A software company in a county town spent 700,000 yuan on an AI system; the vendor ran off and the boss was so furious he was hospitalized

[[**The owner of a software company with fewer than 100 employees in a fifth-tier city spent about 700,000 yuan on an AI system, but in return got accounts banned, payroll data leaked, employees lost, and the vendor running away. The so-called FDE transformation was actually standardized software plus high-priced tokens and agency operations.**

**Key points**

**1. The 700,000 yuan bought standard software, not FDE**

FDE requires long-term on-site customization, often with a team of seven or eight people and a cycle measured in years; localized deployment also requires bearing the operations and maintenance yourself. The owner got only four training sessions, then was repeatedly charged for high-priced tokens and agency operations, and the vendor eventually went missing.

**2. AI customer acquisition, HR, and advisory functions failed one after another**

**3. Using AI to replace veteran employees led to loss of business and customers**

The owner fired those who questioned him and required employees to train AI after work; in the most extreme week, nearly one-third of employees resigned; he used graduates with a monthly salary of 3,000 yuan plus 3 Agents to replace veteran employees, and the newcomers got stuck on details outside the knowledge base, losing a 100,000-yuan contract.

**4. AI implementation lacked the conditions and became a performance show**

The owner made AI use 60% of performance evaluations, which led to fabricated logs, and revenue and customers dropped sharply. Surveys show that 62% of enterprises have launched AI, but only 38% are still running it routinely after three full months, and nearly 30% stopped because there was no sustained on-site support; success requires the owner to understand the business, digitized data, and continuous use by business departments, none of which this company had. 2026-10-28 10:14 The first batch of county-town bosses who spent heavily to find FDE got badly burned AI Story Plan Quick Overview

Wang Hang witnessed a farce firsthand in a small fifth-tier city in Shaanxi. He works at a software company with fewer than 100 employees. Starting in August this year, the owner invested about 700,000 yuan in succession to bring in an AI system claimed to be able to acquire customers, provide customer service, do recruiting, and assist decision-making.

After deployment was completed, accounts were banned one after another, payroll data leaked, and veteran employees left one after another. In the end, even the vendor selling the system ran away. The owner was so angry that he was hospitalized.

On the surface, this is a story about paying for a lesson. What it punctures is one of the most common illusions right now: after FDE and localized deployment became buzzwords, many bosses in small places treated them as a miracle cure for cutting costs and improving efficiency, and wanted to get on the bandwagon without even understanding the basic concepts. In the end, the so-called deep business transformation was just installing a set of standardized software.

Below is Wang Hang's account.

The FDE hired at great expense ran away

After the National Day holiday ended, the boss, who had disappeared for more than ten days, finally showed up. As soon as he entered the company, he organized an all-hands meeting.

At the meeting, the boss announced two things: one, he had found the AI deployment company that had run away and recovered part of the money; the other, he fired the company's vice president on the grounds that the vice president did not understand AI well enough and failed to identify risks in time during the company's AI transformation.

Everyone looked at each other, not knowing how to react.

My name is Wang Hang, and I am 37 this year. Four months ago, I left a big tech company, ended more than ten years of life as a Beijing migrant worker, and returned to a fifth-tier city in Shaanxi.

After going home, I joined a fairly well-known local software company. The company mainly serves government and enterprise clients and is doing well; it is said to have more than 20 stable customers each month and a net profit of more than 400,000. The company's owner is over sixty and is responsible for showing his face outside, while daily operations are handled by the vice president.

The boss can barely even operate a computer, but that did not stop him from being the person in the whole company most anxious about AI. He travels around the country to exhibitions and often scrolls through peers' video accounts; whenever he sees content about OPC, AI employees, Agents, AI marketing, and the like, he forwards it to the work group. For a while, he forwarded things three or four times a day.

In this small fifth-tier city, such bosses are everywhere. We have quite a few AI+OPC communities locally, with bosses from all industries inside: those running supermarkets, shops, and dental clinics. Even when industry associations hold meetings, the topic of AI is unavoidable.

With the idea of transformation in place, the next thing missing was someone to implement it. The boss usually also forwarded articles about FDE and localized deployment to the group, but when it came to actually finding a service provider, he was somewhat deterred by the big companies in Beijing, Shanghai, Guangzhou, and Shenzhen. He often said at meetings that those companies were indeed technically strong, but too far away, which gave him no sense of security.

Our boss firmly believes that "the bold die of overeating and the timid starve to death," and spent 500,000 yuan to buy the full system.

For ordinary reimbursements of office supplies, he would hold them up for weeks; for business trips, the hotel standard was only 300 yuan. On the matter of AI, all his previous investment had been no more than forwarding a few short videos with FDE and OPC in the titles to the work group.

In September, he held an all-hands meeting and announced that the 500,000 yuan had already been paid, and that from then on all energy would be devoted to this system.

Everyone in the audience was stunned. He had always been very cautious in doing things. When cooperating with others, he would require us to first find out the other party's background and hold several rounds of collective meetings before considering signing; whenever there were changes in the company, he would also release word in advance. Even for buying a printer, he would only pay after everyone had tried it out.

The boss was so angry that he was hospitalized

After four brief training sessions, that legendary all-purpose system was handed out to everyone.

Previously, the company's accounts were all specially operated by colleagues in the new media department, with content including screen-recorded operations and talking-head videos shot for the boss and management. The results had always been good, bringing in dozens of valid customers every month.

The editor had resigned just one week earlier when the accident happened. When colleagues in the new media department logged into the accounts, they found that several accounts on major platforms had been banned. They asked platform customer service and learned that the content published every day by these accounts was generated by the same set of prompts, was highly repetitive, and had been judged by the platform as low quality.

Seeing the boss's momentum weaken, the other side pressed the advantage and smoothly pivoted to pitching agency operation.

Agency operation is not cheap: two thousand yuan per account per month, and that does not include Token fees. With this system, Tokens also had to be bought from that company, at more than five times the market price.

Normally I spend ten yuan myself to buy a key on DeepSeek, and it is enough for the whole company for a week. By that calculation, the whole company's monthly cost is less than five thousand. Buying from the AI deployment company, for the same usage, costs more than fifty thousand a month.

The other side promised that agency operation would guarantee no account bans, and also threw in a membership for the boss, allowing him to attend their company's dinner. The boss paid again. Adding the money for buying the system, roughly 700,000 yuan was spent in total.

After the money was paid, the boss came back and held an all-hands meeting, requiring everyone to make full use of it. To prevent duplication, everyone had to make videos with AI themselves. We were not familiar with it to begin with, so we could only revise prompts again and again and generate again and again, and Token costs made things even worse.

At the end of the month, the boss saw the bill, and his vision went black; he decided to stop topping up Tokens. But once the Tokens stopped, the AI customer service that had been running normally also stopped, and customer complaints came one after another.

The boss went over angrily again. This time, the other side still had the same line: insufficient computing power; once computing power catches up, the price will naturally fall.

The digital human was made in just one afternoon. The boss posted a talking-head video generated from his own image, and two hours later he got one comment: "Looks like facial paralysis." After that, he never mentioned digital humans again.

Other management staff were still using their own digital humans. It really did save time; as for the effect, no one cared.

The hole in AI customer acquisition had not yet been filled when AI HR had another accident. Just after September salaries were paid, a colleague next to me suddenly turned his computer toward me, and on the screen was written in plain sight: supervisor salary 35,400.67 yuan.

That week, he interviewed more than fifty people, and almost everyone who came out looked bewildered.

Only one editor who understood AI passed, but refused the offer before joining. I asked her privately why. She said that before signing the contract, she had AI take a look, and AI pointed out a pile of problems in the contract. In the screenshot she sent me, AI's summary was: "This is not an employment contract; this is an unequal treaty."

At the meeting, he demonstrated on the spot, typing into the dialog box, "Predict the company's development over the next six months." Half a minute later, a line popped up on the screen: the company is predicted to go bankrupt within six months.

700,000 yuan, and all you get is a sound

The even more dramatic part was still to come.

Whether because he felt he had lost face or sensed he had been scammed, the boss also disappeared for a while. When he reappeared, it was at that all-hands meeting.

I later thought about it for a long time: what exactly did that 700,000 yuan buy?

The boss had reposted posts about FDE, but what he bought was not that at all.

FDE stands for Forward Deployed Engineer. It refers to engineers stationed long-term at the client site, going deep into the other party's business processes, understanding requirements while customizing development and continuously iterating.

Put plainly, it means accompanying the enterprise in getting its processes running again. A system truly landing often takes years. And it also requires a delivery team of at least seven or eight people: a project manager to coordinate, a solutions engineer to break down the business and translate it for the technical side, a data engineer to connect the data, back-end and algorithm engineers to implement it, and someone stationed at the client to train, accompany, and monitor results. On the client side, there also has to be someone who truly understands the business and cooperates throughout.

This is a completely different business from selling a standard software package, training four times, and then collecting money for Tokens and agency operation.

Local deployment means installing the system and data on the enterprise's own servers, so the data does not leave the premises. It was originally an option for sensitive industries such as government affairs and finance, at the cost of bearing all the hardware, operations and maintenance, and upgrades oneself. For a small company without even dedicated IT staff, the day the system goes live is often the beginning of no one maintaining it.

What the boss bought with that 700,000 yuan was software and anxiety. After four trainings, the other side's work was done; the rest of the money was to be collected through Tokens and agency operation. One deal, money collected three times, and not one of those times was money for getting the process running for the client.

For 700,000 yuan, selling standard software is huge profit, but doing real transformation is not enough. A team of seven or eight people, plus acceptance, would mean the other side loses money. So that company never intended to do FDE from the start. Rather than saying it ran away, it is more accurate to say it was selling a different product all along.

A 2025 survey of 1,200 large-scale manufacturing enterprises by the China Electronics Standardization Institute shows that 62% of enterprises had already purchased or launched AI systems, but only 38% were still running them routinely three months after launch; among the reasons for stopping, nearly 30% were "no sustained on-site accompaniment."

Later, I read online many accounts of bosses whose AI transformation succeeded, and several conditions were indispensable: the boss must understand the business and be willing to sink into the details; the company's data must first be digitized, and if the data has not been incorporated into digital systems, rashly putting all employees on AI office work will instead cause big problems; the system must ultimately be handed over to the business departments and used every day, and if AI is treated only as a simple tool, it is hard to see real results and it sometimes becomes a burden on employees.

Our boss met none of these.

It can still help us make some PPTs, retrieve company materials, and help new employees get familiar with the business. It even helped me make money once.

The boss's explanation was: today's AI is like an 8-year-old child—imperfect, so you have to raise it; when it grows to 18, it will naturally be able to handle things. At first everyone just listened and let it go. It wasn't until he said, "AI will one day replace all of you," that we realized the child we trained with overtime every day would grow up, and its first job would be to take over from us.

In the most extreme week, nearly a third of the company left, and he was not worried at all. He believed their experience had already been "fully uploaded to the system." He even found a trendy word for it: distillation.

Later, he hired two graduates with a monthly salary of 3,000 each, gave each of them three Agents, and was convinced this was enough to replace veteran employees earning 8,000 a month. At first it went smoothly; when the newcomers didn't understand something, they asked AI for company materials from the past few years, and within a few days they could dig up most of what they needed.

The problem soon appeared. Less than a month after the newcomers started, the boss sent one of them to negotiate the final step with a client we had maintained for four months.

The beginning of the talks went well; the reports and PPTs generated by AI satisfied the client. Then the other side began pressing for details that were not in the knowledge base, and the newcomer started to falter, frequently looking down at the screen. The client turned around and gave this 100,000-yuan long-term contract to a competitor.

When the policy comes from above, countermeasures come from below. To fill the performance gap, we started adding AI to our daily work reports. We renamed copywriting "AI-assisted content production," data organization "AI data cleaning," and proposal writing "AI completing full customer lifecycle planning from 0 to 1."

Later, I had a meal with the vice president who had been fired. Talking about this whole chain of events, he said that in the boss's eyes, anyone who says the system is hard to use is rebelling against him personally.

"More than losses, the boss fears not being able to control people," he said. AI industry signal channel: Frontier Technology. This content is authorized by the author for publication; the views represent only the author and do not represent Huxiu's position. If you have objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is in the Huxiu APP[ ]]]