Short answer: it can be fully produced, but it should not be fully published. We once handed publishing to a tool, ran into two glitches, and pulled that step back to a human. Today our pipeline is AI drafts, human review, human publishing.
Here is what went wrong first.
We did let a tool publish for us
A while ago, someone on the team installed an open-source skill that runs the whole "write, illustrate, publish" loop for a social post from a single topic prompt, and with cookies configured it can publish in one command (source: Qingyu Learns AI / 青鱼学AI, 2026-10-09). Convenient. So we ran it for a few days.
The problems showed up fast. In one case a post carried an outdated routing detail and the tool published it anyway. In another, the same post went out two days in a row because the script fired on schedule, and the account's rhythm fell apart. Neither was a major incident, but both pointed to one fact: the tool does not know what "publishing" means.
We had assumed the first version was only a draft
Earlier we had also tested the other direction: using Doubao Work to turn an article into a short video (source: Jianshi / 见实, 2026-10-09). After the seven steps, our own conclusion was that the first version is only a draft and still needs human review.
We kept that sentence, but at the time we read it as a note about production. We did not apply it to publishing.
Then the platform rule changed, and we saw the point.
On 8 October 2026, the WeChat Channels "Announcement on Regulating Traffic-Diversion Behavior" took effect (source: Jianshi / 见实, 2026-10-09). For the first time it allows creators to leave contact details in the account bio, while every other placement remains prohibited. Bulk direct-message diversion, second-hop diversion through Official Account storefronts, and homophone or split-character tricks were all added to the violation list. Xiaohongshu has also begun cracking down on AI-managed operation accounts through a governance announcement (source: Qingyu Learns AI / 青鱼学AI, 2026-10-09).
One spot loosened, several tightened. And the tightening keeps landing on the same question: who is actually publishing.
The fix: separate the pipeline from publishing
After the glitches, we split the workflow into two sections.
The pipeline stays with AI. Drafting, illustration, clipping, batch layout, turning a long article into a video script — all the repetitive execution goes to the tool. The time saved goes back into topic selection and customer conversations.
Publishing stays with people. Whether to publish, when, and to which account are human calls. Scripts do not auto-fire on a schedule.
In between sits a review step. Before anything goes live, a person reads it line by line, checking numbers, routing and quoting language without exception.
And we keep a trail. Source material, drafts and a publishing log are retained, so it is traceable who wrote what, how many revisions it went through, and when it went out.
This is what we call an auditable process. For a freight content operations team, it is not about speed. It is about whether the content has a responsible owner behind it.
Why this line matters for freight lead generation
In freight forwarding lead generation, what a customer remembers is the account that keeps explaining professional problems clearly and consistently.
Once publishing is outsourced to a script, the rhythm and the messaging are no longer controlled by a person, and the trust built up over time can be broken by a single mis-post. So freight forwarding AI adoption is not about how many tools you use. It is about which part of the workflow you use them on.
Shanghai Naili has run freight forwarding marketing operations for 10 years, and our conclusion is one line: AI can speed up the pipeline, but the act of publishing stays with people. That is also what we deliver: a process with a paper trail, auditable and self-verifiable, without promising rankings or results.
FAQ
Q1: Can freight forwarding content really be produced fully automatically?
A: The production part can. Drafting, illustration, clipping and layout are largely handled by tools now. But the output is a first draft and needs human review before use (source: Jianshi / 见实, 2026-10-09).
Q2: Why not publish fully automatically?
A: Because publishing is an act of account identity. A tool does not know what "going live" means, and cannot tell whether routing or pricing in a post is out of date. The WeChat Channels rule, effective 8 October 2026 (source: Jianshi / 见实, 2026-10-09), targets disguised diversion, and Xiaohongshu is cracking down on AI-managed operation accounts (source: Qingyu Learns AI / 青鱼学AI, 2026-10-09). Fully automatic publishing is getting riskier.
Q3: How do you split the work in your own pipeline?
A: AI drafts, people review, people publish, and we keep the source library, drafts and publishing log. The publishing action is never handed to a script.
Q4: Does this make the process too slow?
A: The pipeline speed is unaffected. Only the final human review adds time. In return, content stays traceable and messaging stays controlled. For anyone doing freight forwarding marketing and content operations over the long run, that trade is worth it.