1 Define the experiment
The precise details of an artificial intelligence experiment depend largely on the organization’s goal, which is why the first step focuses on defining these goals. With reference to the Info Support AI Experiment Canvas, we determine the preconditions for the experiment, such as: how will you know if the artificial intelligence experiment has been a success? Within what time frame will the goal be achieved?
We also determine what data are required to achieve the goal:
- What existing data are relevant for the goal?
- What additional data are needed?
- How will the additional data be collected?
2 Perform and interpret the experiment
In step 2, we perform the experiment together, so that knowledge transfer can take place directly. The emphasis is on collecting, analyzing and combining existing and new data. We then interpret the result: what story do the data tell? What patterns and connections have been discovered?
This step is made up of the following parts:
- data collection
- data analysis
- algorithm selection
- interpretation of the results
In this step, it is often a good idea to distance yourself a bit from your own organization. During this step you can therefore make use of the Innovation lab of Info Support and Startup Village at the Amsterdam Science park. At this location, we work together to perform the experiment.
3 Anchor the experiment
A valid hypothesis can be converted into a production system. In this final step, we translate all the data flows found into a software system, so that it becomes part of the organization’s internal processes. This gives you more insight into your organization and helps you be prepared for organizational issues in the future as well.
If you would like to know more about starting an artificial intelligence experiment, contact us.