Brainstorming at the start of a project is hard. Who hasn’t stared at a blank document, wondering how to begin? Reaching for AI can feel like an easy way to get started, especially at first.

But at Atlassian, we believe that AI should amplify human judgment, not replace it. The Teamwork Lab wanted to investigate when it’s most effective to introduce AI into the brainstorming process. Does the timing of AI introduction affect a team’s creativity? How does brainstorming with AI affect employees’ sense of ownership over the work they produce?

The brainstorming challenge we gave our participants for the first experiment is called the Alternative Uses Task (AUT), one of the most widely used measures of creative thinking in psychology research. Participants are asked to come up with as many unusual and creative uses as they can for a common everyday object (in this case, a paperclip). Participants tackled the challenge in three phases, with each phase under one of three conditions:

1. Brainstorming on their own, without using AI
2. Using AI to come up with new ideas
3. Reviewing a list of a colleague’s ideas before brainstorming solo

After each round, we measured how creative the results were on each round. We also asked the participants to estimate how creative they thought they were on each round.

In the second experiment, we looked at how this effect might show up in real-world organizations, with a real-world task. The Teamwork Lab asked 15 Atlassian HR business partners (HRBPs) to participate in two 20-minute brainstorming sessions related to their business — one on talent retention, one on operating model design. Some of the participants brainstormed without AI for the first half of their session, while the rest used AI to brainstorm from the start. Both groups used AI in the last half of their session, to polish their final list of ideas.

Finding #1: Starting with AI gives you an early edge in speed and breadth—at least at first

In both studies, using AI gave brainstormers a boost.

In the paperclip study, participants using AI generated up to 43% more ideas in their first round of brainstorming compared to the control group who did not use AI. That group also received higher overall scores for creativity compared to the control.

Similarly, in the Atlassian HRBP workshop, those brainstorming with AI moved faster initially — their ideas were 36% more complete and covered 45% more topics in the first half of their session, compared to the group that did not use AI at first.

But while brainstorming with AI gave participants an early lead in idea variety and depth at the 10-minute mark, it was only a temporary head start. By minute 20, once both groups polished their work with AI, both groups scored roughly equivalent for creativity.

Finding #2: But when AI does the early lifting, workers feel less connected to the outcome

Ten of 12 participants in the HRBP workshop said that using AI first made it feel easier to get started. But they also noted that speed came at an emotional cost.

When asked how each approach felt:

• Every single participant (12/12) felt more ownership over human-first ideas.
• A strong majority (9/12) found the human-first session more rewarding.
• Two-thirds (8/12) felt human-first unlocked higher creativity.

Finding #3: If you want true originality, look to your team before your tools.

In the paper clip study, peak creativity didn’t come from prompting an AI tool – it came from cross-pollinating with teammates. Workers who reviewed a colleague’s ideas before brainstorming solo (with no AI) outperformed everyone, scoring 18% higher in originality than those who worked entirely solo and 11% higher than those who used AI directly.”

Why the feeling of productivity is misleading—and where AI actually belongs

Generative AI makes brainstorming feel effortless, cutting frustration scores in half compared to working solo. But that sudden volume of ideas creates a deceptive illusion of creative momentum. When asked to identify their most creative round, participants overall were barely better than a random guess – and AI users were the worst at judging their own output, guessing correctly just 32% of the time compared to 40% for those working solo.

The takeaway isn’t that AI has no place in brainstorming. It’s that the creative value of AI lives in the social exchange around the output.

How teams should rethink AI brainstorming:

• Don’t outsource the initial thinking: When individuals or teams jump straight to AI generation, they trade away a sense of ownership and pride for speed.
• Stop evaluating workflows by “how productive it felt”: Because AI slashes the cognitive friction of getting started, people consistently overestimate how creative the session actually was. Teams need to evaluate AI workflows by the novelty and viability of the downstream work, not the volume of text generated in the first five minutes.
• Use AI as a shared catalyst, not a solo shortcut: The highest-performing ideators were those who built upon a curated list of seed ideas shared by a teammate. Curating someone else’s list primed deeper independent thinking than prompting an LLM from scratch.

The bottom line: If you want higher volume on routine tasks, prompt an LLM. But if you want original, high-stakes thinking, don’t prompt from a blank slate. Give your team space to struggle with the problem first, cross-pollinate seed ideas across teammates, and bring in AI to expand and stress-test what you’ve already built. To map out exactly when your team should go solo, collaborate, or bring in AI, explore our AI Collaboration framework tool.

---

Methodology

In April 2026, the Teamwork Lab conducted an experiment with 797 knowledge workers. Each worker completed three brainstorming rounds across several separate conditions with AI and solo brainstorming. After each round, the lab measured how creative their results were and participants also guessed how creative they were on each round. In May 2026, the Teamwork Lab also conducted a separate, converging experiment with 15 Atlassian HRBPs tasked with brainstorming for two 20-minute rounds across two separate conditions: one with using AI throughout, and one with human-first brainstorming before using AI to polish. The HRBPs were also asked to fill out a survey after the session.