How a 90-Year-Old Manufacturer Found the Way to Let Humans Focus on Creative Work Through AI
There is a company in Japan where 30,000 employees and 8,000 AI agents work side-by-side.
And it is not a cutting-edge IT startup.
It is Ricoh, the manufacturing giant founded in 1936, turning 90 this year.
I want people who think, "Our company is too old-fashioned for AI," to read this article.
That is because what Ricoh did was not the result of a group of geniuses. Rather, it was the opposite: they simply followed the "correct order."
By reading this, you will understand why your organization might feel like "we have been using ChatGPT for two years, but it is not showing results."
And starting tomorrow, you can flip the order in which you introduce AI.
A Company Where 30,000 Employees and 8,000 AI Agents Sit Together
First, the numbers.
Ricoh has about 78,000 employees worldwide, with 30,000 in Japan.
As of January 2026, they reportedly have 8,000 AI agents running in Japan alone.
8,000. It is a staggering number.
What is more, these 8,000 agents were created by fewer than 3,000 employees—only about 10% of the workforce.
This 10% is proactively building AI to handle their tasks.
It is important to remember that Ricoh is "the copier company."
Their foundation is in manufacturing. While digital services and AI now account for over half of their revenue, they were originally a traditional manufacturing firm where people drew blueprints by hand and passed them to the next person.
Such a traditional company is iterating on AI faster than many startups.
By the way, there are 45,000 companies in Japan over 100 years old—more than half of the world's centenarian companies.
This means being "old" is not an excuse; rather, old companies have the most untapped potential.
The Real Reason AI Shows No Effect After Two Years
Now for the main point.
Ricoh is seeing a massive surge in consultations like this:
"We put ChatGPT where we thought we needed it two years ago, but it is not showing results. It is not breaking through organizational silos."
Many people likely relate to this.
There are three main reasons.
1. The Productivity Problem
Japan's productivity ranks 29th out of 38 OECD countries (2024 data). It is near the bottom. Compared to the US, which leads the world in digital, it is about half.
Why? The cause is "individualized work styles."
No matter how much you invest in IT, if the way work is done does not change, productivity will not rise.
2. The Data Problem
It is said that 70-90% of data within a company is "unstructured data."
Unstructured data refers to the intuition, tips, and know-how tied to individuals—handwritten drawings or wisdom existing only in a veteran's head. Ricoh calls this "tacit knowledge."
If you ask an AI to help without organizing this first, the data you provide is "dirty."
Therefore, the AI cannot function correctly.
Interestingly, an AI reading documents might suddenly fail when it hits a table. Or there might be conflict because technical secrets should not be on the cloud and must stay on-premise.
In short, throwing AI into a task without building the foundation first will not work.
Step 1: Visualize Work First and Eliminate 20% Waste
So, where did Ricoh start?
It was not AI implementation.
First, they created "room to breathe" for employees.
Updates and new technologies cannot be used by people who have no time. So, Step 1 was to free up time.
Specifically, they visualized the work of 1,000 people across 115 sections company-wide.
They discovered something interesting.
During remote work in the pandemic, "check-in meetings" had increased significantly because managers didn't know what people were doing.
By looking at the data, management realized, "Oh, we don't need to do this task anymore." This eliminated 5-6% of the work.
Next, they found similar tasks being done separately by different organizations and consolidated them. More reduction.
Then, they standardized the remaining "truly necessary work." Once standardized, automation technology becomes effective.
By continuing this persistently for over a year, Ricoh achieved a 20% improvement in operational efficiency.
There is a lesson here for us to steal.
The Japanese style of "everyone picking up the ball and connecting" is a strength, but it also creates "work that doesn't need to be done."
People pick up balls that aren't theirs out of kindness, but the workload just keeps growing.
So, try visualizing your own work for one week.
Just doing that will reveal things like, "Wait, do I really need this meeting?"
Step 2: Everyone Starts Using Just "One" AI
Once they saw a path to 10% of that 20% improvement, Ricoh made the next move:
"Every employee uses AI for exactly one task."
The key here is that they didn't just dump it on everyone.
They first designed "guardrails" for safe AI use. Then, they shared education and success stories through workshops.
Because the foundation (Step 1: visualization and standardization) was there, the data given to the AI was clean. That is why the AI worked.
It is all about this sequence.
The result is the 8,000 agents mentioned earlier.
Here is one specific example.
Ricoh solves management challenges for clients. They talk to CEOs of major corporations, so preparation is critical.
Veteran employees used to spend 4-5 hours per company reading integrated reports and public info to form hypotheses.
By teaching that veteran's tacit knowledge to an AI, the AI agents now verify hypotheses and draft proposals themselves.
As a result, the veteran's time was reduced by 75%. What took 100 units of effort now takes 25.
And there is another benefit.
The veteran's intuition is now passed down to mid-level and junior staff. By working with AI, knowledge is transferred.
This isn't just for marketing. It is happening in back-office, SCM, and sales frontlines.
4-5 hours down to just over 1 hour. The task of "reading documents from scratch every time" in your company could likely be handled the same way.
Step 3: Reassign People to "Creative Work" with the Saved Time
This is the core objective.
Using the freed-up time for work only humans can do.
Ricoh has a meeting room like this:
Behind a large LED display, five AI agents are implemented.
As employees discuss, the AI transcribes, corrects the Japanese, understands the meaning, and structures the information.
This allows employees to focus entirely on discussion and ideation. Finally, they vote and make decisions. The AI even supports the facilitator.
And here is the amazing part.
For the upcoming medium-term management plan, about 10 executives discussed it in this room.
Normally, this would take about two months.
It was finished in four hours.
Two months to four hours.
This idea of "moving people to creative work" resonates with the words of management scholar Ken Kusunoki.
Work consists of "Work" and "Play."
"Work" is providing skills for compensation—tasks within fixed rules. AI is faster, more accurate, and never tires of this.
But "Play" is different. Like Shohei Ohtani, it is work where value is created through unique sense and judgment.
AI takes away the fixed tasks. What remains for humans is sense and judgment.
The better we use AI, the more sophisticated human work becomes.
To foster employee creativity, Ricoh has also run an accelerator program since 2019.
Employees and startups pitch new business ideas. They battle through 200 ideas to select 5-10. They have done this for seven years.
This is how they cultivate employee autonomy and creativity.
Your Company Can Do It Too, Provided You Don't Mess Up the Order
To summarize:
Ricoh's conclusion is simple.
"Do not put AI where you want to use it immediately."
Follow this order:
- Visualize work first to create time.
- Remove waste and consolidate similar tasks.
- Standardize.
- Only then, set up the environment to use AI correctly.
Because of this sequence, employees start moving. If you reverse it, it won't work.
There is one more vital lesson.
AI has two sides.
One is "turning a negative into zero" by removing painful tasks. Getting home early, eliminating drudgery. Everyone loves this immediately.
But that alone doesn't last.
The other is "moving from zero to plus"—where humans create new value. AI implementation only becomes real when this is designed.
The feeling that today is better than yesterday and that you are part of that progress is what moves people.
Takahiro Irisa of Ricoh said:
"If Ricoh could do it, other companies definitely can."
And also:
"I believe AI appeared for the sake of Japanese companies."
Because AI has already learned most open data. What remains is the data sleeping inside companies.
And Japan is the country that holds the most of that internal corporate data in the world.
In a 90-year-old veteran firm or in your company, there are treasures sleeping that no one has dug up yet.
So, for one week starting tomorrow:
Try visualizing your own work.
If you find yourself thinking, "Wait, maybe I don't need this task," that is your starting line.
The lack of results wasn't due to a lack of ability. The order was just backwards.
