We first met when I agreed to be a "buddy" for a cohort of interns to Nesta. I walked this group of able, enthusiastic people around the teams here and asked them to introduce themselves. It was fun listening to them explain who they were and why they were here and how the story sometimes changed according to the responses of the people they were meeting.

Afterwards, Michael hung around. He said that he had some ideas for creating connections across the organisation. I liked the thought of this and asked him to email me with some practical ideas. We met and had coffee.

All of his ideas were good, but I loved the idea of randomly generating coffee meetings. It met (subconscious) criteria of not being too intrusive to anyone's day-job (once a week for 20-30 minutes), non-hierarchical (no-one could direct it) and democratic.

We called it Randomised Coffee Trial (RCT) because Nesta has designed and run Randomised Control Experiments (RCTs) in areas that demand evidenced-based policy making. In a way, our RCT was designed to remove the control, and insert some serendipity.

Over the next week, Michael went off and set it all up, including a write-up of a conceptual framework based on the work of Pedro Medina and others.

We used email, a simple webpage and Excel to randomise the matches. I guess that what I brought to the project at this stage was some institutional permission. For example, I felt ok sending an all-staff email telling people about RCT. As an intern looking to secure employment here, Michael might have found this too risky to do himself.

Within a week, we had signed up over half the organisation, and people were meeting each other once a week and finding out about others and their projects. I had a delightful debate (verging on a row) with the first person I met for coffee and actually this formed a really strong professional bond with someone who I wouldn't otherwise know.

This all happened when I was extremely busy at work. Michael and I must have spent no more than 5 minutes talking about RCT at each decision point, often in corridors, and we pivoted our plans very rapidly. This proved to be a bonus - we didn't have time to decide something was stupid and liked the idea of just doing something and seeing how it turned out.

I would urge other organisations to try RCT, but if you'd like to design your own organisational experiment...

• Ask the people taking part to help design/iterate it
• Try things that go with the flow of people's routines or which aren't too intrusive
• Actively give people permission to try new things
• Build in serendipity - it's delightful