After two and a half years of startup and fully embracing AI, the company has actually become more "traditional."

Author: Digital Life Kazik

Today is a rather special time.

Because I've been running my company for two and a half years now, it's officially the Year of the Earth.

Over the past two and a half years, the company has weathered many storms and faced numerous life-or-death situations, but fortunately, it has survived.

Moreover, I'm doing alright, and the company is moving again soon.

Because of the growing number of people, although the team and business have been expanding very cautiously, there is still not enough room for everyone. So we just moved offices at the end of last year, but now we have to move to a new place again.

I was chatting with some friends the other day, and we were all a bit surprised. They said, "You've been using so much AI, why are you still hiring? Theoretically, shouldn't there be fewer and fewer people and smaller and smaller offices? Shouldn't there be a dozen or so people sitting in one room, each leading a dozen or so digital employees, and doing the work of hundreds of people in the past?"

This is AI! It sounds so advanced and sexy.

I said, "Bullshit."

The reality is that our AI penetration rate is now almost 100%. Almost everyone in the company is using various agents. All positions, including finance, HR, legal, business, brokerage, operations, etc., have used agents to create all sorts of processes and tools.

However, there are some things that cannot be accelerated by AI, and those things have become the most important things for our human employees.

The more advanced AI becomes, the more we use it, yet in everyone's and my understanding, we are becoming more and more "traditional".

This is a very interesting phenomenon. In interviews a few days ago and at dinners with some friends last night, everyone was quite concerned about the AI transformation of enterprises, so they talked to me about it frequently and asked me why I didn't write down these experiences.

So, on this occasion of my company's first anniversary, I'd like to take the liberty of sharing how, as a small, relatively traditional company with fewer than 40 people, we utilize AI within our organizational structure.

Of course, I'm not standing in the position of a so-called successful boss to tell everyone how to manage a company; we're far from that. We're not successful at all. I just feel that I can share some of my experience with you.

So, let's begin.

I. Full AI Integration

I've encountered countless companies where, when managers talk about adopting AI, their first reaction is often remarkably similar: to build a system, or find a person in charge, and establish an AI platform.

This looks professional and makes me feel very safe.

I completely understand this idea; we did the same thing a long time ago when I was doing some consulting work.

However, we later discovered that this method had its pitfalls.

It appears to be the most organized, but it is also the easiest to turn AI into another scheduling center in the company.

For example, if an HR professional encounters a problem with AI-based candidate screening, they should first raise their requirements.

Or, if a business wants to create an AI analysis tool for a client, they should first write a requirements document.

The problem occurs on the front lines of business, but the requirements have to go through several layers of people before reaching the person who actually makes the tool.

We all know that information is lost; with each transmission, much of the context is lost, and by the time something is finally made, the world has changed beyond recognition.

The most troublesome thing is that frontline staff will become increasingly adept at making demands, while those in the middle platform will become increasingly adept at using AI.

In the end, only a few people in the company truly possess creative abilities.

But I don't think organizations in the AI era should be like that. They should be organizations where everyone can use AI or create tools to solve problems, rather than just being limited to so-called AI platforms.

So later, whether in my own company or chatting with friends who own companies, I would give very restrained advice.

Companies with fewer than 100 original employees should be cautious about establishing an AI platform.

It's not that AI platforms are worthless. Large-scale systems, Lark, financial software, and unified permissions and security infrastructure certainly require specialized personnel to manage them.

However, this so-called middle platform should be as thin as possible.

It safeguards permissions, security, cost, data standards, and untouchable red lines, and then solves some server and token issues for everyone. The remaining issues that are closely related to business and change every day must be solved by the people closest to the problem using AI.

Whether you use an agent directly, create an agent-based tool, or do web scraping, scripting, or RPA, it doesn't matter, but you must solve it yourself.

Because only they can truly understand where the pain points lie, thus minimizing information leakage during communication.

So in our company, things can be quite brutal. Many things and processes that need optimization happen, but there's no specific position to help you develop solutions. The only ones who can solve these problems are yourself and the Agent. Whether it's Codex, Workbuddy, Claude Code, or any other platform, it doesn't matter, but the business personnel must solve them themselves.

This process will likely be very unpleasant at the beginning.

Some people can't describe it, some are afraid of code, and some spend a lot of time making a small, barely usable thing.

In many cases, the first time using an agent to solve a task is slower than doing it manually.

But this clumsiness is particularly important.

Believe me, AI isn't that complicated. When someone turns the problems they encounter at work into something functional for the first time, even if it's rough, their perception of their work will change.

Previously, he could only tolerate the procedures and the parts he was dissatisfied with.

Now, for the first time, he realized that he could change the process himself.

I later realized that the most valuable thing AI brings to ordinary employees may not just be efficiency, but also a long-lost sense of agency.

I strongly encourage everyone to use the Agent to optimize aspects of their work that they find unsatisfactory.

This is what I consider AI-driven, so I now appreciate the "thin middle platform, thick front-line" structure more.

II. Data Drives Everything

Of course, sending an agent to everyone will not automatically create a so-called AI-era organization.

Many companies have come to me to discuss AI, and as I listen to them, I always encounter the same question.

They have no data.

In the past, there were no full transcripts of meetings, client communications were scattered across different people's chat logs, and even contracts and quotations could not be found in a unified version. There was no post-project review after the project was completed.

Many of the most important experiences exist only in the minds of a few veteran employees.

This is utterly useless; it's not an agent at all.

For an organization, I now have even greater conviction that:

The agent isn't that important; the data is what matters most.

I personally believe in data emergence.

We at Virtual Media do MCN business, and we've signed hundreds of bloggers and connected with nearly a hundred brands.

We will try our best to preserve past content data, business data, and cooperation data as data assets within our organization.

Furthermore, those things that are difficult to fit into standard fields will be treated as features, labeled as unstructured tags, and then uniformly put into Lark multidimensional tables.

Not to mention that all our full meeting transcripts, documents, knowledge, and SOPs will also be added to the knowledge base.

Our management team's daily routine now consists of tagging many of the bloggers we've worked with.

Doing this is certainly tiring.

Many records don't show their value right now, and tags can't be perfectly applied the first time. Even a detail saved today might be useless for six months.

But I still think we should save it.

Because what a company truly owns is never just the money in its account, the equipment in its office, and the list of employees.

And all the data and context you've accumulated over the years.

Models can be bought, tokens can be bought, and Codex can be used by every company.

But believe me, the context a company leaves behind over countless specific days is something that can almost never be bought outside.

Therefore, the strategy I implemented internally was to retain everything that should be retained.

However, the so-called "store everything" does not mean simply throwing everything into a knowledge base or multidimensional table and calling it a day.

Because there are too many conflicts in the real world, old rules are often mixed with new rules, just like the code we generate with Vibe Coding. For example, two departments may have two different interpretations of the same metric, and outdated contract templates may still appear at the top of search results. Agents will not help organizations digest this chaos.

It will simply pick the one that looks most like the answer and then execute the error at AI speed.

Therefore, the data we store must have a source, a time frame, and a person in charge.

We must dare to discard old things, someone must adjudicate conflicting opinions, and decisions that truly affect money, contracts, and people must be regularly reviewed and the data cleaned.

It's important to understand that an agent won't automatically transform a chaotic company into a sophisticated and organized one.

The data governance behind this is the real hard, tiring, and dirty work, but it is also a sufficient and necessary condition for an organization to make a leap forward in AI.

III. Returning to the Essence of the Position

Once the data gradually became available and the agents were actually assigned to each role, another interesting thing happened.

People didn't become more like machines.

Instead, they are becoming more and more like humans.

For example, a business professional might spend 50% of their time organizing data, creating tables, researching information, and revising contracts, while the other 50% is spent visiting clients.

The agent has taken up most of the initial work, so it might only take 20% of the time. The remaining 80% can be used to meet with the client, understand their concerns, and make the solution more comprehensive and detailed.

Our agent is too.

In the past, they spent a lot of time collecting data, organizing records, and repeatedly confirming information. Now, these tasks can gradually be handed over to the agent, allowing them to talk to one more blogger, listen more carefully to the other person's recent status, and spend more time maintaining a relationship that is difficult to quantify.

HR, legal, content, and operations are all essentially the same.

AI excels at consuming tasks that can be described, repeated, and verified.

As these tasks are peeled away layer by layer, what is finally revealed is precisely what is most difficult to accelerate.

Sincere communication between people.

That's especially true for a company like ours, whose business is very traditional and almost entirely B2B.

For example, whether a customer is still willing to answer your call after a problem has occurred.

Whether a blogger is willing to dedicate the next few years to you, etc.

Codex can help you prepare materials, remember details, and review each communication.

But it can't take your time or communicate with the other person face-to-face for you.

That's why I've always said that in the AI era, trust and brand are more precious than diamonds.

This stuff is so hard to accumulate; it always requires a lot of time to manage.

When serving someone, you earn one point for fulfilling your promises. You earn another point for bravely resolving problems instead of running away. You earn yet another point for genuinely showing up to help when they are going through a difficult time, instead of just offering polite excuses.

It's incredibly slow.

Ironically, the faster AI becomes, the more expensive it is.

This is why the more agents we have, the more traditional the company becomes.

Business is more like the person in the past who had to constantly go out and sit with clients.

The agent is now more like the person who truly needed to get to know the blogger, understand the blogger, and accompany the blogger on a journey.

Managers are increasingly unable to hide behind reports; they must confront conflicts, make judgments, and bear the consequences.

AI has solved all efficiency-related problems, and the oldest relationship between people has been brought to the forefront.

IV. Time Saved

At this point, there's a rather harsh question.

That's where the time saved by the agent ultimately goes.

Many companies talk about AI, and their favorite calculation is to save working hours.

A process that used to take 4 hours now only takes 20 minutes. A position that used to serve 20 people can now serve 50. A plan that used to take two days to produce can now be drafted in half an hour.

These numbers are certainly important.

But if the time saved is eventually filled with more meetings, more reports, more approvals, and more forms that nobody ever reads, then the company just becomes busier.

If an employee who used to do 5 things a day is now required to do 20 things because of using an agent, and all 20 things need to be submitted immediately, he will not feel that technology has liberated him.

He would only feel that the whip had become faster.

I think this is an angle that a boss might easily overlook.

From the company's perspective, improved efficiency naturally seems like a good thing.

We will be excited, feeling that boundaries have been opened, and that we can now do many things that we were unable to do in the past.

I would do that myself.

If you have more capabilities, you want to do more projects. Similarly, if an agent can serve more bloggers, they want to sign more bloggers; if a business development team can process more information, they want to reach more clients; and if content production is faster, they want to cover more topics.

The skills saved up were quickly filled with new ambitions.

I think this is probably one of the reasons why we have more and more people and need to move offices.

AI will not necessarily make companies smaller.

What it amplifies first is probably the ambition of the company and its boss.

I wrote this down as a reminder to myself.

Ambitions are not wrong; a company should move forward.

But if every bit of efficiency ultimately translates into higher numbers, tighter schedules, and more work, then what we call AI-driven processes are nothing more than more overtime that is harder for the average employee to refuse.

So at the company, I'm increasingly less inclined to ask just how many hours the agent saved.

What I really want to ask is, who ultimately gets all those hours?

Give business time back to the client.

Give the blogger back the agent's time.

Give HR time back to the employees who truly need to be heard.

Give legal time back to difficult judgments, instead of mechanically revising the 27th edition format.

Give the content team back their time to experiences, curiosity, and things that are truly worth writing about.

It also gives back some of an ordinary person's time to himself.

He can learn something new, do his job better, and go home earlier to have a proper meal.

Managers often mistake employees for productivity.

But I think that people are not like batteries waiting to be drained dry by agents.

The greatest organizational value of AI, in my opinion, is that everyone should have a more genuine passion and joy for it.

Only in this way can you regain your curiosity about the world and build more trust when you interact with others.

V. What is a manager?

Then comes the cruelest part: the managers themselves.

In the past, a manager could easily prove their value by assigning tasks.

Hold meetings, urge progress, collect daily reports, process approvals, break down a task into 10 steps, and then check whether everyone has strictly followed those 10 steps.

However, as agents consume a large portion of the execution work, this management becomes increasingly awkward.

Employees bring their own Codex tools, which allows them to study materials, develop solutions, write scripts, and get a process that previously required multiple departments running.

At this point, what managers really need to do is not to take so many actions.

Instead, we should ask:

What exactly is the goal?

What absolutely cannot be wrong?

Which risks can a company take on?

To what extent should the result be considered perfect?

In the event of an accident, who makes the final decision?

If something goes wrong, who will take responsibility?

I find these things particularly difficult, and I have absolutely no way of making a nice PPT.

But these are the realities of management.

A manager who doesn't know how to write a Prompt still has time to learn.

If a manager can't clearly define goals, can't make judgments, and only pushes the blame onto subordinates when problems arise, then I think even the most powerful agent can't save him.

Sometimes I even feel that the biggest impact of AI on managers has nothing to do with whether the tools are new or not, but that in the AI era, many incompetences that were previously hidden in the processes are gradually exposed by AI.

In the past, we could say that we didn't have enough manpower, that the information wasn't organized properly, or that the execution below wasn't in place.

But now, what are you even talking about?

Once the agent has gathered all the information, provided solutions, and reduced execution costs, what's left to say if you still can't make that last, indecisive decision?

The same applies to me.

I can't expect everyone to be creative on their own initiative while simultaneously demanding that every detail be done according to my ideas.

I can't say I'm results-oriented while judging a person by their overtime hours, message response speed, and busy expression.

I can't just abandon unclear goals and then use the excuse that you need to learn how to use AI to shift management responsibility to employees.

These are actually signs of my extreme incompetence.

Therefore, how organizations will evolve in the AI era often depends on what the company was like before.

A company that doesn't trust people will use agents to conduct more detailed monitoring.

A boss who is used to being in control will use an agent to issue orders faster.

Only a company that is willing to respect people can truly turn agents into tools in the hands of ordinary employees.

AI will never automatically bring about advanced management.

It will only reveal the far side of the moon.

VI. What should a newcomer do?

This is one area where we're doing poorly, but it's also something we're trying to figure out. When we hire a new person, we don't know how to train them. For example, a newcomer in the content industry might start by finding information, revising titles, compiling case studies, and writing first drafts. A new business development professional would start by organizing client information, attending meetings, writing minutes, and revising proposals. A new legal professional would start by reading through the most basic contracts and terms.

These tasks are tedious and sometimes exhausting.

But in the past, a person often developed their senses through these seemingly mundane tasks.

But now, our Agent can give newcomers an 80-point answer in just a few minutes.

It's especially enjoyable in the short term.

A new employee might be able to produce something in the first week that would have taken six months to complete in the past. From the company's perspective, this reduces training costs, and the new employee feels like they've suddenly become much more capable.

But a sudden improvement in output does not necessarily mean that a person has truly grown.

If he hadn't gone through that foundational work, he would have had no chance to understand why the agent did what it did, and when the AI gave a wrong answer that looked very much like the correct one, he wouldn't even have a second thought.

What I fear most is not that newcomers don't know how to use AI.

What I fear is that he will only use AI and will never have the opportunity to develop his own judgment.

Because, frankly speaking, newcomers are always the weakest group in an organization.

Often, he felt quite lost, unsure of what questions to ask, which rules had expired, or whether he truly had the authority to make decisions when his leader told him to "do it himself."

An agent provides output, but may not necessarily give him the ability to bear the consequences.

If the company only looks at the results, he might be pushed along by that 80-point answer until he makes a huge mistake at a very important point, and then the company will ask him, "Why don't you understand this?"

So I'm also very worried, as this is very detrimental to the growth of newcomers.

For our company, I've been thinking about how to redesign a path for new employees to grow.

This path is certainly not meant to make newcomers continue doing meaningless hard work, nor is it meant to push people back to the manual labor era in order to hone their skills.

Instead, we spend as much time as possible letting him explain why the agent did what he did, let him compare different solutions, let him actually meet the client, let him see the consequences of the mistake, let him make a judgment when someone is backing him up, and let him sign his name on the final product.

I think what a newcomer needs is never just faster output.

He needs to make a few mistakes safely, needs to be corrected by someone who truly understands, and needs to know that one day he too can be the one to support others.

It's easy to give a newcomer an agent that can produce an 80-point answer.

Giving him a path to becoming a master is true management.

VII. Our Essence

So, going back to the beginning, that's why I said the more AI we have in our company, the more we resemble a traditional company.

Because once AI accelerates everything that can be accelerated, the truly difficult parts of an organization that cannot be accelerated by AI will finally emerge from beneath the surface.

Data can be organized automatically, but trust cannot.

Contracts can be generated quickly, but liability cannot.

Sincere communication between people is impossible. Just as customer information can certainly be analyzed, whether a customer is willing to still trust you when bad news comes is something you can't calculate purely through analysis.

These things are old, slow, and unsexy.

But the fate of a small company often hinges on these things.

For our company, we are not one of those companies with technological barriers. We are just a small company in this era that is trying to carve out a little bit of business, support our partners, and find our own way to survive.

We survived because we still have clients willing to entrust their budgets to us, bloggers willing to entrust their careers to us, industry leaders willing to work with us on offline events and even variety shows, and a group of colleagues willing to believe that we can continue to work together to "connect everything in the AI era".

So I'm increasingly convinced that the real value of data, agents, and automation in an organization shouldn't be about removing people from the middle.

We create so many blogger tags not so that one day we won't need to know the bloggers anymore.

It was precisely so that when his agent met him, he would better understand what he had experienced in the past and what he needed now.

Businesses ask agents to compile client information not so that they will never see the clients again.

It is precisely so that when you sit down in front of a client, you don't have to waste time on the homework that should have been done in advance.

We save meetings, documents, and SOPs not so that the organization can rely solely on the system.

It's to ensure that newcomers don't have to beg for help or start from scratch, avoiding the pitfalls that everyone else has already encountered.

AI is responsible for eliminating unnecessary losses in human-to-human interactions.

Then, give one person more time to truly get to know the other.

This is our true nature.

We are not some super company that can operate on its own in the cloud with just a few dozen employees.

We are a group of ordinary people who, with the help of the most advanced technology of this era, strive to do some very traditional things well.

Serving a client well, accompanying a blogger well, mentoring a newcomer well, keeping a promise, and making life better for those you work with.

So we are using AI more and more, and we are becoming more and more traditional, but I don't think this is a regression.

It's just that technology gradually peels away the outer layer of efficiency.

We have finally seen clearly what the most essential thing is between people.

In conclusion

Looking back on the past two and a half years as I write this, I'm actually quite emotional.

We have experienced many life-or-death crises and made many wrong decisions.

Although I'm still alive and about to move to a bigger office, I wouldn't dare say that we've found the right answer.

I don't know if full AI adoption is suitable for every company.

I don't know if a thin middle platform and a thick front-line structure is a good organizational approach for the AI era.

I have no idea whether the growth path we've designed for new employees will actually work out in the end.

But there's at least one thing I'm becoming more and more certain of now.

Organizational change in the AI era may superficially involve changes to tools, processes, data, and efficiency, but ultimately, it tests how a company treats its people.

Are you willing to hand over the power of creation to those on the front lines?

Would you be willing to give back the time saved by the agent to your clients, bloggers, employees, and life?

Are you willing to give a newcomer room to make mistakes and grow while they are still immature?

Are they also willing to step forward and take responsibility for the consequences themselves, instead of hiding behind processes and reports, when problems arise?

These are the things that determine what a company will ultimately become.

Finally, I would like to conclude this article with a quote from Antoine de Saint-Exupéry's *Far From Home*.

He said:

"The greatness of a profession may lie first and foremost in connecting people. There is only one true luxury in the world, and that is the relationship between people."

In the past, when I read such words, I might have thought they were quite romantic.

But after running the company for two and a half years, experiencing those times when the company might not survive, and watching the number of people around me gradually increase, I now feel that this statement is the truth.

Finally, we may also create some special segments in the future to share how our colleagues in various roles, such as finance, legal, HR, operations, and business, use AI. I personally find it quite inspiring to see how they use it.