Public Transportation Company

How to discover 70 ideas for AI

AI Ideation Workshop

In-depth example

A large public transportation company managed to discover 70 ideas for Artificial Intelligence experiments. All within a 4-hour workshop with 11 people. At the end of the workshop, they picked the best three and defined the experiments in greater detail with the AI Experiment Canvas. From thereon, they can start their AI experiments today. In this example, We’ll show you how to discover and define your own AI experiments.

The top priority of this company is to make their travelers’ trips as pleasant and sustainable as possible. During operation, they collect a lot of data, which provides an excellent opportunity to improve their business goals with AI.

After a short introduction of AI and its use cases, we started by stating the business goals. It’s crucial to align experiments with business goals to get them funded. When an experiment doesn’t impact one or more business goals, it’s hard to make it valuable.

The first exercise aimed to set the attendees’ brains into creativity mode. Everyone was making associations between unrelated words. After a couple of rounds, everyone was in a creative mood. So, we jumped to the next exercise, where everyone made associations between the business goals and the use cases. These use cases consisted, among others, of computer vision, speech, and recommendations and resulted in approximately 50 ideas for AI experiments.

A good start! But to stretch the attendees, we made a list of external data sources, including the weather, social media, and football results (which could be an indication for raised risk of aggression in trains). A new round of associations between goals, use cases, and these data sources resulted in 20 more ideas.

From 70 ideas to 3 experiments

Segmenting on business impact, value, and feasibility, the attendees voted for three experiments. In small teams, they defined these experiments in greater detail. We designed the AI Experiment Canvas to define, plan, execute, and evaluate AI experiments. It helps to state a learning goal and set a smart hypothesis. With defining the metrics to measure the hypothesis, each team had to sharpen their hypothesis.

Start experimenting today

So, within 4 hours, this public transportation company discovered 70 ideas, resulting in three defined AI experiments. We aligned the experiments with the business goals with 11 attendees from different backgrounds (business, IT, data scientist). The experiments are well defined, so the attendees can start experimenting today.

Got interested? Get in touch.