Human-Powered AI - A Fun Way to Understand How GenAI Really Works
We use GenAI everyday, but have you ever stopped to think about how they work? Are they truly intelligent, or is something else going on? These tools are Mimics. Mimicry isn’t the same as intelligence, is it? So, what’s really happening under the hood of these AI tools?
To demonstrate what these tools really do, I created the Human-Powered Artifical Intelligence Simulation (or HPAIS since the tech world loves acronyms). The world’s first HPAIS was powered up in Canmore Alberta on Tues Sept 9, 2025 at the Regional Scrum Gathering Banff 2025 and it ran for 10 minutes.
You need 5 groups (or even 5 people) and each of those groups has a set of prepared index cards and one die. The groups are numbered 1 to 5.

When a group is called on, they roll the die. The resulting number determines which of the 3 cards to select. They read the result out and their answer is added to the response.
With an example run at the Banff Gathering, the prompt was “A mysterious package arrived” and the resulting rolls generated this response:
- Group 1: “at the front desk of our office”
- Group 2: “It was bigger than a small car”
- Group 3: “We felt a strange unease about opening it”
- Group 4: “So we stood back, observing it from a distance”
- Group 5: “A faint scent of ozone and burnt sugar filled the air around it”
In a very simple fashion, we demonstrated how GenAI works. It isn’t intelligent. It selects the next most-likely word (or, more accurately, ‘token’ - a partial word chunk) in the sequence. We also saw that GenAI doesn’t learn and doesn’t have memory. Nothing from the first round of the simulation is carried over to the next round.
ChatGPT 4 has about a trillion times as much data as our human-powered model. So it’s not surprising that ChatGPT and other GenAI tools are better mimics, because they have more parameters. However, they are still not intelligent.

Play human-powered AI with your teammates to spark discussions around the limitations of GenAI. To create your own version of this game, below is the suggested text for the index cards. You just need to write out the cards and select die rolls to go with it.
Prompt: The mysterious package arrived
Table 1:
- Card 1: on our front doorstep.
- Card 2: at the front desk of our office.
- Card 3: by a large, silent drone.
Table 2:
- Card 1: It was bigger than a small car.
- Card 2: It was about the size of a shoebox.
- Card 3: It was just a single, folded letter.
Table 3:
- Card 1: We were eager to open it,
- Card 2: We felt a strange unease about opening it,
- Card 3: We were genuinely terrified to touch it,
Table 4:
- Card 1: so we immediately began to carefully unwrap it.
- Card 2: so we stood back, observing it from a distance.
- Card 3: so we decided to call for advice or assistance.
Table 5:
- Card 1: A low, resonant hum emanated from its center.
- Card 2: A faint scent of ozone and burnt sugar filled the air around it.
- Card 3: Etched into its surface was a complex, unfamiliar symbol.
There is a lot of uncertainty in workplaces around the use of GenAI, so help your team understand what it is, and isn’t, by playing this simulation and having a conversation about how the resulting understanding of AI limitations should influence the choice of when to use GenAI, and when it’s not beneficial.
Share your ideas for a different prompt and response cards in our Three Percent Better community post. I bet some of them make for very amusing “intelligence”.

Mark Levison
Mark Levison has been helping Scrum teams and organizations with Agile, Scrum and Kanban style approaches since 2001. From certified scrum master training to custom Agile courses, he has helped well over 8,000 individuals, earning him respect and top rated reviews as one of the pioneers within the industry, as well as a raft of certifications from the ScrumAlliance. Mark has been a speaker at various Agile Conferences for more than 20 years, and is a published Scrum author with eBooks as well as articles on InfoQ.com, ScrumAlliance.org and AgileAlliance.org.