"Use it as you like" is an abandonment of management. The art of "AI Onboarding" to make ChatGPT a force to be reckoned with [The Science of Team Building Vol. 50]

■ [50th Anniversary of the Series] To all our regular readers

"The Science of Team Building" has now reached its milestone 50th installment 🎉🎉🎉
To those who read it every week, and to those who give us likes and comments, thank you so much.
Your reactions are the driving force behind my writing.

Over the past 49 installments, we have studied the science of how "humans" can trust each other and achieve results. But now, a "new member who is not human" is about to join our team.
Yes, it is Generative AI (LLMs like ChatGPT and Gemini).

At this 50-installment milestone, this series is also getting an update.
Starting with this issue, we will deliver a new 10-part series titled "How to Build an AI-Native Team".
It is not about being used by AI, but about the art of welcoming AI as a "teammate."
Please stay with us until the end!

(Below is the main content)

"We've signed up for a corporate plan for ChatGPT, so everyone please use it to improve work efficiency."
"...Yes (how do I do that?)"

AI tools were introduced top-down by the president and management.
However, even after a few months, only a few "gadget lovers" are using them.
Most members logged in a few times at the beginning, but now they don't even touch them.

Doesn't this "AI ghost member" situation happen at your company too?

The IT department and managers must be holding their heads in their hands.
"They should be able to figure out how to use it by looking at the manual."
"Why won't everyone use something so convenient?"

The answer is simple.
It is because you think of AI as a "convenient tool" like Excel or Slack.

Generative AI is not a tool.
In the new series starting now, we will scientifically explain the "onboarding (acceptance) techniques" to make AI a force for your team.

■ AI is not a "tool" but a "new recruit"

Why doesn't the staff use AI? It is because of a gap in expectations.

Many people think of AI as a "vending machine" or "Google search." They think that if they press a button, the correct answer will come out.
However, when you actually put it into ChatGPT, it returns mediocre answers or lies without hesitation.
"What, it's useless," they say, and never log in again.

Please correct this perception.
Generative AI (LLM) is a "super-elite new graduate" from the University of Tokyo.

His (AI's) specs are as follows:

• Knowledge base: Almost infinite (the entire internet).

• Processing speed: 100 times faster than a human.

• Stamina: Can work 24/7.

• Personality: Honest, but waits for instructions.

• Drawback: There is no guarantee of a '100%' correct answer. It pretends to know things and lies (hallucination).

While the latest models have increased computational power, they still make mistakes with confidence from time to time.
What would you do if such a new hire were assigned to your team?
Would you just dump the work on them and say, 'Hey, get this done'? You wouldn't, right?
If you just dump the work on them, it's clear that you'll either get a completely misguided deliverable or a false report.

AI adoption fails because managers are 'neglecting' this new hire.
What is needed is not how to install it (tool explanation), but 'new hire training (onboarding)'.

■ Understand the 'Jagged Frontier'

There is a concept proposed by a research team at Harvard Business School and BCG called 'Jagged Frontier'.

AI capabilities are not uniform.

• Strengths: Ideation, summarization, translation, programming

• Weaknesses: Understanding complex human emotions, accurate searching of the latest news (depending on the model)

This frontier is 'jagged,' and things that are easy for humans are difficult for AI (and vice versa), which differs from intuition.
That is precisely why managers must first share this 'strengths and weaknesses list' with the team and define roles, saying, 'Let's leave this to him (AI)' and 'Let's have a human check this.'

■ You don't have to be able to use it yourself as a manager

'But I'm not good at using AI myself...'
To managers who feel that way, please rest assured.

You can manage engineers even if you can't program, right? It's the same thing.
You don't need to become a 'prompt engineer' (a master of AI operation).
What is needed is the ability to design 'how to incorporate the subordinate called AI into the team'.

'Did you try having the new hire, AI-kun, summarize these meeting minutes?' 'Try asking for AI-kun's opinion as a sounding board too.'

Just by doing that, having AI participate in meetings as a member of the team is enough.

■ First, forget about 'operational efficiency'

There is one thing you must never do in the first step of AI utilization.
That is to make 'operational efficiency' the goal.

The moment you say, 'Use this to reduce overtime' or 'Increase productivity,' AI becomes a 'quota' for the members.
If they are asked for results without even knowing how to use it, they are bound to end up hating it.

Set the goal for the first month to 'play around with it'.

• 'I tried letting AI decide where to have lunch today'

• 'I had AI listen to my complaints about my boss'

• "I'm sharing this because it gave me a really weird answer."

In this way, the first priority is to grasp the feeling (mental model) of "AI is surprisingly cute and silly," or "But sometimes it says amazing things." You wouldn't entrust heavy work to a new hire without a foundation of trust (rapport), would you? It's the same with AI.

■ [This issue's takeaway you can use starting tomorrow]

First, introduce this "elite new hire" to the entire team and give them their first task (training).
I have prepared an "AI Onboarding Sheet" to break the ice.
Please try doing this with everyone at your next regular meeting.

[🤖 Orientation for the AI's first day on the job]

1. Our team's AI (Naming rights)
First, please give him (ChatGPT, etc.) a name.
Name:[　　　　　　　] (e.g., Chiemaru, Robo-tan)
*Calling it by name instead of "AI" will shorten the psychological distance.

2. First task: Create a self-introduction
Instruction (Prompt):
"You are [Name decided above], a super talented but slightly clumsy new hire assigned to this [Department Name] team starting today.
Our team's goal is [　　　].
Please introduce yourself in 200 characters, including some humor, about how you can contribute from now on."

3. Promises within the team (Handshake)
□ What he says often contains lies. Humans must always fact-check.
□ He is "waiting for instructions." He won't move unless you ask him specifically.
□ You don't need to make things efficient. First, try talking to him once a day.

■ Let's welcome the "strongest partner"

AI is not an enemy that will steal your job.
If you master it, it will take on your team's chores, act as a sounding board, and become your "strongest partner" who works 24 hours a day without complaining.

Not "introduction," but "hiring."
Not "operation," but "education."

From the moment this mindset changes, your organization will begin to evolve into "AI-native." Come on, let's start by giving him a name.

■ Finally: Your "Like" is his training data

If you read this article and thought, "I'll try naming the AI" or "I'll welcome it as a new hire," please press the heart-shaped "Like" button.
Your "Likes" become positive feedback (learning data) for my writing algorithm!

Next time, Vol. 51, will be about "Giving Instructions (Prompts)." Giving instructions to AI is the exact same technique as delegating to subordinates.
I will explain management techniques that require no translation.
"Follow" and stay tuned!