd come to rely on it.

But ultimately it was the right decision. Tanguy had been through other acquisitions before, and one of his learnings is that when you keep the legacy product while trying to integrate and rebuild, you end up maintaining 2 products at the same time.

Very costly.

Ultimately it didn’t make sense to spend a lot of effort maintaining a legacy product when we should be entirely focused on the future. The goal is how fast can we get from 20k to 100k customers together, and keeping Cycle running wasn’t a viable component of that strategy.

And from your side as a founder, what tipped you toward joining vs staying solo?

8:23 AM ✓

We had product market fit - the product finally worked with AI - but distribution was a huge challenge.

I was spending 80% of my time on sales and content. I'm a product guy. But I really didn't have much time anymore to spend on the product.

Atlassian has 300,000+ customers. Distribution wouldn't be a problem. Scale wouldn't be a problem. Joining meant I'd be able to focus entirely on the product.

By joining forces, we'd scale our impact many magnitudes further than continuing as a standalone tool.

8:26 AM ✓

Before we get into how it bundles up - walk me through the actual loop for how feedback happens inside JPD.

Step 1, 2, 3 - what's the opinionated POV?

8:27 AM ✓

Here's the model end-to-end:

It starts with capture. Feedback comes in from everywhere - Slack, Zoom, JSM, Zendesk, Teams, Gmail, custom APIs. We pull it all into one place.

Then AI does the heavy lifting. It extracts insights, summarises the customer voice, surfaces the patterns:

That feeds into voice-of-customer dashboards and JPD ideas.

Now here's our opinionated bit. In a JPD idea, you can write a release note - even before you ship the feature. Apple-style. The release note lets you get to the essence of the outcome you want to achieve.

Each idea is linked to Jira. When the work items move to done, the idea moves to done, and the release note can auto-publish to a public changelog.

That’s where it closes: each release note is linked back to every piece of feedback that informed it. Slack, Zendesk, JSM, HubSpot, wherever it came from.

In 1 click, you can fire automated communication back to all of them. Maybe you shipped one release note, but you had 150 feedback items linked to it. One click, and you're back to all those folks: ”Hey, we heard you. 2 months ago you said this. Look at the release note, tell us what you think.”

When you do that systematically, you create a virtuous cycle. Customers turn into co-creators. They want to give you even more feedback. You build trust.

8:33 AM ✓

Love it. So the loop also closes back to the internal teams who logged the feedback, right? Sales, support, etc?

8:34 AM ✓

Exactly. That's the 2nd-order effect.

In B2B, most feedback comes from internal teams - support, CSM, sales - because they spend the most time talking to customers. So the loop has to close back to them too, not just the end customer.

Once it does, the math changes. Sales gets a notification straight in Salesforce: “this deal was blocked because the integration hadn't shipped. Well, it just shipped. Here's a reason to reach out.” CSM gets the same on retention.

Gets them closer to their KPIs.

Without that loop, sales just stops giving feedback. They're busy people with their own KPIs, and if it disappears into a black box they don't bother. The loop isn't a nice-to-have. Without it the whole system breaks.

8:40 AM ✓

So the loop literally helps the sales team hit quota. Love that.

Tell me about this new ‘Product Collection’ I keep seeing. When I was Atlassian we spent a lot of time thinking about the best way to bundle / combine / split / position products.

Would love to know what the latest thinking is.

8:41 AM ✓

The Product Collection is our bundle for product teams.

It includes JPD, the new Feedback App, and Rovo - and there might be other apps later down the line. Rovo enables these next-gen use cases we’re obsessing over atm like agentic roadmapping.

It positions us head-to-head with the roadmap-plus-feedback players in the category. The Feedback App brings the qualitative side - what people actually say about your product. JPD brings the roadmap. Rovo brings AI agents on top.

Imo the other piece that needs to be in there longer-term is product analytics - the quantitative side, what people are doing with the product. You need both to make better product decisions. For now we're integrating with Pendo and other product analytics tools.

8:48 AM ✓

So how is the new Feedback app priced?

8:49 PM ✓

We're experimenting a lot because AI pricing is a totally new frontier. How much does it cost us in tokens to serve a customer? What's their willingness to pay for the value those tokens bring?

But we’re excited because it might end up being Atlassian’s first swing at usage-based pricing.

At Cycle, we removed seats and went 100% usage-based, with AI credits as the unit. The key insight is the value chain - the cost to us in tokens, the price we charge, and the actual value the customer gets (PM time saved, faster decisions, better prioritisation). When that chain has 100x between cost and value, you have a lot of room in the middle for healthy margin while delivering incredible ROI to the customer.

When we made the switch at Cycle, engagement and ACV really skyrocketed.

8:56 AM

The other key thing - the Feedback App isn't a copilot, it's an autopilot. So tokens are predictable.

If a team has 3 million feedback items per year, that's 3 million credits minimum. Every piece gets processed by AI, extracts insights, links them to ideas. You can predict revenue fairly well when it's autopilot.

With copilot, you can't estimate how many times people are going to ask questions to your AI. Autopilot is event-based, not user-input each time.

8:59 AM ✓

What about distribution? JPD is past 20,000+ customers off the back of cross-flow into Jira and the wider Atlassian ecosystem.

What’s been your first taste of that playbook as a founder coming in from the outside?

9:00 PM ✓

I have crazy anecdote on this.

We did a little Pendo form. “Hey, we're building the Feedback App. Who's interested in joining the early access program?”

We built it in 15 minutes. Sent it just to JPD users, not even all of Jira.

In a few days, we had 7,000 teams asking to talk to us.

For a few weeks, we ran 100 customer interviews a week.

This is my colleague’s actual calendar from those weeks - back-to-back Early Adopter Program meetings!

9:06 AM ✓

As a startup, we'd been struggling to get 5 or 6 interviews per week.

We joked with the team: why are people complaining so much about distribution? Just send a Pendo form, get 7,000 signups. Easy.

9:08 AM ✓

<Quietly devastating to anyone in startup land trying to book user interviews 😂>

Let’s go contrarian for a second.

Everyone's talking about AI-native product discovery, the solo PM-builder, the one-person billion-dollar company. What's your take on all this as someone who’s building for builders?

9:09 AM ✓

I actually think there's a broad misconception. What folks are building is mostly prototypes, not working software. We don't hear enough about the difference between a localhost demo and software that actually ships - every edge case accounted for, multiple users, concurrent sources of truth.

What’s changed most is the ability to prototype fast. PMs can test 10 ideas before landing on the one that gets shipped. But the separation of roles will remain. Designers, PMs, engineers all do more, especially on prototyping. The handoff persists at scale.

9:15 AM ✓

Last big one. Anything you'd do differently?

9:16 AM ✓

Post-acquisition, we shut down the product. That was a good decision. The right call, for sure.

But that meant we also shut down all the AI pipelines that were running. We'd had a continuous R&D engine - what works, what doesn't - and we shut everything down.

That created a gap of several months without AI R&D. In this world, several months is a lot. Then there's inertia in restarting.

I'd have kept at least 10 customers. The cutting-edge teams really pushing AI feedback workflows to the limit.

There's a concept I really like: the future already exists, it's just not evenly distributed. We had a couple of teams that were representative of that. They're building the future - they're doing what Atlassian customers will do in a few years. We should have found a way to keep them, keep the AI pipelines running, and not break the momentum.

9:22 AM ✓

That's a great one to land on. Mehdi - thanks for doing this mate!

9:23 AM ✓

Pleasure. Ping me anytime if you need more!

And for the product teams reading this - check out our new Feedback App and Product Collection here :)

9:24 AM ✓