How to work better with AI on your team

This insight is brought to you by… Science Says Platform

The Science Says newsletter keeps you up to date, and I hope you’re loving it 💜

If you’re a pro, we built the Science Says Platform to be the place you turn to before any project to check “What is the latest evidence I should know about, so I get the best possible results?”

 You’ll find all our 300+ insights, like:

• Discount coupons: when to give % off vs $ amount off

• When to use vouchers vs free gifts for promotions

• When AI chatbots boost vs hurt sales

Only paid Platform members can access them all 24/7, along with the exclusive playbook of brand names.

Want to sponsor Science Says to get in front of 30,000+ subscribers? Here’s all you need to know.

Topic: AI | productivity
For: B2C and B2B
Research date: June 2026
Universities: Harvard University, ESSEC Business School, University of Warwick, University of Pennsylvania

📈 Recommendation

When working with AI and a team to solve a problem or come up with ideas (e.g. for a campaign), follow these guidelines to work productively:

• Give the same AI to a small team (e.g. each one has an individual account, or everyone uses the same shared one), then have them work together (e.g. discuss together product ideas while using the AI, generate UI variations)

• Team up people that have different skills, but related to the same task (e.g. UX designer, researcher)

• Make sure the AI is used to complement people’s skills they are lacking (e.g. a creative director and a graphic designer using AI for data analysis)

Your team will produce better results (e.g. high-quality innovations, better campaign concepts).

🎓 Findings

• Scientists analyzed how people perform when they work with an AI compared to alone and in a team. To do so, they had human experts on the task domain rate their work. They found that people:

  • Produced the lowest quality solutions (e.g. in problem-solving a product issue) when working alone (vs with AI or a team)

  • Performed 6.3% better (vs alone) when working with a colleague

  • Performed 9.6% better (vs alone) when working with AI

  • When they worked with AI, and a colleague, they performed 10.2% better (vs alone), and were roughly 3x more likely to be judged top performers by experts in that domain

• The effect disappears when the people in the team have the same skills (e.g. 3 senior salespeople).

🧠 Why it works

• When working in a team, people complement each other’s perspectives and expertise.

• The same happens when working with AI, as we can converse with it as we’d do with humans.

• Human-AI collaboration can outperform human-human collaboration when AI fills the technical gaps.

✋ Limitations

• The study looked at the gains on people inexperienced with AI. It’s likely that gains would be higher for more experienced people.

• The study looked at teams of 2 humans working with or without AI, we don’t know how the effect is when the team is 3 people or more.

• Junior professionals might see higher gains when working with more senior people than with AI, as they likely lack experience to judge the output of human-AI collaboration.

• AI tries to conform to the users’ thoughts. In some tasks that require more creative thinking (e.g. designing campaign creatives ), using AI might provide lower gains.

👀 Real-life example

Optimizely offers AI “teammates” for marketing use. They’re supposed to act as expert colleagues.

• Their positioning can backfire. They should avoid calling their AI a “teammate” as that could make real human workers feel less accountable and prone to not checking mistakes.

• Presenting the product as “perfect” as they’re doing now could make people less likely to use it. They should clarify its limitations to make people trust and use it more.

• For more insights on this, check out our latest blueprint in collaboration with Wharton on AI agent adoption.

📖 Research

Fabrizio, D., Ayoubi, C., Lifshitz, H., Sadun, R., Mollick, E., Mollick, L., Han, Y., Goldman, J., Nair., H., Taub, S., Lakhani, K., R., The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork. Organization Science (June 2026).

Remember: This is a new scientific discovery. In the future it will probably be better understood and could even be proven wrong (that’s how science works). It may also not be generalizable to your situation. If it’s a risky change, always test it on a small scale before rolling it out widely.

🔬 Some of our other work

This week I’m in the Bay Area meeting teams at OpenAI, my ex-colleagues at Google, Microsoft AI, and a few AI startups for our latest research project with Wharton on “Cognitive Surrender” when working with AI.

As we work on this and many other projects we’re running behind the scenes, I want to start sharing more of these insights with you here as we come across them.

Expect more cool stuff in this part of the newsletter, and some more special insights to come up soon!

Here is how else Science Says can help you:

• 📈 Join the Science Says Platform to unlock all 300+ insights and exclusive playbooks

• 📘 Boost your results with Science-based Playbooks (for Pricing, Ad Creatives, Ecommerce, or SaaS)

• 🔬 Get science-based content for your brand. Publish groundbreaking insights about AI, marketing, or management that your customers will love.

🎓 It took our team of PhDs and me 13.5 hours to accurately turn this research into this 3-min insight. 

If this was forwarded by a friend, you can subscribe below for $0 👇