Product teams are shipping faster with AI, and their customers have noticed nothing, which says AI team productivity is being measured in the wrong place. That is the finding Atlassian leads with in its State of Product 2027 page, and it is worth taking seriously because Atlassian sells the software these teams use.
The page says product leads report that AI helps teams ship faster, that customers are not seeing value any sooner, and that gut instinct still overrides customer data. It cites 1,000 product professionals and gives no method. These are Atlassian’s own claims, and we have not seen the data behind them.
Key points
- Where did the AI team productivity gain go? Into more output with the same number of decisions behind it.
- Does the wider evidence point the same way? Loosely, and with caveats. Per a secondary summary of Atlassian’s State of Teams report, 89% of senior leaders said individuals were working faster with AI, only 6% could point to clear ROI, and 14% of teams had turned AI use into real value.
- What should a small firm measure instead? One thing a week, and it should come from the customer’s side.
- What would a weekly check look like? Ten minutes on Monday, with the people who did the work.
Where did the AI team productivity gain go?
Into more output with the same number of decisions behind it. When building gets cheap, the scarce thing is knowing what is worth building. Atlassian’s headline number says as much: 86% of product leads agree that deciding what to build matters more than ever.
Think of it as a pipe that got wider at one end, the same pattern we saw when business teams hand off vibe-coded tools faster than anyone checks them. A team that can draft three prototypes in an afternoon still has one person who decides which to keep. The queue moves to the decision, where it is invisible, so the dashboard shows busy people and flat results.
Does the wider evidence point the same way?
Loosely, and with caveats. Per a secondary summary of Atlassian’s State of Teams report, 89% of senior leaders said individuals were working faster with AI, only 6% could point to clear ROI, and 14% of teams had turned AI use into real value. That summary does not show the survey method or define value, so treat the numbers as signals.
An Atlassian product lead, quoted by diginomica, described organizations measuring tokens consumed and who uses the most AI. He called those vanity metrics. We agree, and would add that they are cheap to collect, which is why they spread.
What should a small firm measure instead?
One thing a week, and it should come from the customer’s side. Before any AI-assisted piece of work starts, write down the behaviour you expect to change: more renewals, fewer support calls, faster quotes accepted. After it ships, check if that moved.
That sounds obvious, and it is rarely done. Most small firms count what AI produced, such as drafts, tickets closed and lines written. Our look at slop grenades showed what that hides, which is the review time that lands on someone else. Output goes up, and the person who checks it gets slower.
Atlassian’s team lead Dr Molly Sands, per that summary, recommends what she calls AI working agreements, short team rules on how AI is used and what to avoid. We would keep them to a page. Who may use AI for what, who reviews it, and which results you will check.
What would a weekly check look like?
Ten minutes on Monday with the people who did the work. List the two or three things AI helped finish last week, name the customer behaviour each was meant to change, and check whether it did. Then drop or change the one that clearly did nothing. That keeps speed tied to customer value. The steps:
Checklist
- List the two or three things AI helped finish last week.
- For each, name the customer behaviour you expected to change.
- Check if it did. If you cannot measure it yet, say so.
- Drop or change the one that clearly did nothing.
It is the same discipline we suggested for counting cost per finished task, applied to value rather than spend.
The sceptic’s view is that a vendor wrote the report and that a vendor survey rarely finds that the tool does not work. Fair, and it cuts both ways. Atlassian does not sell AI output counters, and its finding that speed has not reached customers is not the one a sales deck would choose. We would still want an independent study with a method behind it.
Start the Monday list this week, before approving the next AI tool. For tools that help teams get to customers faster, see AI marketing tools for Canadian teams.
Frequently asked questions
Does AI make product teams faster?
Atlassian’s State of Product page says product leads report AI helps them ship faster. It also says customers are not seeing value any sooner. The survey of 1,000 product professionals is Atlassian’s own, with no method shown.
Why does faster output not mean more customer value?
Because the bottleneck moves from building to deciding. If a team builds more but chooses no better, customers see more changes without more benefit. Atlassian reports that gut instinct still overrides customer data.
How should a small business measure AI results?
Before AI-assisted work starts, write down the customer behaviour you expect to change, such as renewals or fewer support calls. After it ships, check if it moved, instead of counting drafts or tokens used.
Last fact-checked October 10, 2026. Sources are linked on the claims they support.
