AI is accelerating product development. But what happens to decisions, customer feedback and collaboration? A new Atlassian report offers some remarkable answers.
In my previous LoopsWork articles, I put forward two hypotheses: AI amplifies existing organisational patterns. And when teams ship faster than they learn from the results, they accumulate Feedback Debt.
Those ideas were primarily observations and reflections. Atlassian has now published its State of Product 2027. Its survey of 1,000 product professionals provides interesting figures – and some findings that go beyond my earlier articles.
More output does not automatically mean more impact
According to the report, 99% of the product teams surveyed have firmly integrated AI into at least one shared workflow. And 96% consider AI equal to or better than human work in at least one area of responsibility.
So the discussion is increasingly less about whether AI is being used in product development. The more interesting question is what it actually changes.
80% say AI helps them ship faster, without customers seeing value correspondingly sooner. 69% say their decision-making processes cannot keep up with the increased pace of delivery.
That captures almost exactly the tension that interests me about LoopsWork: production gets faster, but the surrounding system does not automatically become more effective. Remove a bottleneck in development, and it may simply reappear in prioritisation, alignment or impact measurement.
Feedback exists. It just does not drive decisions.
The figures on customer feedback are particularly revealing:
- 77% of teams formally document feedback.
- 64% do not consistently use it to make decisions.
- Only 4% can trace every roadmap decision back to customer feedback or other evidence.
- 77% experience leaders' gut feelings overriding data-driven product decisions at least occasionally.
For me, this is the clearest connection to the second article on Feedback Debt. The fundamental problem does not seem to be a lack of feedback. It is much harder to incorporate that feedback into the flow of work so that it actually changes a decision.
Also interesting: 43% cite a lack of standardised processes for using feedback as an obstacle, while 41% point to feedback being scattered across too many tools and channels.
More data alone is unlikely to solve this.
Roles are blurring – collaboration is not automatically improving
The study also shows a positive development: 89% of teams now use AI to perform tasks that previously belonged to other functions. Product managers build prototypes, designers deliver functional solutions, and developers can contribute to product strategy earlier.
Yet another figure stands out: in 82% of teams, developers are not involved from the very beginning of ideation.
That is an important distinction. Just because everyone can do more things themselves does not mean the right people talk to one another earlier.
Professional boundaries are becoming more permeable. Good collaboration still does not happen at the press of a button.
More time. But for what?
The report is by no means purely a warning. Thanks to AI, 53% of respondents have more time for strategy and roadmap development. And 45% can devote more time to product discovery and validation.
That is real progress – at least from the respondents' perspective.
AI frees up time for work where context, judgement and exchange matter. Whether organisations actually use this time to make better decisions is the next question.
What this means for LoopsWork
To me, the findings support two ideas from my previous articles: AI amplifies the effects of existing ways of working, and faster delivery alone does not prevent a backlog of learning from feedback.
That is precisely why the LoopsWork thought model looks beyond implementation in the Build Loop to impact in the Value Loop and changes to the way we work in the Learning Loop. This connection remains essential even if AI dramatically accelerates individual steps.
What the study can – and cannot – tell us
One important caveat: the study is based on self-reports from 1,000 people at companies with 500 to 10,000 employees in the US, Germany and France (surveyed in August 2026). It proves neither cause and effect nor the effectiveness of LoopsWork. And, as a vendor of product management software, Atlassian naturally has its own interest in the subject.
Nevertheless, the findings are a strong signal: the next big opportunity is not just to build faster, but to make better decisions, verify outcomes and learn.
Perhaps this is the more interesting AI question for 2027: not “How much more can we build, faster?” but “What is actually worth building?”
Source: Atlassian – State of Product 2027, published in 2026, based on a survey of 1,000 product professionals.