Over the past few decades in the technology industry, some of the biggest constraints to building products have been about having enough engineers, time, or compute. For the first time, that era is ending.
Tech teams are experiencing a revolution unlike anything they’ve seen before. The barriers to entry for building have all but disappeared. The constraint no longer comes from producing enough output, but from deciding what to build, and the restraint required to build something good.
As a result, Product Managers and Designers can’t keep up with the productivity gains Engineers are making. So, the new superpower for Product Managers is discernment.
This is exactly the problem Atlassian has always been in the business of solving: how do teams decide what to build, align on priorities, and turn speed into outcomes rather than chaos?
In the first session of Atlassian’s new AI Talks series, Atlassian’s Chief Product and AI Officer, Tamar Yehoshua, interviewed Jon Noronha, Co-founder of Gamma, an AI-native design platform and presentation builder.
Jon runs a high-growth company of just over 70 people. He shared that they’re currently at a 1:4 ratio for PMs to Engineers, where the rule of thumb has been 1:10 for decades. This is just one example of how much this shift is already underway.
“If your engineers just got two or three times more productive, I don’t think the PMs did,” says Jon. “I think it’s because PMs deal with human constraints a lot more, and humans haven’t sped up to the same degree. I look at teams we have that are one PM and four engineers, and I do believe they’re getting done what a team of eight to ten engineers used to get done.”
The role of a ‘builder’ in an AI-native company
Across the technology industry, as companies adopt and build with AI, roles are starting to blur.
The PM’s job is drifting away from coordinating developers and towards the building itself. PMs need to remove the friction between the intention of what to build and the execution of it.
PMs have also started to write code. Half of Gamma’s PMs and Designers are now committing code, which brings with it a new array of challenges and opportunities.
“One of the biggest things our PMs use AI for is bug triage, actually,” Jon said. “A customer reports a bug, and PMs do the full investigation of what went wrong, and then decide if it’s worth looping in an engineer, or just make their own ticket.”
Where previously PMs would have had to consult engineers about how the code works, they can now ask AI and start reporting and escalating bugs and typos at speed. At Atlassian we’re starting to see similar patterns and are moving towards a model where everyone ships code, regardless of job title.
But as more and more bug escalations get assigned to engineers to fix, you suddenly need more engineers again. One of the biggest challenges in enterprises today is identifying where the bottleneck is as the pace of output increases.
Why human judgment still matters
In the new world of AI-assisted building, determining the right direction and feature updates for products is becoming increasingly subjective. When evaluating A/B tests, human judgment remains essential, as product updates are often creative and subjective choices where taste matters.
Jon shared an interesting example from Gamma: they tested whether giving users the option to preserve text entirely in a presentation was preferable to the product automatically making copy updates. Using A/B testing, they found that while users said they wanted their text left completely unchanged, they actually preferred it when the product made minor improvements to the copy in the right places.
“People were saying, preserve my exact text, but they didn’t mean preserve my exact text,” says Jon. “They meant ‘mostly preserve my text, but you gotta zhuzh it a little bit in the right places’. And so when we prompted too accurately, even though it got 100% on the evals, it actually totally failed that more subtle, ‘do what I mean, not what I say’ type test.”
A classic example of what people say they want does not always match what they actually want – and a clear demonstration of why experimentation and subjective evaluation still matter enormously.
As we build, it’s still important to sense-check that we are building what customers actually want, not just what we think they do. While parts of the process can be automated, creative human judgment remains irreplaceable.
To open up your product, or not?
There comes a time when building any tech product where you need to decide whether to open up your product ecosystem or keep it closed. Both Atlassian and Gamma have chosen to be open ecosystems, with APIs.
It’s a choice between an offensive or defensive approach. If you don’t build the integration, then someone else will – and your customers may go with them. Choosing to open things up is a growth strategy in itself, and it’s a sign of confidence in your products in a world where anyone can build anything.
“We can’t just build a wall around ourselves and hide in our castle anymore. Maybe once upon a time we could, but not anymore”, says Jon.
In fact, Atlassian recently announced our new partnership with Gamma. You can now assign an agent to Gamma in Jira or Confluence to automatically create a presentation.
How context is becoming one of the most important growth drivers
Tamar asked Jon how his role has changed from being a PM and founder pre-AI era to now leading an AI-native company. His answer was interesting – he feels that nowadays his role has become what he calls a ‘context wrangler’. Having been at the company the longest, one of his biggest challenges is taking the knowledge he has accumulated over the years of running the business and turning it into written documentation so that his team and AI agents have access to it.
“I’m increasingly viewing my job as context wrangler, not just as a leader on the team, but just someone who’s been there for a long time,” says Jon. “I feel like it’s my job to take all the information that is only in my head and get it into writing where my team has access to it, but also all of their agents have access to it.”
Tamar reflected that the keyboard may become less and less central in the future as people turn to voice prompts and AI-generated artifacts to develop context-rich documents. It raises an interesting challenge as businesses scale – how do you ensure that information is recorded and passed on so that learnings can be shared and the right business decisions made at scale?
As startups grow, documentation and processes become increasingly important. This is especially true in remote-first companies, with people working in multiple timezones and office locations that need to collaborate efficiently.
At Atlassian, our System of Work (Jira, Confluence, Loom, etc) helps us record and share information, surfacing key insights quickly through a combination of project management, documentation and AI tools.
Where to from here?
While AI capabilities are rapidly changing the way technology teams work, Tamar reminded us that sometimes we have to remind ourselves that the fundamentals of good business haven’t changed. It’s important to stay close to customers, stay lean as your business grows, and continue running experiments to learn and improve. So while the way we build has changed, the core principles of building a product people love haven’t.
The question to ask yourself as a business leader right now is this:
Where is your biggest bottleneck to growth? Is it in shipping code, deciding what to build, or gaining alignment on priorities?
