In the second part of our human-centric discovery process, we establish the framework for participatory experimentation culture. This article describes a process spanning approximately two to six months, during which we progress from exploring AI usage to adopting it as a working companion.
It is essential to create an environment with a trusting atmosphere for experimentation. While skill levels among experimenters naturally vary, this diversity represents a significant asset, as each insight helps others advance.
During my career, I’ve witnessed a remarkable shift. Today, uncertainty is permitted, and it’s entirely acceptable to acknowledge not knowing something. One can even suggest trying it out. This wasn’t always the case. Considerable resources have been wasted contemplating what someone else might think about an idea that occupies one’s mind.
Artificial intelligence enables numerous possibilities, and what it specifically means in your own environment becomes clear through experimentation. When working with AI, we speak of probable outcomes. Through experimentation and comparing experiences, the boundary values begin to emerge.
It’s advisable to start experiments with relatively small, everyday matters, making it easier to discern the lessons learned. In larger systems, AI may function as a black box, making it difficult to determine why a particular result occurred or what should be done to reproduce it.
Beginning with Research and Experimentation
We assemble a group of interested individuals who want to experiment with AI in their work. This pioneering group constitutes a crucial part—the people—in a trinity where the other components are data and environment.
A former colleague devised an excellent exercise for the beginning of the first experimental project workshop:
- Close your eyes and extend your hands forward.
- Imagine a bar in front of you and grasp it with both hands.
- Once you have a firm grip, lower the bar decisively.
In this new mental state, it’s good to start contemplating your own ideas and thoughts. What have you always wanted to try? Making the experiment is most important, and experiments lead to outcomes. It’s also crucial to allocate time for experimentation—a specific portion of working hours.
What have you always wanted to try?
Initially, it’s also important to establish a safe environment for experiments where one needn’t worry about experiments or sensitive materials spreading worldwide. Some common rules should be discussed, perhaps regarding which materials are suitable for experiments and where results are recorded (Miro’s recently launched Innovation Workspace appears particularly interesting).
Collective Reflections
In workshops guided by Teamit, human-centered service design methods are employed to discover participants’ perspectives on improving work efficiency. A good workshop has a trusting atmosphere, sufficient time, and assurance that one can speak up if desired. The participant group should be assembled from all organizational levels. It’s beneficial to enhance everyone’s visibility both upward, downward, and laterally.
Using AI applications is a fascinatingly interesting subject area. Their user experience is still in its infancy, and various practices continuously come and go. Currently, the most common interface is conversational text-based, which can be somewhat optimized, and a concise two-hour training session can be arranged for this purpose.
Leverage AI: record interviews, create transcriptions using tools like Whisper, anonymize, and transfer to a language model. Request analysis at a general level of topics, pain points, wishes, etc. discussed. Request results in tabular format.
A diverse participant group ensures the best outcome. At this stage, it’s important to listen to as many perspectives as possible.
Self-Learning and Experimentation
Self-study and experimentation are effective approaches when participants have accumulated some knowledge. It’s interesting and useful to utilize AI capabilities when the need arises.
It’s also important to present the results of your experiments to others. Experiments provide much more information than just the outcome, and the actual result itself is rarely the most significant aspect. Instead, the accumulated knowledge about what worked and what didn’t work at different stages in the experimenter’s own context is valuable. Presentations proceed with confidence in a familiar group.
Organically oriented personal experimentation is really fun! Goal-orientation can be gradually incorporated, leading to an examination of one’s work methods. Where is there tedious and repetitive work? Where should one be able to discern something essential from a large amount of information? Where would assistance be needed—perhaps in proofreading or understanding text structure? Workflow modeling is the next step forward.
Leverage AI: provide anonymized material to a language model and ask it to summarize what target groups and user segments emerge from the data.
Expanding Perspective Through Modeling
Modeling means documenting task performance step by step. These paths typically become chains containing dozens of stages. When examining these chains, opportunities to utilize AI assistants begin to emerge.
As the process advances and multiple paths are compared, intersections where different task steps meet begin to appear. These junction points are fascinating development targets and potential pilot project subjects (which will be discussed further in the next part of the series).
Perhaps many units are accumulating customer understanding—what if the content production process included a stage where an AI assistant understanding all units shared information while considering thousands of user feedbacks?
Leverage AI: transfer anonymized response material to a language model. Inquire whether the content contains references to your desired impact. Also, experiment with how the content would appear in different versions (in clear language, explained to a child, as a list…). Use the results as base material for your own content production.
Gathering Information
Soon after experiments begin, enthusiastic accounts of experimenters’ achievements can be expected. Results should be collected and grouped, perhaps in Miro. Visual presentation helps understand what’s on people’s minds, and development opportunities begin to emerge.
Different job roles and business areas typically have completely different needs. All information is necessary material that helps conceptualize technology possibilities in one’s own context.
Leverage AI: transfer material to a language model and request formatting so that each thought occupies its own table cell. Then copy the cells into Miro as notes and group them in the next team meeting. AI could also handle the grouping, but then people would miss experiencing the important phase of familiarizing themselves with the material.
A good way to capture collective thinking is through a simple probe. Read our article here on how to send a probe to gather information. In addition to lowering the bar, the threshold for recording one’s thoughts must also be lowered, and the probe’s anonymous, quickly completed form is a good solution for this.
Ideas and thoughts need to be captured as much as possible.
In group gatherings, reviewing the probe’s yield serves as a good input alongside sharing more concrete experiences. Sharing is essential to receive feedback as soon as possible. Some ideas also serve as building blocks for subsequent ideas. The more sharing, the more gaining.
In the next part of the series, we’ll examine what could be done with all the information collected thus far and how to proceed more purposefully with experiments. Also read the first part of the series if you haven’t yet!
Leverage AI: anonymize received feedback and transfer it to a language model. Ask what kinds of signals the language model sees in the material. What positive and negative mentions can be found?
What Can You Do Yourself?
- Establish a group of enthusiastic experimenters and allocate experimentation time for them
- Create a digital platform for recording experiments
- Agree on rules regarding which tools to use and what content can be used in experiments
- Have the group meet at least once a week to share experiences
- Publish experiment results in internal channels as help and inspiration for everyone
Let’s Meet!
We’d be happy to explain how we’ve launched inspiring processes. One hour is sufficient. Let’s connect!
Aleksi Manninen
aleksi.manninen@teamit.fi
