Using GenAI to prototype personal cultures

Could an AI assistant help build a dialogue about creating a personal culture, which suits the wide variety of needs and desires that people may have? This is part 1 of a series of prototypes.

We heard through our Dayshift studies that people need personalised ways of working, particularly when work takes place from a location that differs from the head office. The once unifying culture of a company now brushes up against hundreds if not thousands of personalised cultures that fan out into the larger organisation.

We’ve been experimenting with generative pre-trained transformers (GPT’s) to help us understand if an assistant could help people build their own personal culture of work. Could a GPT be trained to help individuals give themselves permission, through a personal dialogue, to explore the millions of different cultures that could be possible. Could it eventually turn into a culture discovery coach, helping people find and build their best work-life. Here’s our discoveries so far.

Using OpenAI’s GPT-4, we first noticed there’s a lot of existing GPT’s such as “remote work guru”, “remote work communication tips”, or Work from Home Wizard”. We tried them all, and they’re great but this isn’t it.

Generating tips for productivity success or achieving a work-life balance isn’t what we hear people need. Through our research, we consistently heard people speak about the need for building a personal dialogue and finding their own way. So we made our own GPT.

We started with a prototype about Dayshift Culture Crafting. Crafting because it needs to be all about crafting a personal work culture. Creating something new that feels personal. The introduction asked for a symbol, and we made this little ray of sunshine. A symbol that represents the act of creating something positively new and hopeful.

We first started with one simple button prompt: Can you help me create breaks in today’s schedule? It temporarily changed into a break generating machine. This wasn’t it either.

We then realised that we can train the model for how we’d like it to interact with us. Taking into account everything we’ve seen, played around with and got annoyed with, these were our instructions:

This worked beautifully. It became a discussion that could build over time, with just the right amount of encouragement by asking the right questions. It was on the limit of being slightly pressing by asking follow-up questions, but we then realised that it’s just testing our comfort levels.

Training the model to create a discussion about creating a personal culture of work was a prototype of its own.

As one of my friends recently wrote “It doesn’t have a will or goal, but rather it’s like an un-emotional person reading the conversation as a third-party and literally filling in “based on the discussion so far, what might the person chatting with us say”. It’s essentially guessing the answer through the invisible person you’re talking to.”

In part 2, we’ll show how we used the Dayshift Culture Crafting GPT to create prototypes which helped guide the formation of a personal culture of work. Starting with the simplicity of a day.