[Must-read for Professional Firms] The Essence of AI Implementation Lies Only in the Gritty Details
Hello, I am
Yuki Fukiage, CEO of Lean Stack Inc..
Introduction
Today, I am going to talk about something a bit edgy.
AI service providers have increased, haven't they?
At an incredible pace.
I am one of them, so I don't feel great saying this, haha.
"You can take AI training using subsidies"
"We support AI implementation with DX promotion grants"
Services with these kinds of slogans have exploded in the last year or two.
That in itself is not a bad thing.
Broadening the base of AI is a plus for the industry.
But, let me be honest.
How many companies have truly changed their operations after taking those training courses?
In my experience, very few.
You could almost say none.
In fact, I hear quite a lot of people saying, "I took the training because it was subsidized, but it didn't help me at all..."
This time, I will write about why "I took AI training but nothing changed."
And, I will write the honest truth that people in the industry would never say about where the essence of AI implementation really lies.
Firms that stop at "I took AI training"
Professional firms that took AI training just because it was subsidized.
They have increased quite a bit.
The training itself often has a high satisfaction rate.
"I learned how to use ChatGPT"
"I was taught tips for prompts"
"I was surprised at what it could do"
The post-training survey is marked as "Very Satisfied."
A training satisfaction score of 4.8.
That is a wonderful number.
But you know, if you go to see that firm a month later, nothing has changed.
You know how to use ChatGPT and Gemini.
You have also learned prompt techniques.
But payroll is still done manually.
Grant application checks are still done by hand.
The workflow for creating work rules hasn't changed either.
"I learned how to use AI in training. But I don't know how to apply it to my actual work..."
I hear this voice so often...
This isn't the fault of those who received the training.
The instructors have made teaching the 'functions' of AI their goal.
The latest AI features are constantly being updated. When a new feature comes out, they introduce it. They end it with, 'Isn't that amazing?'
But knowing the functions and changing the work are completely different things.
They just convey fragmented—or rather, temporary—latest feature information and say, 'Okay, that's it.'
That is not the essence.
It is nothing more than self-satisfaction.
Firms that have grittily integrated AI
On the other hand, there are firms that are quietly and grittily integrating AI into their operations.
There is zero glamour.
What they are doing is something like this.
For payroll, they first document the rules of each company one by one.
'This company has a 20th-day closing.'
'This allowance is paid under these conditions.'
'The accounting person here won't accept it unless it's in this format.'
They have all these rules, which were only in the heads of veteran staff, spilled out and written down.
They organize it into a format that AI can read and feed it to the system.
Then, they check the output.
'This part is wrong.'
'This condition is missing.'
'This exception is not being handled.'
Go through the PDCA cycle one step at a time.
When creating work rules, first prepare a template.
Create a hearing sheet and accumulate data for the final version.
For each industry, there are "clauses that must be included."
Codify them.
For irregular cases, set rules for those specific irregular cases in the AI.
It's plain, isn't it?
It's nothing but gritty work.
However, firms that have done this are seeing results.
Subsidy check tasks are now almost entirely completed by AI.
The man-hours for 120 Career Up Subsidy applications have been halved.
Payroll processing time has been reduced to one-fifth.
Human error has also decreased drastically.
If the initial output accuracy was 60%,
change it to 61%. Change it to 62%.
It's nothing more than this accumulation.
A series of plain and tedious tasks that are never discussed in flashy AI seminars.
But this is the real deal.
What creates this difference is just one thing.
The difference between firms that stop at "we took AI training" and firms that have grittily integrated AI into their operations.
What creates this difference is just one thing.
Whether you "learned how to use AI" or "transformed your business with AI."
Teaching how to use AI is a means.
It is not the goal.
The goal is for operations to be improved.
To put it another way, it's to increase the number of cases you can handle and raise the ceiling on your revenue.
However, many AI training programs end after just teaching the tools.
"We achieved a 4.8 satisfaction rating"—but did that actually improve your operations?
If you can specifically say, "This task improved by this much," that is incredibly persuasive.
But saying "the participants are satisfied" proves nothing. I know that sounds extremely harsh.
It is natural for participants to say it was "good."
Because they gain knowledge. Learning new things is fun.
But watching YouTube videos is enough for that.
The issue is whether that knowledge is being translated into actual work.
Training that misses that point is just an event to burn through subsidies.
It is harsh, but that is the truth.
"Field-oriented approach" is the only standard for distinguishing the real thing
So, what is the real thing?
I use the term "field-oriented approach." I believe the Ministry of Economy, Trade and Industry also used this as one of their philosophies.
Setting that aside,
Setting that aside,
On a practical basis, specifically, which tasks are improved and how?
Can you commit to that?
Teaching the functions of AI itself is, of course, necessary.
Because they cannot use it if you do not teach them.
But that is not the main goal.
It is just the entrance.
What lies beyond that—"how to improve operations"—is how much you can walk alongside them. This is everything.
The task of documenting payroll rules for each individual company.
The task of accumulating data for work regulations.
The task of organizing rules by industry.
Designing rules for handling irregularities.
It is all plain and simple.
There is not a shred of glamour.
But, surprisingly, things like this are often the real deal.
And the firms and companies that seek this 'real deal' are the ones that possess the mindset to properly utilize AI. They are the ones that will survive.
Ultimately, AI implementation is a matter of 'determination' on the part of the business owner and the firm.
Will you be satisfied with a flashy training session?
Or will you commit to gritty improvements?
We intend to commit to the latter.
Summary
The essence of AI implementation lies only in the gritty details.
Taking AI training because there are subsidies available.
That in itself is not bad.
But if you think that means you have 'implemented AI,' then nothing has actually begun.
True AI utilization is the polar opposite of glamour.
Codifying each company's rules, feeding them into the AI, and increasing accuracy by 1% at a time.
Only this accumulation of gritty work will...
Transform operations.
Raise the ceiling on revenue.
Build the future of the firm.
Turning 60% into 61%.
That is all it is.
But, surprisingly few firms can do 'just this'.
That is precisely why doing it now will set you apart.
If you do not do it now, you will be left behind.
Will your firm be satisfied with flashy AI training?
Or will you choose the gritty, real deal?
Thank you for reading to the end.
"I've had AI training, but I can't apply it to my actual work."
"I want to consult on which specific tasks I should start using AI for."
"Honestly, the subsidy training I took was useless, so I want to see what real AI implementation looks like! (We show you for free)"
If this sounds like you, please feel free to consult with us via the official LINE below.
Click here for LINE consultation!
↓
https://lin.ee/I73KUiN
We also share information on AI and DX on YouTube.
↓
https://youtube.com/channel/UC-f1ZI9MiME-GC0l1gwNf4w?si=WeyqGjBSvVXwSaDB